1 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
10 2018-06-08 Martin Liska <mliska@suse.cz>
12 * tree-cfg.h (debug_function): Fix argument type to match
15 2018-06-08 Martin Liska <mliska@suse.cz>
17 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
18 Remove usage of MPX-related (and removed) fields.
19 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
21 2018-06-08 David Malcolm <dmalcolm@redhat.com>
23 * cfg.c (debug): Use TDF_NONE rather than 0.
24 * cfghooks.c (debug): Likewise.
25 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
26 (struct dump_option_value_info): Convert to...
27 (struct kv_pair): ...this template type.
28 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
30 (optinfo_verbosity_options): Likewise.
31 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
33 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
34 than int for "optgroup_flags" param.
35 (dump_generic_expr_loc): Use dump_flags_t rather than int for
38 (dump_finish): Use TDF_NONE rather than 0.
39 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
40 rather than int for "optgroup_flags" param. Use TDF_NONE rather
41 than 0. Update for change to option_ptr.
42 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
43 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
44 0. Update for changes to optinfo_verbosity_options and
46 (opt_info_switch_p): Convert optgroup_flags from int to
48 (dump_basic_block): Use dump_flags_t rather than int
49 for "dump_kind" param.
50 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
51 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
52 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
53 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
54 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
55 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
56 TDF_NONE): Convert from macros to...
57 (enum dump_flag): ...this new enum.
58 (dump_flags_t): Update to use enum.
59 (operator|, operator&, operator~, operator|=, operator&=):
60 Implement for dump_flags_t.
61 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
62 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
63 Convert from macros to...
64 (enum optgroup_flag): ...this new enum.
65 (optgroup_flags_t): New typedef.
66 (operator|, operator|=): Implement for optgroup_flags_t.
67 (struct dump_file_info): Convert field "alt_flags" to
68 dump_flags_t. Convert field "optgroup_flags" to
70 (dump_basic_block): Use dump_flags_t rather than int for param.
71 (dump_generic_expr_loc): Likewise.
73 (dump_register): Convert param "optgroup_flags" to
75 (opt_info_enable_passes): Likewise.
76 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
78 * gimple-pretty-print.c (debug): Likewise.
79 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
80 (merged_store_group::apply_stores): Likewise.
81 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
82 * gimple.c (verify_gimple_pp): Likewise.
83 * graphite-poly.c (print_pbb_body): Likewise.
84 * passes.c (pass_manager::register_one_dump_file): Convert
85 local "optgroup_flags" to optgroup_flags_t.
86 * print-tree.c (print_node): Use TDF_NONE rather than 0.
88 (debug_body): Likewise.
89 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
91 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
93 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
94 (convert_mult_to_fma): Likewise.
95 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
96 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
97 * tree-vect-data-refs.c (dump_lower_bound): Convert param
98 "dump_kind" to dump_flags_t.
100 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
102 * config/rs6000/rs6000.c (min, max): Delete.
104 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
106 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
107 -mabi=spe and -mabi=no-spe.
109 2018-06-08 Martin Liska <mliska@suse.cz>
111 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
112 where we expect an existing summary.
114 2018-06-08 Martin Liska <mliska@suse.cz>
116 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
117 * ipa-inline.h (estimate_edge_growth): Likewise.
119 2018-06-08 Martin Liska <mliska@suse.cz>
121 * cgraph.c (function_version_hasher::hash): Use
122 cgraph_node::get_uid ().
123 (function_version_hasher::equal):
124 * cgraph.h (cgraph_node::get_uid): New method.
125 * ipa-inline.c (update_caller_keys): Use
126 cgraph_node::get_uid ().
127 (update_callee_keys): Likewise.
128 * ipa-utils.c (searchc): Likewise.
129 (ipa_reduced_postorder): Likewise.
130 * lto-cgraph.c (input_node): Likewise.
131 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
132 * symbol-summary.h (symtab_insertion): Likewise.
133 (symtab_removal): Likewise.
134 (symtab_duplication): Likewise.
135 * tree-pretty-print.c (dump_function_header): Likewise.
136 * tree-sra.c (convert_callers_for_node): Likewise.
138 2018-06-08 Martin Liska <mliska@suse.cz>
140 * cgraph.c (symbol_table::create_edge): Always assign a new
142 (symbol_table::free_edge): Do not recycle numbers.
143 * cgraph.h (cgraph_edge::get): New method.
144 * symbol-summary.h (symtab_removal): Use it.
145 (symtab_duplication): Likewise.
146 (call_summary::hashable_uid): Remove.
148 2018-06-08 Martin Liska <mliska@suse.cz>
150 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
151 (initialize_growth_caches): Remove.
152 (free_growth_caches): Likewise.
153 (do_estimate_edge_time): Use edge_growth_cache.
154 (do_estimate_edge_size): Likewise.
155 (do_estimate_edge_hints): Likewise.
156 * ipa-inline.c (reset_edge_caches): Likewise.
157 (recursive_inlining): Likewise.
158 (inline_small_functions): Likewise.
159 * ipa-inline.h (initialize_growth_caches): Remove.
160 (estimate_edge_size): Likewise.
161 (estimate_edge_time): Likewise.
162 (estimate_edge_hints): Likewise.
163 (reset_edge_growth_cache): Likewise.
164 * symbol-summary.h (call_summary::remove): New method.
166 2018-06-08 Martin Liska <mliska@suse.cz>
168 * ipa-cp.c (class edge_clone_summary): New summary.
169 (grow_edge_clone_vectors): Remove.
170 (ipcp_edge_duplication_hook): Remove.
171 (class edge_clone_summary_t): New call_summary class.
172 (ipcp_edge_removal_hook): Remove.
173 (edge_clone_summary_t::duplicate): New function.
174 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
175 (create_specialized_node): Likewise.
176 (ipcp_driver): Initialize edge_clone_summaries and do not
179 2018-06-08 Martin Liska <mliska@suse.cz>
181 * symbol-summary.h (get): New function.
182 (call_summary::m_initialize_when_cloning): New class member.
184 2018-06-08 Martin Liska <mliska@suse.cz>
186 * cgraph.c (cgraph_node::remove): Do not recycle uid.
187 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
188 (symbol_table::allocate_cgraph_symbol): Do not set uid.
189 * passes.c (uid_hash_t): Record removed_nodes by their uids.
190 (remove_cgraph_node_from_order): Use the removed_nodes set.
191 (do_per_function_toporder): Likwise.
192 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
193 instead of summary_uid.
194 (symtab_removal): Likewise.
195 (symtab_duplication): Likewise.
197 2018-06-08 Martin Liska <mliska@suse.cz>
199 * ipa-cp.c (ipcp_store_bits_results): Use
200 ipcp_transformation_sum.
201 (ipcp_store_vr_results): Likewise.
202 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
204 (ipcp_transformation_initialize): ... this.
205 (ipa_set_node_agg_value_chain):
206 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
207 (write_ipcp_transformation_info): Likewise.
208 (read_ipcp_transformation_info): Likewise.
209 (ipcp_update_bits): Likewise.
210 (ipcp_update_vr): Likewise.
211 (ipcp_transform_function): Likewise.
212 * ipa-prop.h: Rename ipcp_transformation_summary to
214 (class ipcp_transformation_t): New function summary.
215 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
216 (ipa_get_agg_replacements_for_node): Likewise.
218 2018-06-08 Martin Liska <mliska@suse.cz>
220 * ipa-pure-const.c (struct funct_state_d): Do it class instead
222 (class funct_state_summary_t): New function_summary class.
223 (has_function_state): Remove.
224 (get_function_state): Likewise.
225 (set_function_state): Likewise.
226 (add_new_function): Likewise.
227 (funct_state_summary_t::insert): New function.
228 (duplicate_node_data): Remove.
229 (remove_node_data): Remove.
230 (funct_state_summary_t::duplicate): New function.
231 (register_hooks): Create new funct_state_summaries.
232 (pure_const_generate_summary): Use it.
233 (pure_const_write_summary): Likewise.
234 (pure_const_read_summary): Likewise.
235 (propagate_pure_const): Likewise.
236 (propagate_nothrow): Likewise.
237 (dump_malloc_lattice): Likewise.
238 (propagate_malloc): Likewise.
239 (execute): Do not register hooks, just remove summary
241 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
244 2018-06-08 Martin Liska <mliska@suse.cz>
246 * ipa-reference.c (remove_node_data): Remove.
247 (duplicate_node_data): Likewise.
248 (class ipa_ref_var_info_summary_t): New class.
249 (class ipa_ref_opt_summary_t): Likewise.
250 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
251 (get_reference_optimization_summary): Use
252 ipa_ref_opt_sum_summaries.
253 (set_reference_vars_info): Remove.
254 (set_reference_optimization_summary): Likewise.
255 (ipa_init): Create summaries.
256 (init_function_info): Use function summary.
257 (ipa_ref_opt_summary_t::duplicate): New function.
258 (ipa_ref_opt_summary_t::remove): New function.
259 (get_read_write_all_from_node): Fix GNU coding style.
260 (propagate): Use function summary.
261 (write_node_summary_p): Fix GNU coding style.
262 (stream_out_bitmap): Likewise.
263 (ipa_reference_read_optimization_summary): Use function summary.
264 (ipa_reference_c_finalize): Do not release hooks.
266 2018-06-08 Martin Liska <mliska@suse.cz>
268 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
269 (analyze_function_body): Extract multiple calls of get_create.
270 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
271 * ipa-inline.c (recursive_inlining): Use ::get method.
272 * ipa-inline.h (estimate_edge_growth): Likewise.
274 2018-06-08 Martin Liska <mliska@suse.cz>
276 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
278 (hsa_function_summary::hsa_function_summary): Use the new enum
280 (hsa_gpu_implementation_p): Use hsa_summaries::get.
281 * hsa-gen.c (hsa_get_host_function): Likewise.
282 (get_brig_function_name): Likewise.
283 * ipa-hsa.c (process_hsa_functions): Likewise.
284 (ipa_hsa_write_summary): Likewise.
285 * symbol-summary.h (symtab_duplication): Use ::get function/
288 2018-06-08 Martin Liska <mliska@suse.cz>
290 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
292 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
293 (hsa_register_kernel): Likewise.
294 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
295 * hsa-gen.c (hsa_get_host_function): Likewise.
296 (get_brig_function_name): Likewise.
297 (generate_hsa): Likewise.
298 (pass_gen_hsail::execute): Likewise.
299 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
300 (devirtualization_time_bonus): Likewise.
301 (ipcp_propagate_stage): Likewise.
302 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
303 (edge_set_predicate): Likewise.
304 (evaluate_conditions_for_known_args): Likewise.
305 (evaluate_properties_for_edge): Likewise.
306 (ipa_fn_summary::reset): Likewise.
307 (ipa_fn_summary_t::duplicate): Likewise.
308 (dump_ipa_call_summary): Likewise.
309 (ipa_dump_fn_summary): Likewise.
310 (analyze_function_body): Likewise.
311 (compute_fn_summary): Likewise.
312 (estimate_edge_devirt_benefit): Likewise.
313 (estimate_edge_size_and_time): Likewise.
314 (estimate_calls_size_and_time): Likewise.
315 (estimate_node_size_and_time): Likewise.
316 (inline_update_callee_summaries): Likewise.
317 (remap_edge_change_prob): Likewise.
318 (remap_edge_summaries): Likewise.
319 (ipa_merge_fn_summary_after_inlining): Likewise.
320 (ipa_update_overall_fn_summary): Likewise.
321 (read_ipa_call_summary): Likewise.
322 (inline_read_section): Likewise.
323 (write_ipa_call_summary): Likewise.
324 (ipa_fn_summary_write): Likewise.
325 (ipa_free_fn_summary): Likewise.
326 * ipa-hsa.c (process_hsa_functions): Likewise.
327 (ipa_hsa_write_summary): Likewise.
328 (ipa_hsa_read_section): Likewise.
329 * ipa-icf.c (sem_function::merge): Likewise.
330 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
331 (do_estimate_edge_time): Likewise.
332 (estimate_size_after_inlining): Likewise.
333 (estimate_growth): Likewise.
334 (growth_likely_positive): Likewise.
335 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
336 (inline_call): Likewise.
337 * ipa-inline.c (caller_growth_limits): Likewise.
338 (can_inline_edge_p): Likewise.
339 (can_inline_edge_by_limits_p): Likewise.
340 (compute_uninlined_call_time): Likewise.
341 (compute_inlined_call_time): Likewise.
342 (want_inline_small_function_p): Likewise.
343 (edge_badness): Likewise.
344 (update_caller_keys): Likewise.
345 (update_callee_keys): Likewise.
346 (recursive_inlining): Likewise.
347 (inline_small_functions): Likewise.
348 (inline_to_all_callers_1): Likewise.
349 (dump_overall_stats): Likewise.
350 (early_inline_small_functions): Likewise.
351 (early_inliner): Likewise.
352 * ipa-inline.h (estimate_edge_growth): Likewise.
353 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
354 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
355 * ipa-prop.h (IPA_NODE_REF): Likewise.
356 (IPA_EDGE_REF): Likewise.
357 * ipa-pure-const.c (malloc_candidate_p): Likewise.
358 (propagate_malloc): Likewise.
359 * ipa-split.c (execute_split_functions): Likewise.
360 * symbol-summary.h: Rename get to get_create.
362 (get_create): Likewise.
363 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
365 2018-06-08 Martin Liska <mliska@suse.cz>
367 * symbol-summary.h (release): Move definition out of class
369 (symtab_removal): Likewise.
370 (symtab_duplication): Likewise.
372 2018-06-08 Martin Liska <mliska@suse.cz>
374 * symbol-summary.h (function_summary): Move constructor
375 implementation out of class declaration.
377 (symtab_insertion): Likewise.
378 (symtab_removal): Likewise.
379 (symtab_duplication): Likewise.
382 2018-06-08 Martin Liska <mliska@suse.cz>
384 * Makefile.in: Remove support for MPX (macros, related functions,
385 fields in cgraph_node, ...).
386 * builtin-types.def (BT_BND): Likewise.
387 (BT_FN_BND_CONST_PTR): Likewise.
388 (BT_FN_CONST_PTR_BND): Likewise.
389 (BT_FN_VOID_PTR_BND): Likewise.
390 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
391 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
392 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
393 (expand_builtin_mempcpy_with_bounds): Likewise.
394 (expand_builtin_memset_with_bounds): Likewise.
395 (expand_builtin_memset_args): Likewise.
396 (std_expand_builtin_va_start): Likewise.
397 (expand_builtin): Likewise.
398 (expand_builtin_with_bounds): Likewise.
399 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
400 (DEF_LIB_BUILTIN_CHKP): Likewise.
401 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
402 (DEF_CHKP_BUILTIN): Likewise.
403 (BUILT_IN_MEMCPY): Likewise.
404 (BUILT_IN_MEMMOVE): Likewise.
405 (BUILT_IN_MEMPCPY): Likewise.
406 (BUILT_IN_MEMSET): Likewise.
407 (BUILT_IN_STPCPY): Likewise.
408 (BUILT_IN_STRCAT): Likewise.
409 (BUILT_IN_STRCHR): Likewise.
410 (BUILT_IN_STRCPY): Likewise.
411 (BUILT_IN_STRLEN): Likewise.
412 (BUILT_IN_MEMCPY_CHK): Likewise.
413 (BUILT_IN_MEMMOVE_CHK): Likewise.
414 (BUILT_IN_MEMPCPY_CHK): Likewise.
415 (BUILT_IN_MEMSET_CHK): Likewise.
416 (BUILT_IN_STPCPY_CHK): Likewise.
417 (BUILT_IN_STRCAT_CHK): Likewise.
418 (BUILT_IN_STRCPY_CHK): Likewise.
419 * calls.c (store_bounds): Likewise.
420 (emit_call_1): Likewise.
421 (special_function_p): Likewise.
422 (maybe_warn_nonstring_arg): Likewise.
423 (initialize_argument_information): Likewise.
424 (finalize_must_preallocate): Likewise.
425 (compute_argument_addresses): Likewise.
426 (expand_call): Likewise.
427 * cfgexpand.c (expand_call_stmt): Likewise.
428 (expand_return): Likewise.
429 (expand_gimple_stmt_1): Likewise.
430 (pass_expand::execute): Likewise.
431 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
432 (cgraph_node::remove): Likewise.
433 (cgraph_node::dump): Likewise.
434 (cgraph_node::verify_node): Likewise.
435 * cgraph.h (chkp_function_instrumented_p): Likewise.
436 (symtab_node::get_alias_target): Likewise.
437 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
438 (cgraph_local_p): Likewise.
439 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
440 (cgraph_edge::rebuild_references): Likewise.
441 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
442 (walk_polymorphic_call_targets): Likewise.
443 (cgraph_node::expand_thunk): Likewise.
444 (symbol_table::output_weakrefs): Likewise.
445 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
446 (ix86_handle_option): Likewise.
447 * config/i386/constraints.md: Likewise.
448 * config/i386/i386-builtin-types.def (BND): Likewise.
452 * config/i386/i386-builtin.def (BDESC_END): Likewise.
453 (BDESC_FIRST): Likewise.
455 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
456 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
457 * config/i386/i386.c (enum reg_class): Likewise.
458 (ix86_target_string): Likewise.
459 (ix86_option_override_internal): Likewise.
460 (ix86_conditional_register_usage): Likewise.
461 (ix86_valid_target_attribute_inner_p): Likewise.
462 (ix86_set_indirect_branch_type): Likewise.
463 (ix86_set_current_function): Likewise.
464 (ix86_function_arg_regno_p): Likewise.
465 (init_cumulative_args): Likewise.
466 (ix86_function_arg_advance): Likewise.
467 (ix86_function_arg): Likewise.
468 (ix86_pass_by_reference): Likewise.
469 (ix86_function_value_regno_p): Likewise.
470 (ix86_function_value_1): Likewise.
471 (ix86_function_value_bounds): Likewise.
472 (ix86_return_in_memory): Likewise.
473 (ix86_setup_incoming_vararg_bounds): Likewise.
474 (ix86_va_start): Likewise.
475 (indirect_thunk_need_prefix): Likewise.
476 (print_reg): Likewise.
477 (ix86_print_operand): Likewise.
478 (ix86_expand_call): Likewise.
479 (ix86_output_function_return): Likewise.
480 (reg_encoded_number): Likewise.
481 (BDESC_VERIFYS): Likewise.
482 (ix86_init_mpx_builtins): Likewise.
483 (ix86_init_builtins): Likewise.
484 (ix86_emit_cmove): Likewise.
485 (ix86_emit_move_max): Likewise.
486 (ix86_expand_builtin): Likewise.
487 (ix86_builtin_mpx_function): Likewise.
488 (ix86_get_arg_address_for_bt): Likewise.
489 (ix86_load_bounds): Likewise.
490 (ix86_store_bounds): Likewise.
491 (ix86_load_returned_bounds): Likewise.
492 (ix86_store_returned_bounds): Likewise.
493 (ix86_class_likely_spilled_p): Likewise.
494 (ix86_hard_regno_mode_ok): Likewise.
495 (x86_order_regs_for_local_alloc): Likewise.
496 (ix86_mitigate_rop): Likewise.
497 (ix86_bnd_prefixed_insn_p): Likewise.
498 (ix86_mpx_bound_mode): Likewise.
499 (ix86_make_bounds_constant): Likewise.
500 (ix86_initialize_bounds): Likewise.
501 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
502 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
503 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
504 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
505 (TARGET_CHKP_BOUND_MODE): Likewise.
506 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
507 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
508 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
509 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
510 * config/i386/i386.h (TARGET_MPX): Likewise.
511 (TARGET_MPX_P): Likewise.
512 (VALID_BND_REG_MODE): Likewise.
513 (FIRST_BND_REG): Likewise.
514 (LAST_BND_REG): Likewise.
515 (enum reg_class): Likewise.
516 (BND_REG_P): Likewise.
517 (BND_REGNO_P): Likewise.
519 (ADJUST_INSN_LENGTH): Likewise.
520 * config/i386/i386.md: Likewise.
521 * config/i386/i386.opt: Likewise.
522 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
524 (LINK_MPX): Likewise.
525 (MPX_SPEC): Likewise.
526 (LIBMPX_SPEC): Likewise.
527 (LIBMPXWRAPPERS_SPEC): Likewise.
528 (CHKP_SPEC): Likewise.
529 * config/i386/predicates.md: Likewise.
530 * dbxout.c (dbxout_type): Likewise.
531 * doc/extend.texi: Likewise.
532 * doc/invoke.texi: Likewise.
533 * doc/md.texi: Likewise.
534 * doc/tm.texi: Likewise.
535 * doc/tm.texi.in: Likewise.
536 * dwarf2out.c (is_base_type): Likewise.
537 (gen_formal_types_die): Likewise.
538 (gen_subprogram_die): Likewise.
539 (gen_type_die_with_usage): Likewise.
540 (gen_decl_die): Likewise.
541 (dwarf2out_late_global_decl): Likewise.
542 * expr.c (expand_assignment): Likewise.
543 (emit_storent_insn): Likewise.
544 (store_expr_with_bounds): Likewise.
545 (store_expr): Likewise.
546 (expand_expr_real_1): Likewise.
547 * expr.h (store_expr_with_bounds): Likewise.
548 * function.c (use_register_for_decl): Likewise.
549 (struct bounds_parm_data): Likewise.
550 (assign_parms_augmented_arg_list): Likewise.
551 (assign_parm_find_entry_rtl): Likewise.
552 (assign_parm_is_stack_parm): Likewise.
553 (assign_parm_load_bounds): Likewise.
554 (assign_bounds): Likewise.
555 (assign_parms): Likewise.
556 (expand_function_start): Likewise.
557 * gcc.c (CHKP_SPEC): Likewise.
558 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
559 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
560 (wrestrict_dom_walker::check_call): Likewise.
561 * gimple.c (gimple_build_call_from_tree): Likewise.
562 * gimple.h (enum gf_mask): Likewise.
563 (gimple_call_with_bounds_p): Likewise.
564 (gimple_call_set_with_bounds): Likewise.
565 * gimplify.c (gimplify_init_constructor): Likewise.
566 * ipa-cp.c (initialize_node_lattices): Likewise.
567 (propagate_constants_across_call): Likewise.
568 (find_more_scalar_values_for_callers_subset): Likewise.
569 * ipa-hsa.c (process_hsa_functions): Likewise.
570 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
571 * ipa-icf.c (sem_function::merge): Likewise.
572 * ipa-inline.c (early_inliner): Likewise.
573 * ipa-pure-const.c (warn_function_noreturn): Likewise.
574 (warn_function_cold): Likewise.
575 (propagate_pure_const): Likewise.
576 * ipa-ref.h (enum GTY): Likewise.
577 * ipa-split.c (find_retbnd): Likewise.
578 (consider_split): Likewise.
579 (split_function): Likewise.
580 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
581 * ipa.c (walk_polymorphic_call_targets): Likewise.
582 (symbol_table::remove_unreachable_nodes): Likewise.
583 (process_references): Likewise.
584 (cgraph_build_static_cdtor_1): Likewise.
585 * lto-cgraph.c (lto_output_node): Likewise.
586 (output_refs): Likewise.
587 (compute_ltrans_boundary): Likewise.
588 (input_overwrite_node): Likewise.
589 (input_node): Likewise.
590 (input_cgraph_1): Likewise.
591 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
592 * passes.c (pass_manager::execute_early_local_passes): Likewise.
593 (class pass_chkp_instrumentation_passes): Likewise.
594 (make_pass_chkp_instrumentation_passes): Likewise.
595 * passes.def: Likewise.
596 * rtl.h (struct GTY): Likewise.
597 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
598 * stor-layout.c (layout_type): Likewise.
599 * symtab.c: Likewise.
600 * target.def: Likewise.
601 * targhooks.c (default_chkp_bound_type): Likewise.
602 (default_chkp_bound_mode): Likewise.
603 (default_builtin_chkp_function): Likewise.
604 (default_chkp_function_value_bounds): Likewise.
605 (default_chkp_make_bounds_constant): Likewise.
606 (default_chkp_initialize_bounds): Likewise.
607 * targhooks.h (default_chkp_bound_type): Likewise.
608 (default_chkp_bound_mode): Likewise.
609 (default_builtin_chkp_function): Likewise.
610 (default_chkp_function_value_bounds): Likewise.
611 (default_chkp_make_bounds_constant): Likewise.
612 (default_chkp_initialize_bounds): Likewise.
613 * toplev.c (compile_file): Likewise.
614 (process_options): Likewise.
615 * tree-core.h (DEF_BUILTIN): Likewise.
616 (DEF_BUILTIN_CHKP): Likewise.
617 * tree-inline.c (declare_return_variable): Likewise.
618 (remap_gimple_stmt): Likewise.
620 (initialize_inlined_parameters): Likewise.
621 (expand_call_inline): Likewise.
622 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
623 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
624 (make_pass_ipa_chkp_produce_thunks): Likewise.
625 (make_pass_chkp): Likewise.
626 (make_pass_chkp_opt): Likewise.
627 (make_pass_chkp_instrumentation_passes): Likewise.
628 * tree-pretty-print.c (dump_generic_node): Likewise.
629 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
630 * tree-ssa-dce.c (propagate_necessity): Likewise.
631 (eliminate_unnecessary_stmts): Likewise.
632 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
633 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
634 * tree-ssa-sccvn.h: Likewise.
635 * tree-ssa-strlen.c (get_string_length): Likewise.
636 (valid_builtin_call): Likewise.
637 (adjust_last_stmt): Likewise.
638 (handle_builtin_strchr): Likewise.
639 (handle_builtin_strcpy): Likewise.
640 (handle_builtin_stxncpy): Likewise.
641 (handle_builtin_memcpy): Likewise.
642 (handle_builtin_strcat): Likewise.
643 (strlen_check_and_optimize_stmt): Likewise.
644 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
645 * tree-streamer-in.c: Likewise.
646 * tree-streamer.c (record_common_node): Likewise.
647 * tree.c (tree_code_size): Likewise.
648 (wide_int_to_tree_1): Likewise.
649 (type_contains_placeholder_1): Likewise.
650 (build_common_tree_nodes): Likewise.
651 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
652 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
653 (POINTER_BOUNDS_P): Likewise.
654 (BOUNDED_TYPE_P): Likewise.
655 (BOUNDED_P): Likewise.
656 (CALL_WITH_BOUNDS_P): Likewise.
657 (pointer_bounds_type_node): Likewise.
658 * value-prof.c (gimple_ic): Likewise.
659 * var-tracking.c (vt_add_function_parameters): Likewise.
660 * varasm.c (make_decl_rtl): Likewise.
661 (assemble_start_function): Likewise.
662 (output_constant): Likewise.
663 (maybe_assemble_visibility): Likewise.
664 * varpool.c (ctor_for_folding): Likewise.
665 * chkp-builtins.def: Remove.
666 * ipa-chkp.c: Remove.
667 * ipa-chkp.h: Remove.
668 * rtl-chkp.c: Remove.
669 * rtl-chkp.h: Remove.
670 * tree-chkp-opt.c: Remove.
671 * tree-chkp.c: Remove.
672 * tree-chkp.h: Remove.
674 2018-06-07 Carl Love <cel@us.ibm.com>
676 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
677 vextract_fp_from_shortl): Add BE support.
679 2018-06-07 Paul Koning <ni1d@arrl.net>
681 * compare-elim.c (try_merge_compare): Don't merge compare if
682 address contains a side effect.
683 (try_eliminate_compare): Likewise.
685 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
687 * config.gcc: Support "tremont".
688 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
689 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
691 * config/i386/i386.c (m_TREMONT): Define.
692 (processor_target_table): Add "tremont".
693 (PTA_TREMONT): Define.
694 (ix86_lea_outperforms): Add TARGET_TREMONT.
695 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
696 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
697 and M_INTEL_GOLDMONT_PLUS.
698 (fold_builtin_cpu): Add "tremont".
699 (ix86_add_stmt_cost): Add TARGET_TREMONT.
700 (ix86_option_override_internal): Add "tremont".
701 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
702 (processor_type): Add PROCESSOR_TREMONT.
703 * config/i386/x86-tune.def: Add m_TREMONT.
704 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
706 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
708 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
709 symbol defined for msp430i* devices to be lower case.
711 gcc/testsuite/gcc.target/msp430/
712 * msp430i-device-symbol.c: New test.
713 * msp430f-device-symbol.c: New test.
714 * msp430.h: New test header file.
716 2018-06-07 Richard Biener <rguenther@suse.de>
718 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
719 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
720 Properly wrap signed arithmetic if overflow wraps.
722 2018-06-07 Jakub Jelinek <jakub@redhat.com>
724 PR tree-optimization/69615
725 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
726 of a cast from a same precision integral SSA_NAME in a bb dominated
727 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
728 cast to utype if rhs2 has already a compatible type.
730 2018-06-07 Richard Biener <rguenther@suse.de>
732 PR tree-optimization/85935
733 * graphite-scop-detection.c (find_params_in_bb): Analyze
734 condition operands with respect to the correct loop. Assert
735 the analysis doesn't fail.
737 2018-06-04 Carl Love <cel@us.ibm.com>
739 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
740 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
741 as it is slightly cheaper.
742 (first_match_or_eos_index_<mode>):
743 Calculate index using natural element order.
744 (first_match_index_<mode>):
745 Calculate index using natural element order.
746 (first_match_or_eos_index_<mode>):
747 Calculate index using natural order.
748 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
750 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
751 P9V_BUILTIN_VCLZLSBB_V16QI.
752 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
755 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
757 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
758 indentation and line wrap for many prototypes. Add missing
759 @smallexample directives around block of prototypes for vec_xl and
762 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
764 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
765 track if we pass or return IEEE 128-bit floating point.
766 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
767 C++ mangling that is compatible with GCC 8.1.
768 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
769 (init_cumulative_args): Note if we pass or return IEEE 128-bit
770 floating point types.
771 (rs6000_function_arg_advance_1): Likewise.
772 (rs6000_mangle_type): Optionally generate mangled names that match
773 what GCC 8.1 generated for IEEE 128-bit floating point types.
774 (rs6000_globalize_decl_name): If we have an external function that
775 passes or returns IEEE 128-bit types, generate a weak reference
776 from the mangled name used in GCC 8.1 to the current mangled
778 (rs6000_init_builtins): Make __ibm128 use the long double type if
779 long double is IBM extended double. Make __float128 use the long
780 double type if long double is IEEE 128-bit.
783 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
784 macro for __ibm128 built-in functions.
785 (PACK_IF): Add __ibm128 pack/unpack functions.
786 (UNPACK_IF): Likewise.
787 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
788 enable long double built-in functions if long double is IEEE
789 128-bit floating point.
790 (rs6000_invalid_builtin): Update long double built-in function
792 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
793 functions, adjust the built-in function to use the long double
794 built-in function if __ibm128 and long double are the same type.
795 * doc/extend.texi (PowerPC builtins): Update documention for
796 __builtin_{,un}pack_longdouble. Add documentation for
797 __builtin_{,un}pack_ibm128.
799 2018-06-06 Jim Wilson <jimw@sifive.com>
801 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
802 (struct machine_function): New field interrupt_mode.
803 (riscv_handle_type_attribute): New function. Add forward declaration.
804 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
805 (riscv_expand_epilogue): Check interrupt_mode field.
806 (riscv_set_current_function): Check interrupt attribute args and
807 set interrupt_mode field.
808 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
809 (riscv_sret, riscv_uret): New.
810 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
811 new arguments to interrupt attribute.
813 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
816 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
817 Don't handle -mcpu=power8 if -mpower9-vector is also used.
819 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
821 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
822 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
823 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
824 several redundant entries.
826 2018-06-06 David Malcolm <dmalcolm@redhat.com>
828 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
829 type from "rtx" to "rtx_insn *".
830 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
831 for local "call_insn", removing cast.
832 (ix86_expand_call): Likewise, introducing a "call_insn" local.
834 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
836 PR tree-optimization/86066
837 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
838 for BIT_INSERT_EXPR stores.
840 2018-06-06 Richard Biener <rguenther@suse.de>
842 PR tree-optimization/86062
843 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
845 of to be offsetted base.
847 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
849 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
850 to be static and remove check on interrupt attribute name.
852 gcc/testsuite/gcc.target/msp430/
853 * function-attributes-4.c: New test.
854 * static-interrupts.c: New test.
856 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
858 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
859 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
861 2018-06-05 Steve Ellcey <sellcey@cavium.com>
864 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
866 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
867 Remove second argument, change how error is called.
868 (aarch64_layout_arg): Remove second argument from
869 aarch64_err_no_fpadvsimd call.
870 (aarch64_init_cumulative_args): Ditto.
871 (aarch64_gimplify_va_arg_expr): Ditto.
872 * config/aarch64/aarch64.md (mov<mode>): Ditto.
874 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
876 * config/i386/i386.md (simple_return_indirect_internal): New expander.
877 (*simple_return_indirect_internal<mode>): Rename from
878 simple_return_indirect_internal. Use W mode iterator.
879 (rstorssp): New expander.
880 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
881 (clrssbsy): New expander.
882 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
884 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
886 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
888 (cmse_check_pointed_object): Likewise.
890 2018-06-05 Martin Liska <mliska@suse.cz>
892 PR gcov-profile/47618
893 * doc/invoke.texi: Document how -fprofile-dir format
896 2018-06-05 Richard Biener <rguenther@suse.de>
898 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
899 removal pretend DOM info isn't available so we do not update
900 it and only remove edges, not dominated blocks. Actually free
901 DOM info in case we removed something. Remove unreachable blocks.
902 (mfb_keep_latches): Work with either DOM info or marked backedges.
903 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
904 first. Mark backedges if DOM info isn't available.
905 (Re-)compute DOM info after cleanup_control_flow_pre.
907 2018-06-05 Richard Biener <rguenther@suse.de>
909 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
910 (locus_discrim_hasher::hash): Adjust.
911 (locus_discrim_hasher::equal): Likewise.
912 (next_discriminator_for_locus): Work on line directly.
913 (same_line_p): Pass in expanded locus1 as well.
914 (assign_discriminators): Avoid redundant location expansions.
916 2018-06-05 Richard Biener <rguenther@suse.de>
918 PR tree-optimization/86046
919 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
920 if required after clearing TREE_ADDRESSABLE.
922 2018-06-05 Richard Biener <rguenther@suse.de>
924 PR tree-optimization/86047
925 * tree-ssa-loop.c (for_each_index): Glob handling of all
926 decls and constants and really handle all of them.
928 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
931 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
932 qualifier_void_pointer and qualifier_const_void_pointer.
933 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
934 (arm_init_builtins): Handle the above.
935 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
936 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
937 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
940 2018-06-05 Martin Liska <mliska@suse.cz>
942 * auto-profile.c (read_autofdo_file): Do not use
943 gcov_ctr_summary struct.
944 (afdo_callsite_hot_enough_for_early_inline): Likewise.
945 * coverage.c (struct counts_entry): Likewise.
946 (read_counts_file): Read just single summary entry.
947 (get_coverage_counts): Use gcov_summary struct.
948 * coverage.h (get_coverage_counts): Likewise.
949 * gcov-dump.c (dump_working_sets): Likewise.
950 (tag_summary): Dump just single summary.
951 * gcov-io.c (gcov_write_summary): Write just histogram
953 (gcov_read_summary): Read just single summary.
954 (compute_working_sets): Use gcov_summary struct.
955 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
956 of GCOV_COUNTERS_SUMMABLE.
957 (GCOV_COUNTERS_SUMMABLE): Remove.
958 (GCOV_FIRST_VALUE_COUNTER): Replace with
959 GCOV_COUNTER_V_INTERVAL.
960 (struct gcov_ctr_summary): Remove.
961 (struct gcov_summary): Directly use fields of former
963 (compute_working_sets): Use gcov_summary struct.
964 * gcov.c (read_count_file): Do not use ctrs fields.
965 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
967 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
969 * profile.c: Likewise.
970 * profile.h: Likewise.
972 2018-06-05 Martin Liska <mliska@suse.cz>
974 PR gcov-profile/84846
975 * gcov.c (output_lines): Print working directory only
976 in intermediate format.
978 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
980 * config/s390/s390-builtin-types.def: Add void function type.
981 * config/s390/s390-builtins.def: Use the function type for the
984 2018-06-04 Jim Wilson <jimw@sifive.com>
986 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
988 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
989 and maybe_eh_return. Change regno to unsigned int. Use new args to
990 handle EH_RETURN_DATA_REGNO registers properly.
991 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
992 (riscv_expand_epilogue): Update comment. Change argument name and
993 type. Update code to use new name and type. Pass new args to
994 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
996 * config/riscv/riscv.md (NORMAL_RETURN): New.
997 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
998 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
999 (eh_return): Call gen_eh_return_internal and emit barrier.
1000 (eh_return_internal): Call riscv_expand_epilogue.
1002 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1004 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1005 bit_insertion field and declare can_be_merged_into method.
1006 (merged_store_group::can_be_merged_into): New method.
1007 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1008 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1009 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1011 2018-06-04 Richard Biener <rguenther@suse.de>
1013 PR tree-optimization/85955
1014 * builtins.c (fold_builtin_sincos): Convert pointers to
1015 destination to appropriate type before dereferencing.
1017 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1019 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1021 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1023 * expr.c (expand_expr_real_1): Force the operand into memory if
1024 its TYPE_MODE is BLKmode and if there is no integer mode for
1025 the number of bits being extracted.
1027 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1031 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1032 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1034 2018-06-04 Richard Biener <rguenther@suse.de>
1036 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1037 (cleanup_tree_cfg_noloop): ... single caller. Do
1038 start_recording_case_labels later.
1040 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1042 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1043 to _IMMINTRIN_H_INCLUDED.
1044 * config/i386/pconfigintrin.h: Ditto.
1045 * config/i386/waitpkgintrin.h: Ditto.
1046 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1047 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1048 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1049 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1050 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1051 waitpkgintrin.h and cldemoteintrin.h.
1053 2018-06-04 Richard Biener <rguenther@suse.de>
1055 PR tree-optimization/86038
1056 * tracer.c (find_best_successor): Check probability for
1057 being initialized, bail out if not.
1059 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1062 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1063 of bits to ignore when comparing architectures.
1065 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1067 PR tree-optimization/69615
1068 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1069 maximum or minimum of the type, try to merge it also as if
1070 range1 is + [-, x - 1] or + [x + 1, -].
1073 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1075 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1077 PR tree-optimization/86034
1078 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1079 the unsigned bitfield type in a bit insertion sequence if it does not
1080 have a larger precision than the bitfield size.
1081 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1083 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1085 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1087 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1089 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1090 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1091 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1092 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1094 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1096 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1097 Disable -fdelete-null-pointer-checks for ELF toolchain.
1099 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1100 Kito Cheng <kito.cheng@gmail.com>
1102 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1103 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1104 (nds32*-*-*): Add glibc and uclibc conditions.
1105 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1106 (TARGET_EXCEPT_UNWIND_INFO): Define.
1107 * config/nds32/elf.h: New file.
1108 * config/nds32/linux.h: New file.
1109 * config/nds32/nds32-elf.opt: New file.
1110 * config/nds32/nds32-linux.opt: New file.
1111 * config/nds32/nds32-fp-as-gp.c
1112 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1113 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1115 (nds32_asm_file_end): Ditto.
1116 (nds32_print_operand): Ditto.
1117 (nds32_insert_attributes): Ditto.
1118 (nds32_init_libfuncs): New function.
1119 (TARGET_HAVE_TLS): Define.
1120 (TARGET_INIT_LIBFUNCS): Define.
1121 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1123 (TARGET_ELF): Apply different mcmodel setting.
1124 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1125 been migrated into elf.h and linux.h files.
1126 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1127 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1128 (mcmodel): The content has been migrated into nds32-elf.opt and
1129 nds32-linux.opt files.
1130 * config/nds32/t-elf: New file.
1131 * config/nds32/t-linux: New file.
1133 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1134 Shiva Chen <shiva0217@gmail.com>
1136 * config/nds32/constants.md (unspec_volatile_element): Add
1137 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1138 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1140 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1141 (make_pass_nds32_fp_as_gp): Declare.
1142 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1144 (nds32_asm_function_end_prologue): Remove unused asm output.
1145 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1146 (nds32_asm_file_start): Output necessary fp_as_gp information.
1147 (nds32_option_override): Adjust register usage.
1148 (nds32_expand_prologue): Consider fp_as_gp situation.
1149 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1150 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1153 (simple_return): Ditto.
1154 (omit_fp_begin): Output special directive for fp_as_gp.
1155 (omit_fp_end): Output special directive for fp_as_gp.
1156 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1157 mforbid-fp-as-gp): New options.
1159 2018-06-01 Mark Wielaard <mark@klomp.org>
1161 * dwarf2out.c (dwarf2out_finish): Remove generation of
1162 DW_AT_loclists_base.
1164 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1166 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1167 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1168 (struct merged_store_group): Add bit_insertion field.
1169 (dump_char_array): Use standard hexadecimal format.
1170 (merged_store_group::merged_store_group): Set bit_insertion to false.
1171 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1172 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1173 also print the mask in the dump file.
1174 (pass_store_merging::gate): Minor tweak.
1175 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1176 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1177 stores with INTEGER_CST stores.
1178 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1179 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1180 and use it throughout. Generate bit insertion sequences if need be.
1181 (pass_store_merging::process_store): Remove redundant condition.
1182 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1184 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1186 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1187 the 128-bit floating point types. Fix function comment.
1189 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1191 * config/aarch64/aarch64-simd.md
1192 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1194 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1197 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1199 PR tree-optimization/85989
1200 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1202 (backprop::intersect_uses): Check it when deciding whether this
1203 is a backedge reference.
1204 (backprop::process_block): Add each phi to m_visited_phis
1205 after visiting it, then clear it at the end.
1207 2018-06-01 Richard Biener <rguenther@suse.de>
1209 * tree-vectorizer.h (vect_dr_stmt): New function.
1210 (vect_get_load_cost): Adjust.
1211 (vect_get_store_cost): Likewise.
1212 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1213 Use vect_dr_stmt instead of DR_SMTT.
1214 (vect_record_base_alignments): Likewise.
1215 (vect_calculate_target_alignment): Likewise.
1216 (vect_compute_data_ref_alignment): Likewise and make static.
1217 (vect_update_misalignment_for_peel): Likewise.
1218 (vect_verify_datarefs_alignment): Likewise.
1219 (vector_alignment_reachable_p): Likewise.
1220 (vect_get_data_access_cost): Likewise. Pass down
1221 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1222 (vect_get_peeling_costs_all_drs): Likewise.
1223 (vect_peeling_hash_get_lowest_cost): Likewise.
1224 (vect_enhance_data_refs_alignment): Likewise.
1225 (vect_find_same_alignment_drs): Likewise.
1226 (vect_analyze_data_refs_alignment): Likewise.
1227 (vect_analyze_group_access_1): Likewise.
1228 (vect_analyze_group_access): Likewise.
1229 (vect_analyze_data_ref_access): Likewise.
1230 (vect_analyze_data_ref_accesses): Likewise.
1231 (vect_vfa_segment_size): Likewise.
1232 (vect_small_gap_p): Likewise.
1233 (vectorizable_with_step_bound_p): Likewise.
1234 (vect_prune_runtime_alias_test_list): Likewise.
1235 (vect_analyze_data_refs): Likewise.
1236 (vect_supportable_dr_alignment): Likewise.
1237 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1238 (vect_gen_prolog_loop_niters): Likewise.
1239 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1240 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1242 (vect_recog_mask_conversion_pattern): Likewise.
1243 (vect_try_gather_scatter_pattern): Likewise.
1244 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1245 to vect_get_store_cost.
1246 (vect_get_store_cost): Get stmt_info instead of DR.
1247 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1248 (vect_get_load_cost): Get stmt_info instead of DR.
1250 2018-06-01 Richard Biener <rguenther@suse.de>
1253 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1254 vars wrapped in MEM_REFs.
1256 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1258 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1259 Fix subreg tests so that we only return a choice between
1260 GENERAL_REGS and FP_REGS if the original classes included both.
1262 2018-06-01 Richard Biener <rguenther@suse.de>
1265 * tree-ssa-structalias.c (get_function_part_constraint):
1266 Handle NULL fi->decl.
1267 (find_func_aliases_for_call): Properly handle indirect
1268 fi from direct call.
1269 (find_func_clobbers): Likewise.
1270 (ipa_pta_execute): Likewise.
1271 (create_variable_info_for): For functions that are ifunc_resolver
1272 resolve to a varinfo that contains the result of the resolver call.
1273 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1276 2018-05-31 Michael Collison <michael.collison@arm.com>
1278 * config/aarch64/aarch64.md:
1279 (*fix_to_zero_extenddfdi2): New pattern.
1280 * gcc.target/aarch64/fix_extend1.c: New testcase.
1282 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1286 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1287 and BUILT_IN_STRNCMP_EQ.
1288 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1289 BUILT_IN_STRNCMP_EQ.
1290 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1291 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1292 (gimple_fold_builtin): Likewise.
1293 * tree-ssa-strlen.c (compute_string_length): New function.
1294 (determine_min_obsize): New function.
1295 (handle_builtin_string_cmp): New function to handle calls to
1296 string compare functions.
1297 (strlen_optimize_stmt): Add handling to builtin string compare
1299 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1300 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1301 * tree.c (build_common_builtin_nodes): Add new defines of
1302 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1304 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1307 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1308 functions with naked attribute.
1310 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1312 * config/i386/sse.md (avx_vec_concat<mode>):
1313 Substitute concat_tg_mode mode attribute with xtg_mode.
1314 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1315 (concat_tg_mode): Remove mode attribute.
1317 2018-05-31 Martin Sebor <msebor@redhat.com>
1320 * calls.c (alloc_max_size): Correct a logic error/typo.
1321 Treat excessive arguments as infinite. Warn for invalid arguments.
1322 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1324 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1327 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1328 and movx for Haswell.
1330 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1331 Cesar Philippidis <cesar@codesourcery.com>
1334 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1335 when emitting error on private/firstprivate reductions.
1336 * omp-low.c (lower_omp_target): Avoid reference-type processing
1337 on pointers for firstprivate clause.
1339 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1341 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1344 * config/aarch64/aarch64-simd.md
1345 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1346 (aarch64_ld1_x3_<mode>): Likewise
1347 (aarch64_st1x2<VALLDIF:mode>): Likewise
1348 (aarch64_st1_x2_<mode>): Likewise
1349 (aarch64_st1x3<VALLDIF:mode>): Likewise
1350 (aarch64_st1_x3_<mode>): Likewise
1351 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1352 (vld1_s8_x3): Likewise.
1353 (vld1_u16_x3): Likewise.
1354 (vld1_s16_x3): Likewise.
1355 (vld1_u32_x3): Likewise.
1356 (vld1_s32_x3): Likewise.
1357 (vld1_u64_x3): Likewise.
1358 (vld1_s64_x3): Likewise.
1359 (vld1_f16_x3): Likewise.
1360 (vld1_f32_x3): Likewise.
1361 (vld1_f64_x3): Likewise.
1362 (vld1_p8_x3): Likewise.
1363 (vld1_p16_x3): Likewise.
1364 (vld1_p64_x3): Likewise.
1365 (vld1q_u8_x3): Likewise.
1366 (vld1q_s8_x3): Likewise.
1367 (vld1q_u16_x3): Likewise.
1368 (vld1q_s16_x3): Likewise.
1369 (vld1q_u32_x3): Likewise.
1370 (vld1q_s32_x3): Likewise.
1371 (vld1q_u64_x3): Likewise.
1372 (vld1q_s64_x3): Likewise.
1373 (vld1q_f16_x3): Likewise.
1374 (vld1q_f32_x3): Likewise.
1375 (vld1q_f64_x3): Likewise.
1376 (vld1q_p8_x3): Likewise.
1377 (vld1q_p16_x3): Likewise.
1378 (vld1q_p64_x3): Likewise.
1379 (vst1_s64_x2): Likewise.
1380 (vst1_u64_x2): Likewise.
1381 (vst1_f64_x2): Likewise.
1382 (vst1_s8_x2): Likewise.
1383 (vst1_p8_x2): Likewise.
1384 (vst1_s16_x2): Likewise.
1385 (vst1_p16_x2): Likewise.
1386 (vst1_s32_x2): Likewise.
1387 (vst1_u8_x2): Likewise.
1388 (vst1_u16_x2): Likewise.
1389 (vst1_u32_x2): Likewise.
1390 (vst1_f16_x2): Likewise.
1391 (vst1_f32_x2): Likewise.
1392 (vst1_p64_x2): Likewise.
1393 (vst1q_s8_x2): Likewise.
1394 (vst1q_p8_x2): Likewise.
1395 (vst1q_s16_x2): Likewise.
1396 (vst1q_p16_x2): Likewise.
1397 (vst1q_s32_x2): Likewise.
1398 (vst1q_s64_x2): Likewise.
1399 (vst1q_u8_x2): Likewise.
1400 (vst1q_u16_x2): Likewise.
1401 (vst1q_u32_x2): Likewise.
1402 (vst1q_u64_x2): Likewise.
1403 (vst1q_f16_x2): Likewise.
1404 (vst1q_f32_x2): Likewise.
1405 (vst1q_f64_x2): Likewise.
1406 (vst1q_p64_x2): Likewise.
1407 (vst1_s64_x3): Likewise.
1408 (vst1_u64_x3): Likewise.
1409 (vst1_f64_x3): Likewise.
1410 (vst1_s8_x3): Likewise.
1411 (vst1_p8_x3): Likewise.
1412 (vst1_s16_x3): Likewise.
1413 (vst1_p16_x3): Likewise.
1414 (vst1_s32_x3): Likewise.
1415 (vst1_u8_x3): Likewise.
1416 (vst1_u16_x3): Likewise.
1417 (vst1_u32_x3): Likewise.
1418 (vst1_f16_x3): Likewise.
1419 (vst1_f32_x3): Likewise.
1420 (vst1_p64_x3): Likewise.
1421 (vst1q_s8_x3): Likewise.
1422 (vst1q_p8_x3): Likewise.
1423 (vst1q_s16_x3): Likewise.
1424 (vst1q_p16_x3): Likewise.
1425 (vst1q_s32_x3): Likewise.
1426 (vst1q_s64_x3): Likewise.
1427 (vst1q_u8_x3): Likewise.
1428 (vst1q_u16_x3): Likewise.
1429 (vst1q_u32_x3): Likewise.
1430 (vst1q_u64_x3): Likewise.
1431 (vst1q_f16_x3): Likewise.
1432 (vst1q_f32_x3): Likewise.
1433 (vst1q_f64_x3): Likewise.
1434 (vst1q_p64_x3): Likewise.
1436 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1438 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1439 user_label_prefix to name.
1441 * tree-core.h: Update comment about the format of NAME string
1442 passed to handler in attribute_spec.
1444 * config/msp430/msp430.md: Remove erroneous subreg expression from
1445 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1446 zero_extend{q,h}isi2.
1448 2018-05-30 Borislav Petkov <bp@suse.de>
1450 * doc/extend.texi: Document some architecture specific
1451 constraints and sort entries.
1453 2018-05-30 Martin Sebor <msebor@redhat.com>
1456 * builtins.c (expand_builtin_stpcpy_1): New function.
1457 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1458 only if the former succeeds.
1460 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1462 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1465 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1467 * doc/invoke.texi (-flinker-output): Document
1469 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1471 * passes.c (ipa_write_summaries): Only modify statements if body
1473 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1474 incrementally linking.
1475 (ipa_passes): Likewise.
1476 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1477 pass down resolution info.
1478 * common.opt (flag_incremental_link): Update info.
1479 * gcc.c (plugin specs): Turn flinker-output=* to
1480 -plugin-opt=-linker-output-known
1481 * toplev.c (compile_file): Also cut compilation when doing incremental
1483 * flag-types. (enum lto_partition_model): Add
1484 LTO_LINKER_OUTPUT_NOLTOREL.
1485 (invoke.texi): Add -flinker-output docs.
1486 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1487 link same way as WPA; do not stream in dead initializers.
1489 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1490 darf2out_register_external_decl): Support incremental link.
1492 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1494 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1496 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1498 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1499 it down to simple_object_copy_lto_debug_sections.
1500 (run_gcc): Determine incremental LTO link time and configure
1501 lto1 into non-wpa mode, disable renaming of debug sections.
1503 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1505 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1506 descriptions of various incorrectly documented functions.
1508 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1511 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1512 address check not strict.
1514 2018-05-30 Richard Biener <rguenther@suse.de>
1516 PR tree-optimization/85964
1517 * tracer.c (better_p): Drop initialized count check, we only
1518 call the function with initialized counts now.
1519 (find_best_successor): Do find a best edge if one
1520 has uninitialized count.
1521 (find_best_predecessor): Likewise. Do BB frequency check only
1522 if count is initialized.
1524 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1526 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1527 (aarch64_ldrstr_offset_compare): New.
1528 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1529 load/store orderings.
1530 (aarch64_gen_adjusted_ldpstp): Likewise.
1532 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1534 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1535 Check for subset of GENERAL_REGS and FP_REGS.
1536 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1539 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1541 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1542 and wi::to_poly_offset. Add the current offset and then check
1543 whether the sum fits, rather than using an unchecked addition of
1544 a checked term. Check for a shwi rather than a uhwi.
1545 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1546 (store_constructor): Use poly_int_tree_p.
1547 (expand_expr_real_1): Likewise.
1548 * function.c (assign_temp): Likewise.
1549 * fold-const.c (const_binop): Use poly_int_tree_p and
1551 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1553 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1554 to_poly_offset for MEM offsets.
1555 * ipa-icf.c (sem_variable::equals): Likewise.
1556 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1557 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1558 wi::to_poly_offset for BIT_FIELD_REF offsets.
1559 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1561 * var-tracking.c (emit_note_insn_var_location): Use
1562 tree_to_poly_uint64.
1564 2018-05-29 Jim Wilson <jimw@sifive.com>
1566 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1568 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1571 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1572 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1574 (sse4_1_round<mode>2): Use nonimmediate_operand
1575 for operand 1 predicate.
1577 2018-05-29 Martin Sebor <msebor@redhat.com>
1578 Richard Biener <rguenther@suse.de>
1581 * calls.c (get_size_range): Call determine_value_range instead
1582 of get_value_range..
1583 * tree-vrp.h (determine_value_range): Declared new function.
1584 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1586 2018-05-29 Richard Biener <rguenther@suse.de>
1588 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1589 sure to use non-pattern stmts for get_earlier_stmt arguments.
1590 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1591 called on pattern stmts.
1592 (get_later_stmt): Likewise.
1594 2018-05-29 Martin Liska <mliska@suse.cz>
1596 PR gcov-profile/85759
1597 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1600 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1602 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1604 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1608 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1609 VEC_PACK_FLOAT_EXPR): New tree codes.
1610 * tree-pretty-print.c (op_code_prio): Handle
1611 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1612 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1613 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1614 * tree-inline.c (estimate_operator_cost): Likewise.
1615 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1616 * fold-const.c (const_binop): Likewise.
1617 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1618 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1619 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1620 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1621 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1622 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1623 * expr.c (expand_expr_real_2): Likewise.
1624 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1625 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1626 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1628 * optabs.c (expand_widen_pattern_expr): For
1629 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1630 sign from result type rather than operand's type.
1631 (expand_binop_directly): For vec_packu_float_optab and
1632 vec_packs_float_optab allow result type to be different from operand's
1634 * optabs-tree.c (optab_for_tree_code): Handle
1635 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1636 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1637 * tree-vect-generic.c (expand_vector_operations_1): Handle
1638 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1639 VEC_PACK_FLOAT_EXPR.
1640 * tree-vect-stmts.c (supportable_widening_operation): Handle
1642 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1643 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1644 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1645 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1646 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1648 (vec_pack<floatprefix>_float_<mode>): New expander.
1649 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1651 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1652 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1653 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1654 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1655 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1657 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1658 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1659 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1660 VEC_PACK_FLOAT_EXPR): Document.
1662 2018-05-29 Richard Biener <rguenther@suse.de>
1664 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1666 (stmt_vec_info_vec): Make pointer.
1667 (init_stmt_vec_info_vec): Remove.
1668 (free_stmt_vec_info_vec): Likewise.
1669 (set_stmt_vec_info_vec): New function.
1670 (free_stmt_vec_infos): Likewise.
1671 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1672 (set_vinfo_for_stmt): Likewise.
1673 (get_earlier_stmt): Likewise.
1674 (get_later_stmt): Likewise.
1675 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1676 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1677 (vec_info::~vec_info): Free stmt_vec_infos.
1678 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1679 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1680 (pass_slp_vectorize::execute): Likewise.
1681 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1682 (free_stmt_vec_info_vec): Likewise.
1683 (set_stmt_vec_info_vec): New function.
1684 (free_stmt_vec_infos): Likewise.
1685 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1686 the global stmt_vec_info_vec.
1687 * tree-parloops.c (gather_scalar_reductions): Use
1688 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1691 2018-05-29 Richard Biener <rguenther@suse.de>
1693 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1695 2018-05-29 Martin Liska <mliska@suse.cz>
1696 David Malcolm <dmalcolm@redhat.com>
1698 * vec.c (test_reverse): New.
1699 (vec_c_tests): Add new test.
1700 * vec.h (vl_ptr>::reverse): New function.
1702 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1704 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1706 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1709 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1711 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1713 2018-05-28 Richard Biener <rguenther@suse.de>
1715 PR tree-optimization/85933
1716 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1717 look at stmts marked as vectorizable.
1719 2018-05-28 Richard Biener <rguenther@suse.de>
1721 PR tree-optimization/85934
1722 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1723 vector boolean check before scalar optimization.
1725 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1727 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1730 2018-05-28 Mark Wielaard <mark@klomp.org>
1732 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1733 if it is an expression containing a minus sign.
1735 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1737 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1739 2018-05-27 Paul Koning <ni1d@arrl.net>
1741 * config/pdp11/pdp11.md (truncsihi2): Remove.
1743 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1744 Chung-Ju Wu <jasonwucj@gmail.com>
1746 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1748 (unaligned_store_dw): Ditto.
1749 * config/nds32/nds32-memory-manipulation.c
1750 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1751 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1752 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1753 (emit_setmem_word_loop): Rename to ...
1754 (emit_setmem_doubleword_loop): ... this.
1755 (nds32_gen_dup_4_byte_to_word_value): New function.
1756 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1757 (nds32_expand_setmem_loop): Refine implementation.
1758 (nds32_expand_setmem_loop_v3m): Ditto.
1759 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1762 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1764 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1766 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1768 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1769 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1770 machine->attr_no_prologue_p.
1771 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1772 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1773 (nds32_expand_epilogue): Consider attr_naked_p.
1774 (nds32_expand_epilogue_v3pop): Likewise.
1775 (nds32_can_use_return_insn): Likewise.
1776 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1777 attr_no_prologue_p fields.
1778 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1780 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1783 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1785 * config/i386/sse.md
1786 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1788 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1790 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1792 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1794 (*<floatsuffix>floatv2div2sf2): Rename to ...
1795 (*float<floatunssuffix>v2div2sf2): ... this.
1796 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1797 (float<floatunssuffix>v2div2sf2_mask): ... this.
1798 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1799 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1800 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1802 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1804 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1806 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1808 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1810 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1812 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1813 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1814 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1815 gen_ufix_truncv8dfv8si2.
1816 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1817 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1818 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1819 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1820 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1821 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1822 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1823 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1825 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1829 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1831 2018-05-25 Jim Wilson <jimw@sifive.com>
1833 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1834 * config/riscv/riscv.c (struct machine_function): Add
1835 interrupt_handler_p and attribute_checked_p fields.
1836 (riscv_attribute_table): Add interrupt.
1837 (riscv_interrupt_type_p): New.
1838 (riscv_save_reg_p): Save extra regs for interrupt handler.
1839 (riscv_use_save_libcall): Return false for interrupt handler.
1840 (riscv_first_stack_step): Add forward declaration.
1841 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1842 for interrupt handler with large frame. Use it for saved reg list.
1843 (riscv_expand_prologue): Move flag_stack_usage_info support to
1844 eliminate duplication.
1845 (riscv_expand_epilogue): Generate mret for interrupt handler.
1846 (riscv_epilogue_uses): New.
1847 (riscv_can_use_return_insn): Return false for interrupt handler.
1848 (riscv_function_ok_for_sibcall): Likewise.
1849 (riscv_set_current_function): Add interrupt handler support.
1850 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1851 * config/riscv/riscv.md (UNSPECV_MRET): New.
1853 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1855 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1857 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1859 PR tree-optimization/85712
1860 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1861 this candidate has already been replaced in-situ by a copy.
1863 2018-05-25 Jason Merrill <jason@redhat.com>
1865 PR c++/80485 - inline function non-zero address.
1866 * symtab.c (nonzero_address): Check DECL_COMDAT.
1868 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1871 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1872 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1873 extension of SImode operation. Use const123_operand predicate.
1874 (*saddsi_1): Remove.
1875 (*saddl_se_1): Ditto.
1877 (*ssubl_se_1): Ditto.
1878 * config/alpha/predicates.md (const123_operand): New predicate.
1879 * config/alpha/constraints.md (P): Use IN_RANGE.
1881 2018-05-25 Richard Biener <rguenther@suse.de>
1883 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1885 (ref_maybe_used_by_stmt_p): Likewise.
1886 (stmt_may_clobber_ref_p): Likewise.
1887 (stmt_may_clobber_ref_p_1): Likewise.
1888 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1890 (ref_maybe_used_by_stmt_p): Likewise.
1891 (stmt_may_clobber_ref_p): Likewise.
1892 (stmt_may_clobber_ref_p_1): Likewise.
1893 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1894 the alias oracle to disambiguate DRs with stmts DR analysis
1896 (vect_analyze_data_refs): Do not give up on not analyzable
1897 DRs for BB vectorization. Remove code truncating the dataref
1900 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1903 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1904 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1905 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1907 2018-05-25 Richard Biener <rguenther@suse.de>
1909 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1910 function, combining stmt data ref gathering and fatal analysis
1912 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1913 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1914 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1915 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1916 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1918 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1920 PR tree-optimization/85720
1921 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1922 SCC if all partitions are builtins.
1923 (version_loop_by_alias_check): New parameter. Generate cancelable
1924 runtime alias check if all partitions are builtins.
1925 (distribute_loop): Update call to above function.
1927 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1929 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1930 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1931 (parm_default_def_partition_arg): Ditto.
1932 (set_parm_default_def_partition): Ditto.
1933 (get_parm_default_def_partitions): Ditto and make it static.
1934 (get_undefined_value_partitions): Ditto and make it static.
1935 (remove_ssa_form): Refactor call to init_var_map here.
1936 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1937 computation for loop region.
1938 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1939 (register_default_def): Delete.
1940 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1941 (parm_default_def_partition_arg): Ditto.
1942 (set_parm_default_def_partition): Ditto.
1943 (get_parm_default_def_partitions): Ditto and make it static.
1944 (get_undefined_value_partitions): Ditto and make it static.
1945 (coalesce_with_default, coalesce_with_default): Update comment.
1946 (create_coalesce_list_for_region): New func factored out from
1947 create_outofssa_var_map.
1948 (populate_coalesce_list_for_outofssa): New func factored out from
1949 create_outofssa_var_map and coalesce_ssa_name.
1950 (create_outofssa_var_map): Delete.
1951 (coalesce_ssa_name): Refactor to support live range computation.
1952 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1953 (get_parm_default_def_partitions): Delete.
1954 (get_undefined_value_partitions): Ditto.
1955 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1956 computation for loop region.
1957 (new_tree_live_info, loe_visit_block): Ditto.
1958 (live_worklist, set_var_live_on_entry): Ditto.
1959 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1960 * tree-ssa-live.h (struct _var_map): New fields.
1961 (init_var_map): Change decl.
1962 (region_contains_p): New.
1964 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1966 * tree-ssa-live.h (live_merge_and_clear): Delete.
1968 2018-05-25 Richard Biener <rguenther@suse.de>
1971 * tree-dump.c (dequeue_and_dump): Remove access to removed
1972 operand 2 of a SWITCH_EXPR.
1974 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1976 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1977 multiplication and division.
1978 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1979 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1980 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1981 (cond_udiv_optab, cond_umod_optab): New optabs.
1982 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1983 (IFN_COND_RDIV): New internal functions.
1984 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1985 TRUNC_MOD_EXPR and RDIV_EXPR.
1986 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1987 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1989 (SVE_INT_BINARY): Include mult.
1990 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1991 (optab, sve_int_op): Handle mult.
1992 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1994 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1995 for SVE_INT_BINARY_SD.
1997 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1999 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2000 (optab, sve_int_op): Handle div and udiv.
2001 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2002 for SVE_INT_BINARY_SD.
2003 (*<optab><mode>3): New insn for the same.
2005 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2007 * tree-vect-patterns.c: Include predict.h.
2008 (vect_recog_divmod_pattern): Restrict check for division support
2009 to when optimizing for size.
2011 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2013 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2014 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2015 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2016 (gimple_match_op::set_op): Likewise.
2017 (gimple_resimplify4): Declare.
2018 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2019 (expr::gen_transform): Likewise.
2020 (decision_tree::gen): Generate a simplification routine for 4 operands.
2021 * gimple-match-head.c (gimple_simplify): Add an overload for
2022 4 operands. In the top-level function, handle up to 4 call
2023 arguments and call gimple_resimplify4.
2024 (gimple_resimplify4): New function.
2025 (build_call_internal): Pass a fourth operand.
2026 (maybe_push_to_seq): Likewise.
2027 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2028 Fold VEC_COND_EXPRs of an operation and a default value into
2029 an IFN_COND_* function if possible.
2030 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2032 (SVE_COND_FP_BINARY): Include them.
2033 (optab, sve_fp_op): Handle them.
2034 (SVE_INT_BINARY_REV): New code iterator.
2035 (SVE_COND_FP_BINARY_REV): New int iterator.
2036 (commutative): New int attribute.
2037 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2039 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2041 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2042 (*cond_<optab><mode>): New patterns for reversed operands.
2044 2018-05-25 Richard Biener <rguenther@suse.de>
2046 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2047 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2048 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2049 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2050 * tree-vect-data-refs.c (everywhere): Adjust users.
2051 * tree-vect-loop.c (everywhere): Likewise.
2052 * tree-vect-slp.c (everywhere): Likewise.
2053 * tree-vect-stmts.c (everywhere): Likewise.
2054 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2056 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2058 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2060 (gcc_cv_as_section_exclude): ... this.
2061 Try Solaris as #exclude syntax.
2062 * configure: Regenerate.
2063 * config.in: Regenerate.
2064 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2066 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2067 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2069 * varasm.c (default_elf_asm_named_section): Don't check if
2070 HAVE_GAS_SECTION_EXCLUDE is defined.
2072 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2074 * doc/md.texi: Update the documentation of the cond_* optabs
2075 to mention the new final operand. Fix GET_MODE_NUNITS call.
2076 Describe the scalar case too.
2077 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2078 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2080 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2081 (get_conditional_internal_fn): Update comment.
2082 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2083 accumulator value as a final argument to conditional functions.
2084 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2085 a define_expand and add an "else" operand. Assert for now that
2086 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2087 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2088 (*cond_<optab><mode>): New patterns.
2089 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2090 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2091 (UNSPEC_COND_EOR): Delete.
2092 (optab): Remove associated mappings.
2093 (SVE_INT_BINARY): New code iterator.
2094 (sve_int_op): Remove int attribute and add "minus" to the code
2096 (SVE_COND_INT_OP): Delete.
2097 (SVE_COND_FP_OP): Rename to...
2098 (SVE_COND_FP_BINARY): ...this.
2100 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2102 * optabs.c (can_reuse_operands_p): New function.
2103 (maybe_legitimize_operands): Try to reuse the results for
2106 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2108 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2109 Add {q} suffix to insn mnemonic.
2111 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2113 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2114 (msp430_warn_func_return): New.
2116 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2118 * fold-const.c (tree_nonzero_bits): New function.
2119 * fold-const.h (tree_nonzero_bits): Likewise.
2120 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2121 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2123 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2127 * varasm.c (assemble_alias): Check ifunc_resolver only on
2130 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2133 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2134 when memory input operand is handled.
2136 2018-05-24 Luis Machado <luis.machado@linaro.org>
2138 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2140 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2142 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2144 * match.pd: Delay FMA folds until after vectorization.
2146 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2149 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2150 address check not strict.
2152 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2154 * gimple-match.h (gimple_match_op): New class.
2155 (mprts_hook): Replace parameters with a gimple_match_op *.
2156 (maybe_build_generic_op): Likewise.
2157 (gimple_simplified_result_is_gimple_val): Replace parameters with
2158 a const gimple_match_op *.
2159 (gimple_simplify): Replace code_helper * and tree * parameters with
2160 a gimple_match_op * parameter.
2161 (gimple_resimplify1): Replace code_helper *, tree and tree *
2162 parameters with a gimple_match_op * parameter.
2163 (gimple_resimplify2): Likewise.
2164 (gimple_resimplify3): Likewise.
2165 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2166 parameters with a gimple_match_op * parameter.
2167 * gimple-match-head.c (gimple_simplify): Change prototypes of
2168 auto-generated functions to take a gimple_match_op * instead of
2169 separate code_helper * and tree * parameters. Make the same
2170 change in the top-level overload and update calls to the
2171 gimple_resimplify routines. Update calls to the auto-generated
2172 functions and to maybe_push_res_to_seq in the publicly-facing
2173 operation-specific gimple_simplify overloads.
2174 (gimple_match_op::MAX_NUM_OPS): Define.
2175 (gimple_resimplify1): Replace rcode and ops with a single res_op
2176 parameter. Update call to gimple_simplify.
2177 (gimple_resimplify2): Likewise.
2178 (gimple_resimplify3): Likewise.
2179 (mprts_hook): Replace parameters with a gimple_match_op *.
2180 (maybe_build_generic_op): Likewise.
2181 (build_call_internal): Replace type, nargs and ops with
2182 a gimple_match_op *.
2183 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2184 with a single gimple_match_op *. Update calls to mprts_hook,
2185 build_call_internal and gimple_simplified_result_is_gimple_val.
2186 Factor out code that is common to the tree_code and combined_fn cases.
2187 * genmatch.c (expr::gen_transform): Replace tem_code and
2188 tem_ops with a gimple_match_op called tem_op. Update calls
2189 to the gimple_resimplify functions and maybe_push_res_to_seq.
2190 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2191 res_ops. Update call to the gimple_resimplify functions.
2192 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2193 (decision_tree::gen): Make the functions take a gimple_match_op *
2194 called res_op instead of separate res_code and res_ops parameters.
2195 Update call accordingly.
2196 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2197 and ops with a single res_op parameter. Update calls to
2198 maybe_build_generic_op and maybe_push_res_to_seq.
2199 (fold_stmt_1): Update calls to gimple_simplify and
2200 replace_stmt_with_simplification.
2201 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2202 and gimple_simplified_result_is_gimple_val.
2203 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2205 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2206 with a gimple_match_op *.
2207 (vn_nary_build_or_lookup): Likewise. Update call to
2208 vn_nary_build_or_lookup_1.
2209 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2210 gimple_match_op *. Update calls to the gimple_resimplify routines
2211 and to gimple_simplified_result_is_gimple_val.
2212 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2213 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2214 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2215 (visit_nary_op): Likewise.
2216 (visit_reference_op_load): Likewise.
2218 2018-05-23 Luis Machado <luis.machado@linaro.org>
2220 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2221 modifier for printing the step amount.
2223 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2226 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2229 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2231 * doc/sourcebuild.texi (Endianness): New subsubsection.
2233 2018-05-23 Luis Machado <luis.machado@linaro.org>
2235 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2236 <prefetch_dynamic_strides>: New const bool field.
2237 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2238 prefetch_dynamic_strides.
2239 (exynosm1_prefetch_tune): Likewise.
2240 (thunderxt88_prefetch_tune): Likewise.
2241 (thunderx_prefetch_tune): Likewise.
2242 (thunderx2t99_prefetch_tune): Likewise.
2243 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2245 (aarch64_override_options_internal): Update to set
2246 PARAM_PREFETCH_DYNAMIC_STRIDES.
2247 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2248 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2249 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2250 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2251 prefetch-dynamic-strides setting.
2253 2018-05-23 Luis Machado <luis.machado@linaro.org>
2255 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2256 <minimum_stride>: New const int field.
2257 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2258 minimum_stride field defaulting to -1.
2259 (exynosm1_prefetch_tune): Likewise.
2260 (thunderxt88_prefetch_tune): Likewise.
2261 (thunderx_prefetch_tune): Likewise.
2262 (thunderx2t99_prefetch_tune): Likewise.
2263 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2264 <default_opt_level>: Set to 3.
2265 (aarch64_override_options_internal): Update to set
2266 PARAM_PREFETCH_MINIMUM_STRIDE.
2267 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2268 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2269 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2270 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2271 stride is constant and is below the minimum stride threshold.
2273 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2275 * config/arm/arm-cpus.in (mode26): Delete.
2276 (armv4): Delete mode26 reference.
2277 * config/arm/arm.c (arm_configure_build_target): Delete use of
2280 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2282 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2284 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2285 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2287 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2288 (floatunsdidf2): Ditto.
2290 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2292 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2293 (fixuns_trunc<mode>si2_avx512f): Ditto.
2294 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2295 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2296 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2298 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2300 PR rtl-optimization/79985
2301 * df-scan.c (df_insn_refs_collect): Remove special case for
2302 global registers and asm statements.
2304 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2306 * extend.texi (Global Register Variables): Rewrite the bullet list.
2307 Note that the register is available for allocation. Note that access
2308 via inline asm must use constraints. Add note about async-signal
2309 handlers. Remove paragraph about automagic register selection.
2311 2018-05-23 Richard Biener <rguenther@suse.de>
2313 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2314 of fixed offset from memset VN.
2316 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2318 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2320 (alloc_cand_and_find_basis): Initialize first_interp field.
2321 (slsr_process_mul): Modify first_interp field.
2322 (slsr_process_add): Likewise.
2323 (slsr_process_cast): Modify first_interp field for each new
2325 (slsr_process_copy): Likewise.
2326 (dump_candidate): Dump first_interp field.
2327 (replace_mult_candidate): Process all interpretations, not just
2329 (replace_rhs_if_not_dup): Likewise.
2330 (replace_one_candidate): Likewise.
2332 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2334 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2336 (aarch64_needs_frame_chain): New function.
2337 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2339 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2342 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2343 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2344 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2345 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2346 as true for strict-align.
2347 (aarch64_can_inline_p): Perform checks even when callee has no
2348 attributes to check for strict alignment.
2349 * doc/extend.texi (AArch64 Function Attributes): Document
2351 * doc/invoke.texi: (AArch64 Options): Likewise.
2353 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2355 PR tree-optimization/85853
2356 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2357 the handling of the root of the node to...
2358 (vect_slp_analyze_node_operations_1): ...this new function,
2359 and run the whole thing with the child nodes' def types
2360 set according to their SLP node's def type.
2362 2018-05-23 Richard Biener <rguenther@suse.de>
2365 * tree-data-ref.c (create_runtime_alias_checks): Defer
2366 and ignore overflow warnings.
2368 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2370 PR tree-optimization/85822
2371 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2374 2018-05-23 Richard Biener <rguenther@suse.de>
2376 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2377 memset constants via native_interpret_expr.
2379 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2382 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2384 (cgraph_node::create_alias): Likewise.
2385 (cgraph_node::get_availability): Check ifunc_resolver instead
2386 of looking up ifunc attribute.
2387 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2388 * varasm.c (do_assemble_alias): Likewise.
2389 (assemble_alias): Likewise.
2390 (default_binds_local_p_3): Likewise.
2391 * cgraph.h (cgraph_node): Add ifunc_resolver.
2392 (cgraph_node::only_called_directly_or_aliased_p): Return false
2394 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2396 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2397 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2398 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2399 instead of looking up ifunc attribute.
2401 2018-05-22 Luis Machado <luis.machado@linaro.org>
2403 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2405 2018-05-22 Martin Sebor <msebor@redhat.com>
2408 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2409 only when expasion succeeds.
2410 (expand_builtin_strcmp): Same.
2411 (expand_builtin_strncmp): Same.
2413 2018-05-22 Martin Sebor <msebor@redhat.com>
2415 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2417 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2418 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2420 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2421 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2423 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2424 Add check for legitimate_address.
2425 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2426 (aarch64_swap_ldrstr_operands): New.
2427 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2430 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2431 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2433 * config/aarch64/aarch64.md: New patterns to generate stp
2435 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2436 single words and double words.
2437 (load_pair_sw, load_pair_dw): Likewise.
2438 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2440 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2442 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2443 for different mode ldpstp and add peephole for merged zero stores.
2445 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2447 (aarch64_gen_store_pair): Rename calls to match new patterns.
2448 (aarch64_gen_load_pair): Rename calls to match new patterns.
2449 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2450 (load_pair<DREG:mode><DREG2:mode>): ... This.
2451 (store_pair<mode>): Rename to...
2452 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2453 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2455 (V_INT_EQUIV): Handle SImode.
2456 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2459 2018-05-22 Martin Sebor <msebor@redhat.com>
2462 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2463 or ajust the presumed bound on an operation to avoid unnecessary
2466 2018-05-22 Martin Sebor <msebor@redhat.com>
2468 PR tree-optimization/85826
2469 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2470 assuming that a DECL necesarily has a constant size.
2472 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2475 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2477 2018-05-22 Richard Biener <rguenther@suse.de>
2479 PR tree-optimization/85834
2480 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2481 non-constant and non-zero memset arguments.
2483 2018-05-22 Martin Liska <mliska@suse.cz>
2486 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2488 2018-05-22 Richard Biener <rguenther@suse.de>
2490 PR tree-optimization/85863
2491 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2492 comparisons when vectype is specified.
2493 (vectorizable_condition): Do not specify vectype for
2494 vect_is_simple_cond when SLP vectorizing.
2496 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2499 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2500 define __ibm128 as long double.
2501 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2502 as a distinct type when IEEE 128-bit support is enabled.
2503 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2504 128-bit types to use the correct functions.
2505 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2506 convert between 128-bit floating point types that have different
2507 modes but the same representation, instead of using gen_lowpart to
2509 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2511 (IFKF_reg): New attributes to give the register constraints for
2513 (extend<mode>tf2_internal): New insns to mark an explicit
2514 conversion between 128-bit floating point types that have a
2515 different mode but share the same representation.
2517 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2519 PR tree-optimization/85814
2520 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2521 a null return from get_strinfo when unsharing the next
2522 strinfo in the chain.
2524 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2527 * varasm.c (weak_finish): Clean up weak_decls.
2529 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2531 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2532 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2533 UNSPEC_UADALP values.
2534 * config/aarch64/iterators.md (ABAL): New int iterator.
2537 (sur): Add mappings for the above.
2538 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2540 (aarch64_<sur>abal<mode>_4): Likewise.
2541 (aarch64_<sur>adalp<mode>_3): Likewise.
2542 (<sur>sadv16qi): New define_expand.
2544 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2546 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2547 (*movdf_internal): Ditto.
2548 (*rcpsf2_sse): Ditto.
2549 (*rsqrtsf2_sse): Ditto.
2550 (*sqrt<mode>2_sse): Ditto.
2552 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2554 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2556 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2557 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2558 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2559 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2561 * config/aarch64/arm_neon.h: Likewise.
2562 * config/aarch64/iterators.md (VQ_I): New.
2564 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2566 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2568 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2571 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2573 * config/nds32/constraints.md (S): New constraint.
2574 * config/nds32/nds32.md (call_internal): Use constraint S.
2575 (call_value_internal): Likewise.
2576 (sibcall_internal): Likewise.
2577 (sibcall_value_internal): Likewise.
2579 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2580 Chung-Ju Wu <jasonwucj@gmail.com>
2582 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2585 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2586 Chung-Ju Wu <jasonwucj@gmail.com>
2588 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2589 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2590 (nds32_rtx_costs_impl): Simplify.
2591 (nds32_address_cost_impl): Simplify.
2592 (nds32_init_rtx_costs): New function.
2593 (nds32_rtx_costs_speed_prefer): Likewise.
2594 (nds32_rtx_costs_size_prefer): Likewise.
2595 (nds32_address_cost_speed_prefer): Likewise.
2596 (nds32_address_cost_speed_fwprop): Likewise.
2597 (nds32_address_cost_size_prefer): Likewise.
2598 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2599 * config/nds32/nds32.c (nds32_option_override): Use
2600 nds32_init_rtx_costs function.
2602 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2604 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2605 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2606 (TARGET_PIPELINE_N8): Likewise.
2607 (TARGET_PIPELINE_N10): Likewise.
2608 (TARGET_PIPELINE_N13): Likewise.
2609 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2611 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2613 * config/nds32/nds32-fpu.md: Update copyright year.
2615 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2617 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2619 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2621 * config/nds32/nds32.c
2622 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2623 * config/nds32/nds32.opt (minline-asm-r15): New option.
2625 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2627 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2629 * config/nds32/nds32.md (abssi2): New pattern.
2631 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2633 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2634 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2635 Merge insn pattern from sse_cvtsi2ss<round_name> and
2636 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2637 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2638 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2639 using SWI48 mode iterator.
2640 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2641 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2642 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2643 pattern from sse_cvttss2si<round_saeonly_name>
2644 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2645 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2646 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2647 using SWI48 mode iterator.
2648 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2649 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2650 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2651 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2652 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2653 using SWI48 mode iterator.
2654 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2655 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2656 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2657 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2658 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2659 SWI48 mode iterator.
2660 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2661 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2662 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2663 pattern from sse_cvttsd2si<round_saeonly_name>
2664 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2666 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2668 * config/nds32/nds32-md-auxiliary.c
2669 (nds32_valid_smw_lwm_base_p): Refine.
2670 (nds32_output_smw_single_word): Refine.
2671 (nds32_output_smw_double_word): New.
2672 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2674 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2676 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2677 (nds32_output_stack_pop): Refine.
2678 (nds32_expand_unaligned_load): Refine.
2679 (nds32_expand_unaligned_store): Refine.
2681 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2682 Chung-Ju Wu <jasonwucj@gmail.com>
2684 * config/nds32/constants.md: Add TP_REGNUM constant.
2685 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2686 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2688 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2689 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2690 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2691 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2692 and PIC code generation.
2693 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2695 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2697 * config/nds32/nds32.md: Support TLS and PIC.
2698 * config/nds32/nds32.c: Support TLS and PIC.
2699 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2700 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2703 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2705 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2706 mode with E_ prefix.
2708 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2709 Chung-Ju Wu <jasonwucj@gmail.com>
2711 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2712 * config/nds32/nds32-md-auxiliary.c
2713 (symbolic_reference_mentioned_p): New.
2714 (nds32_legitimize_ict_address): New.
2715 (nds32_expand_ict_move): New.
2716 (nds32_indirect_call_referenced_p): New.
2717 (nds32_symbol_binds_local_p): Delete.
2718 (nds32_long_call_p): Modify.
2719 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2720 * config/nds32/nds32-protos.h
2721 (symbolic_reference_mentioned_p): Declare.
2722 (nds32_legitimize_ict_address): Declare.
2723 (nds32_expand_ict_move): Declare.
2724 (nds32_indirect_call_referenced_p): Declare.
2725 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2726 (nds32_relax_group): Use nds32_ict_const_p as condition.
2727 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2728 (nds32_asm_file_start): Output ict_model directive in asm code.
2729 (nds32_legitimate_address_p): Consider indirect call.
2730 (nds32_print_operand): Consider indirect call.
2731 (nds32_print_operand_address): Consider indirect call.
2732 (nds32_insert_attributes): Handle "indirect_call" attribute.
2733 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2734 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2735 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2736 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2737 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2738 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2739 (TARGET_ICT_MODEL_SMALL): Define.
2740 (TARGET_ICT_MODEL_LARGE): Define.
2741 * config/nds32/nds32.md (movsi): Consider ict model.
2742 (call, call_value): Consider ict model.
2743 (sibcall, sibcall_value): Consider ict model.
2744 * config/nds32/nds32.opt (mict-model): New option.
2745 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2748 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2749 Monk Chiang <sh.chiang04@gmail.com>
2750 Jim Wilson <jimw@sifive.com>
2752 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2753 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2754 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2755 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2756 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2757 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2758 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2759 compute save_libcall_adjustment properly.
2760 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2761 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2762 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2763 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2764 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2765 (ABI_SPEC): Handle mabi=ilp32e.
2766 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2767 (RVE): Add RVE mask.
2768 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2769 <-march>: Add rv32e as an example.
2771 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2774 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2775 (intra_create_variable_infos): Handle C++ constructors.
2777 2018-05-18 Martin Liska <mliska@suse.cz>
2779 * passes.def: Remove a redundant pass.
2781 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2784 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2786 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2788 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2790 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2792 (armv2, armv2a, armv3, armv3m): Delete architectures.
2793 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2794 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2795 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2797 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2798 (*mulsidi3adddi): Likewise.
2799 (mulsidi3): Likewise.
2800 (*mulsidi3_nov6): Likewise.
2801 (umulsidi3): Likewise.
2802 (umulsidi3_nov6): Likewise.
2803 (umaddsidi4): Likewise.
2804 (*umulsidi3adddi): Likewise.
2805 (smulsi3_highpart): Likewise.
2806 (*smulsi3_highpart_nov6): Likewise.
2807 (umulsi3_highpart): Likewise.
2808 (*umulsi3_highpart_nov6): Likewise.
2809 * config/arm/arm.h (arm_arch3m): Delete.
2810 * config/arm/arm.c (arm_arch3m): Delete.
2811 (arm_option_override_internal): Update armv3-related comment.
2812 (arm_configure_build_target): Delete use of isa_bit_mode32.
2813 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2814 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2815 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2817 (mulusa3): Likewise.
2818 * config/arm/arm-protos.h (arm_arch3m): Delete.
2819 * config/arm/arm-tables.opt: Regenerate.
2820 * config/arm/arm-tune.md: Likewise.
2821 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2822 deleted architectures.
2824 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2826 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2827 (armv5t, armv5te): New features.
2828 (ARMv5, ARMv5e): Delete fgroups.
2829 (ARMv5t, ARMv5te): Adjust for above changes.
2831 (armv5, armv5e): Delete arches.
2832 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2834 (*call_reg_arm): Likewise.
2835 (*call_value_reg_armv5): Likewise.
2836 (*call_value_reg_arm): Likewise.
2837 (*call_symbol): Likewise.
2838 (*call_value_symbol): Likewise.
2839 (*sibcall_insn): Likewise.
2840 (*sibcall_value_insn): Likewise.
2842 (prefetch): Likewise.
2843 (define_split and define_peephole2 dependent on arm_arch5):
2845 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2847 (TARGET_ARM_QBIT): Likewise.
2848 (TARGET_DSP_MULTIPLY): Likewise.
2849 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2850 (arm_arch5, arm_arch5e): Delete.
2851 (arm_arch5t, arm_arch5te): Declare.
2852 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2853 (arm_arch5t): Declare.
2854 (arm_option_reconfigure_globals): Update for the above.
2855 (arm_options_perform_arch_sanity_checks): Update comment, replace
2856 use of arm_arch5 with arm_arch5t.
2857 (use_return_insn): Likewise.
2858 (arm_emit_call_insn): Likewise.
2859 (output_return_instruction): Likewise.
2860 (arm_final_prescan_insn): Likewise.
2861 (arm_coproc_builtin_available): Likewise.
2862 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2863 arm_arch5e with arm_arch5t and arm_arch5te.
2864 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2865 (arm_arch5t, arm_arch5te): Declare.
2866 * config/arm/arm-tables.opt: Regenerate.
2867 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2868 * config/arm/t-multilib: Likewise.
2869 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2870 instead of arm_arch5.
2871 (*call_reg_thumb1): Likewise.
2872 (*call_value_reg_thumb1_v5): Likewise.
2873 (*call_value_reg_thumb1): Likewise.
2874 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2876 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2878 2018-05-18 Martin Liska <mliska@suse.cz>
2880 PR gcov-profile/84846
2881 * doc/gcov.texi: Document -t option of gcov tool.
2883 2018-05-18 Martin Liska <mliska@suse.cz>
2885 PR gcov-profile/84846
2886 * gcov.c (print_usage): Add new -t option.
2887 (process_args): Handle the option.
2888 (generate_results): Use stdout as output when requested by
2891 2018-05-18 Martin Liska <mliska@suse.cz>
2893 PR gcov-profile/84846
2894 * coverage.c (coverage_init): Write PWD to .gcno file.
2895 * doc/gcov.texi: Document how working directory is printed.
2896 * gcov-dump.c (dump_gcov_file): Print PWD.
2897 * gcov.c (output_intermediate_file): Likewise.
2898 (read_graph_file): Read PWD string.
2899 (output_lines): Print PWD.
2901 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2904 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2905 for retval and return false if all args to phi are zero.
2907 2018-05-18 Richard Biener <rguenther@suse.de>
2909 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2911 (evrp_dom_walker::before_dom_children): Call it.
2913 2018-05-18 Richard Biener <rguenther@suse.de>
2915 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2916 results when processing array refs with variable index.
2918 2018-05-18 Toon Moene <toon@moene.org>
2920 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2921 directly after that of -floop-interchange. Indicate that both
2922 options are enabled by default when specifying -O3.
2924 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2926 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2927 iterator. Delete separate integer-mode vec_set<mode> expander.
2928 (aarch64_simd_vec_setv2di): Delete.
2929 (vec_setv2di): Delete.
2930 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2931 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2932 the "w, r" alternative.
2934 2018-05-18 Martin Liska <mliska@suse.cz>
2936 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2937 * tree-pass.h (make_pass_lower_switch_O0): New function.
2938 * tree-switch-conversion.c (node_has_low_bound): Remove.
2939 (node_has_high_bound): Likewise.
2940 (node_is_bounded): Likewise.
2941 (class pass_lower_switch): Make it a template type and create
2943 (pass_lower_switch::execute): Add template argument.
2944 (make_pass_lower_switch): New function.
2945 (make_pass_lower_switch_O0): New function.
2946 (do_jump_if_equal): Remove.
2947 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2948 all the hard work to tree optimization passes.
2950 2018-05-18 Martin Liska <mliska@suse.cz>
2952 * dbgcnt.c (limit_low): Renamed from limit.
2953 (limit_high): New variable.
2954 (dbg_cnt_is_enabled): Check for upper limit.
2955 (dbg_cnt): Adjust dumping.
2956 (dbg_cnt_set_limit_by_index): Add new argument for high
2958 (dbg_cnt_set_limit_by_name): Likewise.
2959 (dbg_cnt_process_single_pair): Parse new format.
2960 (dbg_cnt_process_opt): Use strtok.
2961 (dbg_cnt_list_all_counters): Remove 'value' and add
2963 * doc/invoke.texi: Document changes.
2965 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2967 * doc/sourcebuild.texi (scalar_all_fma): Document.
2968 * tree.def (FMA_EXPR): Delete.
2969 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2970 * internal-fn.c (ternary_direct): New macro.
2971 (expand_ternary_optab_fn): Likewise.
2972 (direct_ternary_optab_supported_p): Likewise.
2973 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2974 * builtins.c (fold_builtin_fma): Delete.
2975 (fold_builtin_3): Don't call it.
2976 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2977 * expr.c (expand_expr_real_2): Likewise.
2978 * fold-const.c (operand_equal_p): Likewise.
2979 (fold_ternary_loc): Likewise.
2980 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2981 * gimple.c (DEFTREECODE): Likewise.
2982 * gimplify.c (gimplify_expr): Likewise.
2983 * optabs-tree.c (optab_for_tree_code): Likewise.
2984 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2985 * tree-eh.c (operation_could_trap_p): Likewise.
2986 (stmt_could_throw_1_p): Likewise.
2987 * tree-inline.c (estimate_operator_cost): Likewise.
2988 * tree-pretty-print.c (dump_generic_node): Likewise.
2989 (op_code_prio): Likewise.
2990 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2991 * tree-ssa-operands.c (get_expr_operands): Likewise.
2992 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2993 * fold-const-call.h (fold_fma): Delete.
2994 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2995 CFN_FNMA and CFN_FNMS.
2997 * genmatch.c (combined_fn): New enum.
2998 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2999 (commutative_op): New function.
3000 (commutate): Use it. Handle more than 2 operands.
3001 (dt_operand::gen_gimple_expr): Use commutative_op.
3002 (parser::parse_expr): Allow :c to be used with non-binary
3003 operators if the commutative operand is known.
3004 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3005 CFN_FMS, CFN_FNMA and CFN_FNMS.
3006 (backprop::process_assign_use): Remove FMA_EXPR handling.
3007 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3008 (gen_hsa_fma): New function.
3009 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3010 IFN_FNMA and IFN_FNMS.
3011 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3012 * gimple-fold.h (follow_all_ssa_edges): Declare.
3013 * gimple-fold.c (follow_all_ssa_edges): New function.
3014 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3015 gimple_build interface and use follow_all_ssa_edges to fold the result.
3016 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3017 instead of checking for optabs directly.
3018 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3019 rather than FMA_EXPRs.
3020 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3021 call to IFN_FMA instead of an FMA_EXPR.
3023 2018-05-17 Jim Wilson <jimw@sifive.com>
3025 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3026 sign extended promoted subreg, and the range does not have the sign bit
3027 set, then do a sign extend.
3029 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3030 test, check for sign extended subreg and/or constant operands, and
3031 do a sign extend in that case.
3033 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3035 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3036 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3038 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3039 Change logics_shift_reg to logics_shift_imm.
3040 (thunderx2t99_fp_loadpair_basic): Delete.
3041 (thunderx2t99_fp_storepair_basic): Delete.
3042 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3043 (thunderx2t99_asimd_polynomial): Delete.
3044 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3045 and neon_fp_mul_d_scalar_q.
3046 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3047 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3048 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3049 (thunderx2t99_asimd_lut): Add missing tbl types.
3050 (thunderx2t99_asimd_ext): Delete.
3051 (thunderx2t99_asimd_load1_1_mult): Delete.
3052 (thunderx2t99_asimd_load1_2_mult): Delete.
3053 (thunderx2t99_asimd_load1_ldp): New.
3054 (thunderx2t99_asimd_load1): New.
3055 (thunderx2t99_asimd_load2): Add missing *load2* types.
3056 (thunderx2t99_asimd_load3): New.
3057 (thunderx2t99_asimd_load4): New.
3058 (thunderx2t99_asimd_store1_1_mult): Delete.
3059 (thunderx2t99_asimd_store1_2_mult): Delete.
3060 (thunderx2t99_asimd_store2_mult): Delete.
3061 (thunderx2t99_asimd_store2_onelane): Delete.
3062 (thunderx2t99_asimd_store_stp): New.
3063 (thunderx2t99_asimd_store1): New.
3064 (thunderx2t99_asimd_store2): New.
3065 (thunderx2t99_asimd_store3): New.
3066 (thunderx2t99_asimd_store4): New.
3068 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3070 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3071 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3073 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3074 Segher Boessenkool <segher@kernel.crashing.org>
3077 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3080 2018-05-17 Richard Biener <rguenther@suse.de>
3082 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3083 for pruning loop and prune defs feeding only already visited PHIs.
3085 2018-05-17 Richard Biener <rguenther@suse.de>
3087 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3089 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3090 Richard Biener <rguenther@suse.de>
3092 PR tree-optimization/85793
3093 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3094 for VMAT_ELEMENTWISE.
3096 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3098 * internal-fn.h (lookup_internal_fn): Declare
3099 * internal-fn.c (lookup_internal_fn): New function.
3100 * gimple.c (gimple_build_call_from_tree): Handle calls to
3102 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3103 internal function names.
3104 * tree-pretty-print.c (dump_generic_node): Likewise.
3105 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3107 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3109 * gimple-fold.h (gimple_build): Make the function forms take
3110 combined_fn rather than built_in_function.
3111 (gimple_simplify): Likewise.
3112 * gimple-match-head.c (gimple_simplify): Likewise.
3113 * gimple-fold.c (gimple_build): Likewise.
3114 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3115 rather than gimple_build_call_internal.
3116 (get_initial_defs_for_reduction): Likewise.
3117 (vect_create_epilog_for_reduction): Likewise.
3118 (vectorizable_live_operation): Likewise.
3120 2018-05-17 Martin Liska <mliska@suse.cz>
3122 * gimple-ssa-sprintf.c (format_directive): Do not use
3123 space in between 'G_' and '('.
3125 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3128 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3129 even if the mask is not all ones.
3132 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3134 (ix86_gimple_fold_builtin): Likewise.
3137 * config/i386/i386.c: Include tree-vector-builder.h.
3138 (ix86_vector_shift_count): New function.
3139 (ix86_fold_builtin): Fold shift builtins by scalar count.
3140 (ix86_gimple_fold_builtin): Likewise.
3142 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3143 _mm512_setzero): New intrinsics.
3145 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3146 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3148 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3149 code generation for cases where splatting a value is not useful.
3150 * simplify-rtx.c (simplify_ternary_operation): Simplify
3151 vec_merge across a vec_duplicate and a paradoxical subreg forming
3152 a vector mode to a vec_concat.
3154 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3156 * config.gcc: Support "goldmont-plus".
3157 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3159 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3160 PROCESSOR_GOLDMONT_PLUS.
3161 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3162 (processor_target_table): Add "goldmont-plus".
3163 (PTA_GOLDMONT_PLUS): Define.
3164 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3165 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3166 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3167 (fold_builtin_cpu): Add "goldmont-plus".
3168 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3169 (ix86_option_override_internal): Add "goldmont-plus".
3170 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3171 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3172 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3173 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3175 2018-05-17 Richard Biener <rguenther@suse.de>
3177 PR tree-optimization/85757
3178 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3179 remove defs that only feed that PHI from further processing.
3181 2018-05-16 Jim Wilson <jimw@sifive.com>
3183 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3185 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3187 2018-05-16 Mark Wielaard <mark@klomp.org>
3189 * dwarf2out.c (count_index_strings): New function.
3190 (output_indirect_strings): Call count_index_strings and generate
3191 header for dwarf_version >= 5.
3193 2018-05-16 Mark Wielaard <mark@klomp.org>
3195 * dwarf2out.c (dwarf_FORM): New function.
3196 (set_indirect_string): Use dwarf_FORM.
3197 (reset_indirect_string): Likewise.
3198 (size_of_die): Likewise.
3199 (value_format): Likewise.
3200 (output_die): Likewise.
3201 (add_skeleton_AT_string): Likewise.
3202 (output_macinfo_op): Likewise.
3203 (index_string): Likewise.
3204 (output_index_string_offset): Likewise.
3205 (output_index_string): Likewise.
3206 (count_index_strings): Likewise.
3208 2018-05-16 Carl Love <cel@us.ibm.com>
3210 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3211 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3213 2018-05-16 Martin Jambor <mjambor@suse.cz>
3215 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3216 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3218 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3220 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3221 (fnma<mode>4): Likewise.
3222 (fms<mode>4): Likewise.
3223 (fnms<mode>4): Likewise.
3224 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3225 (aarch64_fnma<mode>4): Likewise.
3226 (aarch64_fms<mode>4): Likewise.
3227 (aarch64_fnms<mode>4): Likewise.
3228 (aarch64_fnmadd<mode>4): Likewise.
3230 2018-05-16 Jason Merrill <jason@redhat.com>
3232 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3234 2018-05-16 Richard Biener <rguenther@suse.de>
3236 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3237 (dump_stmt_cost): Declare.
3238 (add_stmt_cost): Dump cost we add.
3239 (add_stmt_costs): New function.
3240 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3242 (vect_analyze_stmt): Adjust prototype.
3243 (vectorizable_condition): Likewise.
3244 (vectorizable_live_operation): Likewise.
3245 (vectorizable_reduction): Likewise.
3246 (vectorizable_induction): Likewise.
3247 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3248 cost vector to pass to vectorizable_ and record afterwards.
3249 (vect_model_reduction_cost): Take cost vector argument and adjust.
3250 (vect_model_induction_cost): Likewise.
3251 (vectorizable_reduction): Likewise.
3252 (vectorizable_induction): Likewise.
3253 (vectorizable_live_operation): Likewise.
3254 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3255 SLP_TREE_NUMBER_OF_VEC_STMTS.
3256 (vect_analyze_slp_cost_1): Remove.
3257 (vect_analyze_slp_cost): Likewise.
3258 (vect_slp_analyze_node_operations): Take visited args and
3259 a target cost vector. Avoid processing already visited stmt sets.
3260 (vect_slp_analyze_operations): Use a local cost vector to gather
3261 costs and register those of non-discarded instances.
3262 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3263 (vect_schedule_slp_instance): Remove copying of
3264 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3266 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3267 adding cost. Record cost entry location.
3268 (vect_prologue_cost_for_slp_op): Function to compute cost of
3269 a constant or invariant generated for SLP vect in the prologue,
3270 split out from vect_analyze_slp_cost_1.
3271 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3272 (vect_model_promotion_demotion_cost): Likewise.
3273 (vect_model_store_cost): Likewise, make static.
3274 (vect_model_load_cost): Likewise.
3275 (vectorizable_bswap): Add cost vector arg and adjust.
3276 (vectorizable_call): Likewise.
3277 (vectorizable_simd_clone_call): Likewise.
3278 (vectorizable_conversion): Likewise.
3279 (vectorizable_assignment): Likewise.
3280 (vectorizable_shift): Likewise.
3281 (vectorizable_operation): Likewise.
3282 (vectorizable_store): Likewise.
3283 (vectorizable_load): Likewise.
3284 (vectorizable_condition): Likewise.
3285 (vectorizable_comparison): Likewise.
3286 (can_vectorize_live_stmts): Likewise.
3287 (vect_analyze_stmt): Likewise.
3288 (vect_transform_stmt): Adjust calls to vectorizable_*.
3289 * tree-vectorizer.c: Include gimple-pretty-print.h.
3290 (dump_stmt_cost): New function.
3292 2018-05-16 Richard Biener <rguenther@suse.de>
3294 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3295 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3296 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3297 (check_name): New callback.
3298 (dse_classify_store): Track cycles via a visited bitmap of PHI
3299 defs and simplify handling of in-loop and across loop dead stores
3300 and properly fail for loop-variant refs. Handle byte-tracking with
3301 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3304 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3306 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3307 (vect_get_mask_type_for_stmt): Likewise.
3308 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3310 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3311 to determine the statement's vector type and the vector type that
3312 should be used for calculating nunits. Deal with cases in which
3313 the type has to be deferred.
3314 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3315 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3316 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3317 (vect_determine_vf_for_stmt): New functions, split out from...
3318 (vect_determine_vectorization_factor): ...here.
3319 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3320 (vect_get_mask_type_for_stmt): New functions, split out from
3321 vect_determine_vectorization_factor.
3323 2018-05-16 Richard Biener <rguenther@suse.de>
3325 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3326 verify the [VEC_]COND_EXPR embedded comparison.
3328 2018-05-15 Martin Sebor <msebor@redhat.com>
3330 PR tree-optimization/85753
3331 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3332 RECORD_TYPE in addition to ARRAY_TYPE.
3334 2018-05-15 Martin Sebor <msebor@redhat.com>
3337 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3339 2018-05-15 Richard Biener <rguenther@suse.de>
3341 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3342 add by_clobber_p one. Change algorithm to collect all defs
3343 representing uses we need to walk and try reducing them to
3344 a single one before failing.
3345 (dse_dom_walker::dse_optimize_stmt): Adjust.
3347 2018-05-13 Mark Wielaard <mark@klomp.org>
3349 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3350 (size_of_loc_descr): Likewise.
3351 (output_loc_operands): Likewise.
3352 (output_loc_operands_raw): Likewise.
3353 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3354 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3355 (hash_loc_operands): Likewise.
3356 (compare_loc_operands): Likewise.
3358 2018-05-14 Mark Wielaard <mark@klomp.org>
3360 * dwarf2out.c (count_index_addrs): New function.
3361 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3363 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3365 PR tree-optimization/83648
3366 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3367 return value as malloc candidate.
3369 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3372 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3373 param as true in call to suggest_attribute.
3375 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3377 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3378 -mreadonly-in-sdata.
3380 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3382 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3384 (aarch64_crypto_aesd_fused): Likewise.
3386 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3388 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3389 (movsi_aarch64): Likewise.
3390 (load_pairsi): Likewise.
3391 (load_pairdi): Likewise.
3392 (store_pairsi): Likewise.
3393 (store_pairdi): Likewise.
3394 (load_pairsf): Likewise.
3395 (load_pairdf): Likewise.
3396 (store_pairsf): Likewise.
3397 (store_pairdf): Likewise.
3398 (zero_extend): Likewise.
3399 (trunc): Swap alternatives.
3400 (fcvt_target): Add '?' to prefer w over r.
3402 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3405 * config/i386/i386.md: Disallow non-commutative arithmetics in
3406 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3407 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3408 in the peephole2 before it.
3410 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3412 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3413 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3414 (ix86_handle_option): Handle -mcldemote.
3415 * config.gcc: New header.
3416 * config/i386/cldemoteintrin.h: New file.
3417 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3418 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3420 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3421 OPTION_MASK_ISA_CLDEMOTE.
3422 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3423 (ix86_valid_target_attribute_inner_p): Ditto.
3424 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3425 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3426 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3427 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3428 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3430 * config/i386/i386.opt: Add -mcldemote.
3431 * config/i386/x86intrin.h: New header.
3432 * doc/invoke.texi: Add -mcldemote.
3434 2018-05-14 Richard Biener <rguenther@suse.de>
3436 * doc/match-and-simplify.texi: Adjust :s documentation.
3438 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3440 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3441 intended memcpy size.
3442 (REORDER_45): Likewise.
3444 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3446 * sort.cc: New file.
3447 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3448 * vec.c (qsort_chk): Use gcc_qsort.
3449 * Makefile.in (OBJS-libcommon): Add sort.o.
3450 (build/sort.o): New target. Use it...
3451 (BUILD_RTL): ... here, and...
3452 (build/gencfn-macros): ... here, and...
3453 (build/genmatch): ... here.
3455 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3456 Chung-Ju Wu <jasonwucj@gmail.com>
3458 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3459 * config/nds32/nds32-graywolf.md: New file.
3460 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3461 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3463 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3464 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3465 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3466 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3467 * config/nds32/pipelines.md: Include n15 settings.
3469 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3470 Chung-Ju Wu <jasonwucj@gmail.com>
3472 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3473 * config/nds32/nds32-n13.md: New file.
3474 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3475 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3477 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3478 * config/nds32/nds32.md (pipeline_model): Add n13.
3479 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3480 * config/nds32/pipelines.md: Include n13 settings.
3482 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3483 Chung-Ju Wu <jasonwucj@gmail.com>
3485 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3486 * config/nds32/nds32-n10.md: New file.
3487 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3488 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3490 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3491 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3492 * config/nds32/nds32.md (pipeline_model): Add n10.
3493 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3494 * config/nds32/pipelines.md: Include n10 settings.
3496 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3497 Kito Cheng <kito.cheng@gmail.com>
3498 Chung-Ju Wu <jasonwucj@gmail.com>
3500 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3501 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3502 Add enum values for DSP extension instructions.
3503 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3505 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3506 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3508 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3509 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3510 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3511 * config/nds32/nds32-intrinsic.md: Likewise.
3512 * config/nds32/nds32_intrinsic.h: Likewise.
3513 * config/nds32/nds32-md-auxiliary.c: Likewise.
3514 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3515 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3516 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3517 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3518 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3519 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3520 TYPE_DMAC in switch statement.
3521 * config/nds32/nds32.c: New checking and implementation for DSP
3522 extension instructions.
3523 * config/nds32/nds32.h: Likewise.
3524 * config/nds32/nds32.md: Likewise.
3525 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3526 * config/nds32/predicates.md: Implement new predicates for DSP
3529 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3531 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3532 Reformat alternatives and attributes so it is easier to identify
3533 which constraints/attributes go with which instruction.
3534 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3535 (mov<mode>_softfloat32, FMOVE64): Likewise.
3536 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3537 (mov<mode>_softfloat64, FMOVE64): Likewise.
3539 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3541 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3543 (Basic PowerPC Built-in Functions): The new name of the
3544 subsection previously known as "PowerPC Built-in Functions".
3545 (Basic PowerPC Built-in Functions Available on all Configurations):
3547 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3548 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3549 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3550 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3552 2018-05-11 Martin Jambor <mjambor@suse.cz>
3555 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3558 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3561 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3563 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3565 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3566 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3567 (ix86_handle_option): Handle -mwaitpkg.
3568 * config.gcc: New header.
3569 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3570 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3571 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3573 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3574 OPTION_MASK_ISA_WAITPKG.
3575 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3576 (ix86_option_override_internal): Add PTA_WAITPKG.
3577 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3578 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3579 IX86_BUILTIN_TPAUSE.
3580 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3581 __builtin_ia32_umwait and __builtin_ia32_tpause.
3582 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3583 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3584 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3585 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3586 UNSPECV_TPAUSE): New.
3587 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3588 * config/i386/i386.opt: Add -mwaitpkg.
3589 * config/i386/waitpkgintrin.h: New file.
3590 * config/i386/x86intrin.h: New header.
3591 * doc/invoke.texi: Add -mwaitpkg.
3593 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3596 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3598 (cortex-m0): Use armv6s-m isa.
3599 (cortex-m0plus): Likewise.
3600 (cortex-m1): Likewise.
3601 (cortex-m0.small-multiply): Likewise.
3602 (cortex-m0plus.small-multiply): Likewise.
3603 (cortex-m1.small-multiply): Likewise.
3605 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3606 Jakub Jelinek <jakub@redhat.com>
3608 PR tree-optimization/85692
3609 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3610 source permute as well.
3612 2018-05-11 Martin Liska <mliska@suse.cz>
3615 * doc/extend.texi: Document LLVM style format for no_sanitize
3618 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3620 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3621 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3622 (mode_supports_vsx_dform_quad): Likewise.
3623 (mode_supports_vmx_dform): Move these functions to be next to the
3624 other mode_supports functions.
3625 (mode_supports_dq_form): Likewise.
3626 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3627 mode_supports_dq_form.
3628 (reg_offset_addressing_ok_p): Likewise.
3629 (offsettable_ok_by_alignment): Likewise.
3630 (rs6000_legitimate_offset_address_p): Likewise.
3631 (legitimate_lo_sum_address_p): Likewise.
3632 (rs6000_legitimize_address): Likewise.
3633 (rs6000_legitimize_reload_address): Likewise.
3634 (rs6000_secondary_reload_inner): Likewise.
3635 (rs6000_preferred_reload_class): Likewise.
3636 (rs6000_output_move_128bit): Likewise.
3638 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3640 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3641 Generate SImode target register for null target.
3642 <case IX86_BUILTIN_XGETBV>: Ditto.
3643 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3644 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3646 2018-05-10 Carl Love <cel@us.ibm.com>
3648 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3649 dcbtt and dcbtstt if operands[2] is 0.
3651 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3654 * config/i386/sse.md (usadv64qi): New expander.
3656 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3658 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3659 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3660 -maltivec=be support.
3661 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3662 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3663 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3664 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3665 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3666 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3667 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3668 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3669 altivec_vsumsws): Adjust.
3670 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3671 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3672 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3674 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3675 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3676 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3677 (altivec_lve<VI_char>x): Delete expand.
3678 (*altivec_lve<VI_char>x_internal): Rename to...
3679 (altivec_lve<VI_char>x): ... this.
3680 (altivec_lvxl_<mode>): Delete expand.
3681 (*altivec_lvxl_<mode>_internal): Rename to ...
3682 (altivec_lvxl_<mode>): ... this.
3683 (altivec_stvxl_<mode>): Delete expand.
3684 (*altivec_stvxl_<mode>_internal): Rename to ...
3685 (altivec_stvxl_<mode>): ... this.
3686 (altivec_stve<VI_char>x): Delete expand.
3687 (*altivec_stve<VI_char>x_internal): Rename to ...
3688 (altivec_stve<VI_char>x): ... this.
3689 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3690 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3691 reduc_plus_scal_<mode>): Adjust.
3692 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3694 (rs6000_cpu_cpp_builtins): Adjust.
3695 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3696 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3697 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3698 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3699 -maltivec=be support.
3700 (rs6000_split_vec_extract_var): Adjust.
3701 (rs6000_split_v4si_init): Adjust.
3702 (swap_selector_for_mode): Delete.
3703 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3704 altivec_expand_stvex_be): Delete.
3705 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3706 -maltivec=be support.
3707 (rs6000_gimple_fold_builtin): Ditto.
3708 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3710 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3711 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3712 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3713 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3714 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3715 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3716 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3717 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3718 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3719 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3720 anonymous split): Adjust.
3721 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3722 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3724 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3726 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3727 when --with-gxx-include-dir is also specified.
3728 * configure: Regenerate.
3730 2018-05-09 Jim Wilson <jimw@sifive.com>
3733 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3734 * config/riscv/t-withmultilib: New.
3735 * config/riscv/withmultilib.h: New.
3736 * doc/install.texi: Document RISC-V --with-multilib-list support.
3738 2018-05-09 Richard Biener <rguenther@suse.de>
3740 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3742 (vect_bb_vectorization_profitable_p): Adjust. Compute
3743 actual scalar cost using the cost vector and the add_stmt_cost
3746 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3748 PR rtl-optimization/85645
3749 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3750 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3752 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3754 PR rtl-optimization/85645
3755 * shrink-wrap.c (spread_components): Return a boolean saying if
3756 anything was changed.
3757 (try_shrink_wrapping_separate): Iterate spread_components until
3758 nothing changes anymore.
3760 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3762 PR rtl-optimization/85645
3763 * regrename.c (build_def_use): Also kill the chains that include the
3764 destination of a REG_CFA_REGISTER note.
3766 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3768 PR rtl-optimization/85645
3769 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3770 insn that has a REG_CFA_REGISTER note.
3772 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3774 * cfgexpand.c (expand_clobber): New function.
3775 (expand_gimple_stmt_1): Use it.
3776 * tree-vect-stmts.c (vect_clobber_variable): New function,
3778 (vectorizable_simd_clone_call): ...here.
3779 (vectorizable_store): Emit a clobber either side of an
3780 IFN_STORE_LANES sequence.
3781 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3783 2018-05-09 Tom de Vries <tom@codesourcery.com>
3786 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3787 (define_insn "trap_if_false"): Add exit after trap.
3789 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3791 PR rtl-optimization/85638
3792 * bb-reorder.c: Include common/common-target.h.
3793 (create_forwarder_block): New function extracted from...
3794 (fix_up_crossing_landing_pad): ...here. Rename into...
3795 (dw2_fix_up_crossing_landing_pad): ...this.
3796 (sjlj_fix_up_crossing_landing_pad): New function.
3797 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3798 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3799 from both partitions and exit the loop after one iteration.
3801 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3804 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3806 (Basic PowerPC Built-in Functions): The new name of the
3807 subsection previously known as "PowerPC Built-in Functions".
3808 (Basic PowerPC Built-in Functions Available on all Configurations):
3810 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3812 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3814 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3816 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3819 2018-05-08 Jim Wilson <jimw@sifive.com>
3821 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3822 (LD_EMUL_SUFFIX): New.
3823 (LINK_SPEC): Use it.
3825 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3827 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3829 (Basic PowerPC Built-in Functions): The new name of the
3830 subsection previously known as "PowerPC Built-in Functions".
3831 (Basic PowerPC Built-in Functions Available on all Configurations):
3833 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3835 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3837 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3839 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3842 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3845 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3846 after cmpelim optimization.
3848 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3850 * config.gcc: Support "goldmont".
3851 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3852 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3854 * config/i386/i386.c (m_GOLDMONT): Define.
3855 (processor_target_table): Add "goldmont".
3856 (PTA_GOLDMONT): Define.
3857 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3858 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3859 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3860 (fold_builtin_cpu): Add "goldmont".
3861 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3862 (ix86_option_override_internal): Add "goldmont".
3863 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3864 (processor_type): Add PROCESSOR_GOLDMONT.
3865 * config/i386/i386.md: Add CPU "glm".
3866 * config/i386/glm.md: New file.
3867 * config/i386/x86-tune.def: Add m_GOLDMONT.
3868 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3870 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3873 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3875 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3876 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3877 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3880 * config/i386/i386.c (ix86_fold_builtin): Handle
3881 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3884 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3885 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3887 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3890 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3891 (check_arch): Likewise.
3892 (check_fpu): Return the result rather than printing it.
3893 (end arch): Fix operator precedence.
3894 (end cpu): Likewise.
3895 (END): Print the result from check_fpu.
3897 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3898 Alan Hayward <alan.hayward@arm.com>
3899 David Sherwood <david.sherwood@arm.com>
3901 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3902 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3903 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3904 (*fcmuo<mode>_and): New patterns.
3906 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3908 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3909 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3910 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3911 (cmp_op, sve_imm_con): New code attributes.
3912 (SVE_COND_INT_CMP, imm_con): Delete.
3913 (cmp_op): Remove above unspecs from int attribute.
3914 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3916 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3917 comparison-specific unspecs.
3918 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3919 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3920 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3921 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3922 (*vec_fcm<cmp_op><mode>): Rename to...
3923 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3924 (*vec_fcmuo<mode>): Rename to...
3925 (*fcmuo<mode>): ...this and adjust likewise.
3926 (*pred_fcm<cmp_op><mode>): New pattern.
3927 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3928 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3930 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3932 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3933 (aarch64_emit_sve_predicated_cond): New function.
3934 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3935 (aarch64_emit_unspec_cond_or): Replace with...
3936 (aarch64_emit_sve_or_conds): ...this new function. Use
3937 aarch64_emit_sve_ptrue_op for the individual comparisons and
3938 aarch64_emit_binop to OR them together.
3939 (aarch64_emit_inverted_unspec_cond): Replace with...
3940 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3941 aarch64_emit_sve_ptrue_op for the comparison and
3942 aarch64_emit_unop to invert the result.
3943 (aarch64_expand_sve_vec_cmp_float): Update after the above
3944 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3946 2018-05-07 Nathan Sidwell <nathan@acm.org>
3948 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3949 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3950 (Backwards Compatibility): Likewise.
3952 2018-05-07 Luis Machado <luis.machado@linaro.org>
3956 2018-05-07 Luis Machado <luis.machado@linaro.org>
3958 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3959 <prefetch_dynamic_strides>: New const bool field.
3960 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3961 prefetch_dynamic_strides.
3962 (exynosm1_prefetch_tune): Likewise.
3963 (thunderxt88_prefetch_tune): Likewise.
3964 (thunderx_prefetch_tune): Likewise.
3965 (thunderx2t99_prefetch_tune): Likewise.
3966 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3968 (aarch64_override_options_internal): Update to set
3969 PARAM_PREFETCH_DYNAMIC_STRIDES.
3970 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3971 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3972 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3973 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3974 prefetch-dynamic-strides setting.
3976 2018-05-07 Luis Machado <luis.machado@linaro.org>
3978 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3979 <minimum_stride>: New const int field.
3980 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3981 minimum_stride field.
3982 (exynosm1_prefetch_tune): Likewise.
3983 (thunderxt88_prefetch_tune): Likewise.
3984 (thunderx_prefetch_tune): Likewise.
3985 (thunderx2t99_prefetch_tune): Likewise.
3986 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3987 (aarch64_override_options_internal): Update to set
3988 PARAM_PREFETCH_MINIMUM_STRIDE.
3989 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3990 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3991 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3992 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3993 stride is constant and is below the minimum stride threshold.
3995 2018-05-07 Luis Machado <luis.machado@linaro.org>
3997 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4000 2018-05-07 Luis Machado <luis.machado@linaro.org>
4002 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4003 <prefetch_dynamic_strides>: New const bool field.
4004 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4005 prefetch_dynamic_strides.
4006 (exynosm1_prefetch_tune): Likewise.
4007 (thunderxt88_prefetch_tune): Likewise.
4008 (thunderx_prefetch_tune): Likewise.
4009 (thunderx2t99_prefetch_tune): Likewise.
4010 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4012 (aarch64_override_options_internal): Update to set
4013 PARAM_PREFETCH_DYNAMIC_STRIDES.
4014 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4015 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4016 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4017 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4018 prefetch-dynamic-strides setting.
4020 2018-05-07 Luis Machado <luis.machado@linaro.org>
4022 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4023 <minimum_stride>: New const int field.
4024 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4025 minimum_stride field.
4026 (exynosm1_prefetch_tune): Likewise.
4027 (thunderxt88_prefetch_tune): Likewise.
4028 (thunderx_prefetch_tune): Likewise.
4029 (thunderx2t99_prefetch_tune): Likewise.
4030 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4031 (aarch64_override_options_internal): Update to set
4032 PARAM_PREFETCH_MINIMUM_STRIDE.
4033 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4034 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4035 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4036 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4037 stride is constant and is below the minimum stride threshold.
4039 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4042 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4043 the type is addressable. Don't force op into register if it has
4046 2018-05-05 Roland McGrath <mcgrathr@google.com>
4049 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4050 any section for which we don't know a specific type it should have,
4051 regardless of name. Previously this was done only for the exact
4052 names ".init_array", ".fini_array", and ".preinit_array".
4053 (default_elf_asm_named_section): Add comment about
4054 relationship with default_section_type_flags and SECTION_NOTYPE.
4055 (get_section): Don't consider it a type conflict if one side has
4056 SECTION_NOTYPE and the other doesn't, as long as neither has the
4057 SECTION_BSS et al used in the default_section_type_flags logic.
4059 2018-05-05 Tom de Vries <tom@codesourcery.com>
4062 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4063 (workaround_barsyncs): New function.
4064 (nvptx_reorg): Use workaround_barsyncs.
4065 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4066 (define_expand "nvptx_membar_cta"): New define_expand.
4067 (define_insn "*nvptx_membar_cta"): New insn.
4069 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4071 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4072 To improve optimization opportunities.
4073 * builtin-types.def: The new needed builtin types for the above.
4075 2018-05-04 Richard Biener <rguenther@suse.de>
4077 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4078 * gimple-ssa-store-merging.c
4079 (imm_store_chain_info::output_merged_store): Remove redundant create,
4080 release split_store vector contents on failure.
4081 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4082 scalar stmt vector on cache hit.
4084 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4086 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4088 * config.gcc (powerpc-xilinx-eabi*): Remove.
4089 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4091 (fusion_addis_mem_combo_load): Ditto.
4092 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4094 (rs6000_cpu_cpp_builtins): Ditto.
4095 * config/rs6000/rs6000-linux.c
4096 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4097 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4098 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4100 (rs6000_setup_reg_addr_masks): Ditto.
4101 (rs6000_init_hard_regno_mode_ok): Ditto.
4102 (rs6000_option_override_internal): Ditto.
4103 (legitimate_lo_sum_address_p): Ditto.
4104 (rs6000_legitimize_address): Ditto.
4105 (rs6000_legitimize_reload_address): Ditto.
4106 (rs6000_legitimate_address_p): Ditto.
4107 (abi_v4_pass_in_fpr): Ditto.
4108 (setup_incoming_varargs): Ditto.
4109 (rs6000_gimplify_va_arg): Ditto.
4110 (rs6000_split_multireg_move): Ditto.
4111 (rs6000_savres_strategy): Ditto.
4112 (rs6000_emit_prologue_components): Ditto.
4113 (rs6000_emit_epilogue_components): Ditto.
4114 (rs6000_emit_prologue): Ditto.
4115 (rs6000_emit_epilogue): Ditto.
4116 (rs6000_elf_file_end): Ditto.
4117 (rs6000_function_value): Ditto.
4118 (rs6000_libcall_value): Ditto.
4119 * config/rs6000/rs6000.h: Ditto.
4120 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4121 (TARGET_MINMAX): ... this. New.
4122 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4123 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4124 (*movsi_internal1_single): Delete.
4125 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4126 mfpu=, mxilinx-fpu): Delete.
4127 * config/rs6000/singlefp.h: Delete.
4128 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4129 * config/rs6000/t-rs6000: Ditto.
4130 * config/rs6000/t-xilinx: Delete.
4131 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4132 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4133 (VStype_simple): Delete.
4134 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4135 * config/rs6000/xfpu.h: Delete.
4136 * config/rs6000/xfpu.md: Delete.
4137 * config/rs6000/xilinx.h: Delete.
4138 * config/rs6000/xilinx.opt: Delete.
4139 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4140 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4142 2018-05-04 Tom de Vries <tom@codesourcery.com>
4145 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4148 2018-05-04 Richard Biener <rguenther@suse.de>
4151 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4152 (expand_complex_div_wide): Likewise.
4153 (expand_complex_operations_1): Likewise.
4154 (expand_complex_libcall): Preserve EH info of the original stmt.
4155 (tree_lower_complex): Handle removed blocks.
4156 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4157 on complex multiplication and division libcall builtins.
4159 2018-05-04 Richard Biener <rguenther@suse.de>
4162 * fold-const.c (negate_expr_p): Restrict negation of operand
4163 zero of a division to when we know that can happen without
4165 (fold_negate_expr_1): Likewise.
4167 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4170 * real.h (real_nextafter): Declare.
4171 * real.c (real_nextafter): New function.
4172 * fold-const-call.c (fold_const_nextafter): New function.
4173 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4174 CASE_CFN_NEXTTOWARD.
4175 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4176 even when arg1_mode is different from arg0_mode.
4178 2018-05-03 Nathan Sidwell <nathan@acm.org>
4180 * doc/extend.texi (Deprecated Features): Remove
4182 (Backwards Compatibility): Likewise.
4183 * doc/invoke.texi (C++ Language Options): Likewise.
4184 (C++ Dialect Options): Likewise.
4186 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4189 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4190 _mm512_mask_mullox_epi64): New intrinsics.
4192 2018-05-03 Tom de Vries <tom@codesourcery.com>
4195 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4196 dump files): Add offload-tree.
4198 2018-05-03 Richard Biener <rguenther@suse.de>
4200 PR tree-optimization/85615
4201 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4202 to loops not nested in BBs loop father to avoid creating multi-entry
4205 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4207 PR tree-optimization/70291
4208 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4209 arguments. Change return type to tree. Emit libcall as a new
4210 statement rather than replacing existing one when inplace_p is true.
4211 (expand_complex_multiplication_components): New function.
4212 (expand_complex_multiplication): Expand floating-point complex
4213 multiplication using the above.
4214 (expand_complex_division): Rename inner_type parameter to type.
4215 Update expand_complex_libcall call-site.
4216 (expand_complex_operations_1): Update expand_complex_multiplication
4217 and expand_complex_division call-sites.
4219 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4222 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4223 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4224 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4225 the highest significant bit of the shift count mask is clear. In
4226 check whether and[sq]i3 is needed verify that all significant bits
4227 of the shift count other than the highest are set.
4229 2018-05-02 Tom de Vries <tom@codesourcery.com>
4232 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4233 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4234 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4235 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4236 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4237 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4238 __builtin_goacc_parlevel_size.
4240 2018-05-02 Richard Biener <rguenther@suse.de>
4242 PR tree-optimization/85597
4243 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4244 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4247 2018-05-02 Tom de Vries <tom@codesourcery.com>
4250 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4251 dump files): Add ltrans-tree.
4253 2018-05-02 Tom de Vries <tom@codesourcery.com>
4256 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4257 dump files): Add wpa-ipa.
4259 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4261 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4262 powerpc*-*-linux*paired* target.
4263 * config/rs6000/750cl.h: Delete.
4264 * config/rs6000/paired.h: Delete.
4265 * config/rs6000/paired.md: Delete.
4266 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4268 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4269 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4270 comment. Remove paired float support.
4271 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4272 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4274 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4275 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4277 * config/rs6000/rs6000.c: Remove paired float support.
4278 (paired_expand_vector_init, paired_expand_vector_move,
4279 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4280 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4281 paired_expand_builtin, paired_expand_predicate_builtin,
4282 paired_init_builtins): Delete.
4283 * config/rs6000/rs6000.h: Remove paired float support.
4284 * config/rs6000/rs6000.md: Remove paired float support.
4285 (move_from_CR_ov_bit): Delete.
4286 * config/rs6000/rs6000.opt (mpaired): Delete.
4287 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4288 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4290 2018-05-02 Richard Biener <rguenther@suse.de>
4293 * gimplify.c (gimplify_save_expr): When in SSA form allow
4294 SAVE_EXPRs to compute to SSA vars.
4296 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4299 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4300 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4301 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4302 clobber operands[2], instead use a new pseudo. Formatting fixes.
4304 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4306 PR tree-optimization/85586
4307 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4308 exit early for statements in the same group if the accesses are
4311 2018-05-02 Tom de Vries <tom@codesourcery.com>
4314 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4317 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4319 PR tree-optimization/85143
4320 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4322 2018-05-01 Tom de Vries <tom@codesourcery.com>
4325 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4326 not found" error message.
4328 2018-05-01 Tom de Vries <tom@codesourcery.com>
4331 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4332 * vec.c (test_ordered_remove_if): New function.
4333 (vec_c_tests): Call test_ordered_remove_if.
4334 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4335 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4336 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4337 VEC_ORDERED_REMOVE_IF.
4339 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4341 PR tree-optimization/82665
4342 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4343 pointer subtraction where arguments come from a memchr call.
4345 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4347 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4348 --push-state --as-needed and --pop-state instead of --as-needed and
4349 --no-as-needed if ld supports it.
4350 * configure: Regenerated.
4353 * doc/install.texi2html: Replace _002d with - and _002a with * in
4354 generated html files using sed.
4356 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4359 * gcc-rich-location.c (blank_line_before_p): New function.
4360 (use_new_line): New function.
4361 (gcc_rich_location::add_fixit_insert_formatted): New function.
4362 * gcc-rich-location.h
4363 (gcc_rich_location::add_fixit_insert_formatted): New function.
4365 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4367 * selftest.c (assert_streq): Rename "expected" and "actual" to
4368 "val1" and "val2". Extend NULL-handling to cover both inputs
4369 symmetrically, while still requiring both to be non-NULL for a pass.
4370 * selftest.h (assert_streq): Rename "expected" and "actual" to
4372 (ASSERT_EQ): Likewise.
4373 (ASSERT_EQ_AT): Likewise.
4374 (ASSERT_KNOWN_EQ): Likewise.
4375 (ASSERT_KNOWN_EQ_AT): Likewise.
4376 (ASSERT_NE): Likewise.
4377 (ASSERT_MAYBE_NE): Likewise.
4378 (ASSERT_MAYBE_NE_AT): Likewise.
4379 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4380 the assertion to pass.
4381 (ASSERT_STREQ_AT): Likewise.
4383 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4385 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4386 interaction with -pie.
4388 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4390 * selftest.h: Fix alphabetization of per-source-file selftest
4393 2018-04-30 Jason Merrill <jason@redhat.com>
4395 PR c++/61982 - dead stores to destroyed objects.
4396 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4399 2018-04-30 Jason Merrill <jason@redhat.com>
4401 * tree.c (build_clobber): New.
4402 * tree.h: Declare it.
4403 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4405 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4407 * diagnostic-show-locus.c (layout::layout): Update for
4408 location_get_source_line returning a char_span.
4409 (struct char_span): Move to input.h.
4410 (struct correction): Update for fields in char_span becoming
4412 (struct source_line): Update for location_get_source_line
4413 returning a char_span.
4414 (layout::print_line): Likewise.
4415 * edit-context.c (edited_file::print_content): Likewise.
4416 (edited_file::print_diff_hunk): Likewise.
4417 (edited_file::print_run_of_changed_lines): Likewise.
4418 (edited_file::get_num_lines): Likewise.
4419 (edited_line::edited_line): Likewise.
4420 * final.c (asm_show_source): Likewise.
4421 * input.c (location_get_source_line): Convert return type
4422 from const char * to char_span, losing the final "line_len"
4424 (dump_location_info): Update for the above.
4425 (get_substring_ranges_for_loc): Likewise. Use a char_span
4426 when handling the literal within the line.
4427 (test_reading_source_line): Update for location_get_source_line
4428 returning a char_span.
4429 * input.h (class char_span): Move here from
4430 diagnostic-show-locus.c, converting from a struct to a class.
4431 Make data members private.
4432 (char_span::operator bool): New.
4433 (char_span::length): New.
4434 (char_span::get_buffer): New.
4435 (char_span::operator[]): New.
4436 (char_span::subspan): Make const.
4437 (char_span::xstrdup): New.
4438 (location_get_source_line): Convert return type from const char *
4439 to char_span, losing the final "line_size" param.
4441 2018-04-30 Jan Hubicka <jh@suse.cz>
4443 * lto-wrapper.c (ltrans_priorities): New static var.
4444 (cmp_priority): New.
4445 (run_gcc): Read priorities and if doing parallel build order
4446 the Makefile by them.
4448 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4450 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4452 2018-04-30 Richard Biener <rguenther@suse.de>
4454 * tree-cfg.c (verify_address): Remove base argument, add
4455 flag whether to check TREE_ADDRESSABLE and do that.
4456 (verify_expr): Remove.
4457 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4458 (verify_gimple_assign_single): Likewise.
4459 (verify_gimple_switch): Likewise.
4460 (verify_expr_location_1): Dereference tp once. Add (disabled)
4461 piece from verify_expr.
4462 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4464 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4466 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4468 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4470 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4471 (small_data_pattern): Likewise.
4472 (arc_rewrite_small_data): Likewise.
4473 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4474 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4475 (get_symbol_alignment): New function.
4476 (legitimate_small_data_address_p): Likewise.
4477 (legitimate_scaled_address): Update, call
4478 legitimate_small_data_address_p.
4479 (output_sdata): New static variable.
4480 (arc_print_operand): Update how we handle small data operands.
4481 (arc_print_operand_address): Likewise.
4482 (arc_legitimate_address_p): Update, use
4483 legitimate_small_data_address_p.
4484 (arc_rewrite_small_data_p): Remove.
4485 (arc_rewrite_small_data_1): Likewise.
4486 (arc_rewrite_small_data): Likewise.
4487 (small_data_pattern): Likewise.
4488 (compact_sda_memory_operand): Update to use
4489 legitimate_small_data_address_p and get_symbol_alignment.
4490 (prepare_move_operands): Don't rewite sdata pattern.
4491 (prepare_extend_operands): Remove.
4492 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4494 (zero_extendqisi2): Likewise.
4495 (zero_extendhisi2): Likewise.
4496 (extendqihi2): Likewise.
4497 (extendqisi2): Likewise.
4498 (extendhisi2): Likewise.
4502 * config/arc/constraints.md (Usd): Change it to memory constraint.
4504 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4506 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4507 as source of std instructions.
4508 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4509 allow 6-bit constants as source for store instructions.
4510 (movdi_insn): Update instruction pattern to allow 6-bit constants
4511 as source for store instructions.
4513 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4515 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4517 2018-04-30 Nathan Sidwell <nathan@acm.org>
4518 Sandra Loosemore <sandra@codesourcery.com>
4520 * dumpfile.c (dump_open): Allow '-' for stdout.
4521 * doc/invoke.texi (Developer Options): Document dump filename
4522 determination early. Document stdin/stdout selection.
4524 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4526 Microblaze Target: PIC data text relative
4528 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4529 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4531 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4532 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4533 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4534 New addressing mode for data-text relative position indepenedent code.
4535 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4536 'ADDRESS_SYMBOLIC_TXT_REL'.
4537 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4538 (microblaze_legitimate_pic_operand): Exclude function calls from
4539 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4540 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4542 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4543 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4544 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4545 for 'address + offset'.
4546 (microblaze_expand_prologue): Add new function prologue call for
4548 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4549 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4550 table in case of TARGET_PIC_DATA_TEXT_REL.
4551 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4552 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4553 Add new macros 'UNSPEC_TEXT',
4554 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4555 + exclude function calls from 'UNSPEC_PLT' in case of data text
4557 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4558 new target hook for generating address diff vector tables in case of
4560 * doc/tm.texi : Regenerate.
4561 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4562 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4563 of addr diff vector generation.
4564 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4565 target hook definition.
4566 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4567 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4568 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4569 option for data text relative.
4571 2018-04-30 Richard Biener <rguenther@suse.de>
4573 * tree-chrec.h (evolution_function_is_constant_p): Remove
4575 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4577 2018-04-30 Richard Biener <rguenther@suse.de>
4580 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4582 2018-04-30 Richard Biener <rguenther@suse.de>
4584 PR tree-optimization/28364
4585 PR tree-optimization/85275
4586 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4587 copying first exit test.
4589 2018-04-28 Mark Wielaard <mark@klomp.org>
4591 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4593 (dwarf_AT): Handle DW_AT_addr_base.
4594 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4596 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4599 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4600 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4601 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4602 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4604 2018-04-28 Richard Biener <rguenther@suse.de>
4606 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4607 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4608 to reflect use. Only add interesting stmts.
4610 2018-04-27 Martin Jambor <mjambor@suse.cz>
4613 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4614 the jump function allows for passing through aggregate values.
4616 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4618 * input.h (in_system_header_at): Convert from macro to inline
4620 (from_macro_expansion_at): Likewise.
4621 (from_macro_definition_at): Likewise.
4623 2018-04-27 Jeff Law <law@redhat.com>
4625 * config.gcc: Mark tile* targets as deprecated/obsolete.
4627 2018-04-27 Richard Biener <rguenther@suse.de>
4629 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4632 2018-04-27 Richard Biener <rguenther@suse.de>
4634 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4636 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4638 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4639 with Yd constraint. Set "preferred_for_speed" attribute from
4640 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4642 (*movdi_internal): Ditto.
4643 (movti_interunit splitters): Remove
4644 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4645 (movdi_interunit splitters): Ditto.
4646 * config/i386/constraints.md (Ye): Remove.
4647 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4649 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4652 * config/aarch64/constraints.md (Usg): Limit to 31.
4655 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4657 PR tree-optimization/85529
4658 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4659 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4660 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4661 zero extension or masking of the MSB bit.
4662 (optimize_range_tests): Add FIRST_BB argument, pass it through
4663 to optimize_range_tests_var_bound.
4664 (maybe_optimize_range_tests, reassociate_bb): Adjust
4665 optimize_range_tests callers.
4667 2018-04-26 Richard Biener <rguenther@suse.de>
4668 Jakub Jelinek <jakub@redhat.com>
4670 * cgraph.h (symbol_table): Just declare debug method here.
4671 * symtab.c (symbol_table::debug): Define.
4673 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4675 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4677 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4679 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4680 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4681 (*movdi_internal): Substitute Yi and Yj constraint with x
4682 and Ym and Yn constraint with y constraint. Update "isa"
4683 attribute and set "preferred_for_speed" attribute from
4684 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4685 (*movsi_internal): Ditto.
4686 (*movdf_internal): Ditto.
4687 (*movsf_internal): Ditto.
4688 (*zero_extendsidi2): Ditto.
4689 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4690 (sse2_loadld): Ditto.
4691 (*vec_extract<ssevecmodelower>_0): Ditto.
4692 (*vec_extractv4si_0_zext_sse4): Ditto.
4693 (vec_concatv2di): Ditto.
4694 (*vec_dup<mode>): Ditto.
4695 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4696 * config/i386/constraints.md (Yi): Remove.
4701 2018-04-26 Nathan Sidwell <nathan@acm.org>
4703 * dumpfile.c (dump_open): New.
4704 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4705 (dump_finish): Detect stdio/stderr by value not name.
4707 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4709 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4711 2018-04-26 Tom de Vries <tom@codesourcery.com>
4714 * config/nvptx/nvptx.c (verify_neutering_jumps)
4715 (verify_neutering_labels): New function
4716 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4718 2018-04-26 Tom de Vries <tom@codesourcery.com>
4721 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4722 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4725 2018-04-26 Richard Biener <rguenther@suse.de>
4726 Tom de Vries <tom@codesourcery.com>
4729 * lto-streamer-out.c (output_function): Fixup loops if required to match
4730 discovery done in the reader.
4732 2018-04-26 Richard Biener <rguenther@suse.de>
4734 PR tree-optimization/85116
4735 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4736 have a loop exit from the single latch predecessor. Remove
4737 case of header with just condition.
4738 (ch_base::copy_headers): Exclude infinite loops from any
4740 (pass_ch::execute): Record exits.
4742 2018-04-26 Richard Biener <rguenther@suse.de>
4744 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4745 prologue cost vector and pass it to vect_get_load_cost.
4746 (vect_get_peeling_costs_all_drs): Likewise.
4747 (vect_peeling_hash_get_lowest_cost): Likewise.
4748 (vect_enhance_data_refs_alignment): Likewise.
4750 2018-04-26 Richard Biener <rguenther@suse.de>
4753 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4754 checking of integer<->pointer conversions.
4755 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4756 sign-/zero-extending pointer types.
4757 (expand_omp_for_static_chunk): Likewise.
4759 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4760 Jean Lee <xiaoyur347@gmail.com>
4762 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4763 (TARGET_ASAN_SHADOW_OFFSET): Define.
4764 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4765 true for -fsanitize=address.
4767 2018-04-25 Mark Wielaard <mark@klomp.org>
4769 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4772 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4774 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4775 than "alu", remove explicit "memory" and "imm_disp" attributes.
4776 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4779 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4780 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4781 gen_lowpart_no_emit.
4783 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4786 * config/i386/i386.c (ix86_expand_builtin): Change memory
4787 operand to XI, extend p0 to Pmode.
4788 * config/i386/i386.md: Change unspec volatile and operand
4789 1 mode to XI, change operand 0 mode to P.
4791 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4793 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4794 GET_MODE_MASK before any checking.
4795 (nds32_can_use_bset_p): Likewise.
4796 (nds32_can_use_btgl_p): Likewise.
4798 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4800 * config/nds32/nds32-doubleword.md: New define_split pattern for
4801 illegal register number.
4803 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4805 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4807 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4809 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4811 2018-04-25 Richard Biener <rguenther@suse.de>
4813 * lto-streamer.h (LTO_major_version): Bump to 8.
4815 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4817 * BASE-VER: Set to 9.0.0.
4819 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4821 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4822 in __abskf2 and __powikf2.
4824 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4827 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4828 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4829 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4830 Use the above on operand 2. Reindent.
4831 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4833 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4836 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4838 * config/i386/i386.opt (mcet): Removed.
4839 * doc/install.texi: Remove -mcet documentation.
4840 * doc/invoke.texi: Likewise.
4842 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4845 * doc/install.texi: Remove -mcet from bootstrap-cet.
4847 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4850 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4851 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4855 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4856 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4857 containing a CONST_VECTOR.
4859 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4861 * doc/install.texi: Update newlib dependency for nvptx.
4863 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4866 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4867 instead of INTVAL when shifting x left.
4869 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4871 PR tree-optimization/85478
4872 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4873 vect_grouped_store_supported for single element vectors.
4875 2018-04-24 Richard Biener <rguenther@suse.de>
4878 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4879 load cost increase to the case of non-constant step.
4881 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4884 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4885 destination if any_malformed_asm.
4887 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4890 * expr.c (store_field): In the bitfield case, if the value comes from
4891 a function call and is returned in registers by means of a PARALLEL,
4892 do not change the mode of the temporary unless BLKmode and VOIDmode.
4894 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4896 PR rtl-optimization/85423
4897 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4898 dependencies to debug insns when the previous insn is non-debug.
4900 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4902 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4903 enums into a single definition.
4904 (fls): Fix predicates and printing.
4907 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4909 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4910 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4911 and short u6 immediate.
4912 (check_if_valid_sleep_operand): Remove.
4913 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4915 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4917 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4918 flag_always_save_lp condition.
4919 * config/nds32/nds32.opt (malways-save-lp): New option.
4921 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4923 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4924 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4925 * config/nds32/nds32.h
4926 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4927 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4929 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4931 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4932 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4934 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4935 Chung-Ju Wu <jasonwucj@gmail.com>
4937 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4939 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4940 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4942 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4944 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4946 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4948 * config/nds32/nds32-protos.h (nds32_data_alignment,
4949 nds32_local_alignment): Declare.
4950 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4951 nds32_local_alignment): New functions.
4952 (TARGET_CONSTANT_ALIGNMENT): Define.
4953 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4955 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4957 * config/nds32/nds32.c
4958 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4959 (TARGET_MODES_TIEABLE_P): Likewise.
4961 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4963 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4966 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4967 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4970 for unaligned access.
4971 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4973 * config/nds32/nds32-intrinsic.md: Likewise.
4974 * config/nds32/nds32_intrinsic.h: Likewise.
4975 * config/nds32/nds32.h (nds32_builtins): Likewise.
4976 * config/nds32/nds32.opt (munaligned-access): New option.
4977 * config/nds32/nds32.c (nds32_asm_file_start): Display
4978 flag_unaligned_access status.
4980 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4982 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4983 -mno-relax is present.
4984 * config/riscv/linux.h (LINK_SPEC): Ditto.
4986 2018-04-20 Martin Sebor <msebor@redhat.com>
4989 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4991 (gimple_fold_builtin_stxcpy_chk): Same.
4992 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4994 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4997 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4998 __powikf2 when long double is IEEE 128-bit.
5000 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5002 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5003 step to make sure stack always aligned.
5005 2018-04-20 Carl Love <cel@us.ibm.com>
5008 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5009 size check for arg0.
5011 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5012 Tom de Vries <tom@codesourcery.com>
5015 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5016 Emit insns for calls too.
5017 (nvptx_find_par): Always look for worker-level predecessor insn.
5018 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5020 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5021 (nvptx_process_pars): Propagate frames for calls.
5023 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5026 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5028 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5029 (ix86_handle_option): Don't handle OPT_mibt.
5030 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5032 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5034 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5035 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5036 (ix86_target_macros): Define __CET__ with flag_cf_protection
5037 for -fcf-protection.
5038 * config/i386/i386.c (isa2_opts): Remove -mibt.
5039 * config/i386/i386.h (TARGET_IBT): Removed.
5040 (TARGET_IBT_P): Likewise.
5041 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5042 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5043 * config/i386/i386.opt (mcet): Update help message.
5046 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5047 -mcet as an alias for -mshstk.
5049 2018-04-20 Richard Biener <rguenther@suse.de>
5052 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5053 complexity by forcing a single use of the multiply operand.
5055 2018-04-20 Martin Jambor <mjambor@suse.cz>
5058 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5059 recursion dependency to only apply to non-clones.
5061 2018-04-20 Martin Jambor <mjambor@suse.cz>
5064 * ipa-cp.c (create_specialized_node): Check that clones of
5065 self-recursive edges exist during IPA-CP.
5067 2018-04-19 Toon Moene <toon@moene.org>
5069 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5072 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5074 PR tree-optimization/85467
5075 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5076 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5077 VECTOR_CST element to type.
5079 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5082 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5083 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5084 (builtin_longjmp): Likewise.
5085 (save_stack_nonlocal): New pattern.
5086 (restore_stack_nonlocal): Likewise.
5088 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5091 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5092 Replace ASM_OUTPUT_LABEL with fprintf.
5094 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5097 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5098 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5099 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5100 define __IBT__ and __SHSTK__ for -fcf-protection.
5101 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5103 (ix86_trampoline_init): Likewise.
5104 (x86_output_mi_thunk): Likewise.
5105 (ix86_notrack_prefixed_insn_p): Likewise.
5106 (ix86_option_override_internal): Don't disallow -fcf-protection.
5107 * config/i386/i386.md (rdssp<mode>): Also enable for
5109 (incssp<mode>): Likewise.
5110 (nop_endbr): Likewise.
5111 * config/i386/i386.opt (mcet): Change help message to built-in
5115 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5116 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5117 enable CET built-in functions.
5119 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5121 * common/config/i386/i386-common.c
5122 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5123 OPTION_MASK_ISA_MOVDIRI_UNSET,
5124 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5125 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5126 * config.gcc (movdirintrin.h): New header.
5127 * config/i386/cpuid.h (bit_MOVDIRI,
5128 bit_MOVDIR64B): New bits.
5129 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5131 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5132 (VOID, PVOID, PCVOID)): New function types.
5133 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5134 __builtin_ia32_directstoreu_u64,
5135 __builtin_ia32_movdir64b): New builtins.
5136 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5137 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5139 (ix86_valid_target_attribute_inner_p): Ditto.
5140 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5141 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5142 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5143 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5144 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5145 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5146 (movdiri<mode>, movdir64b_<mode>): New.
5147 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5148 * config/i386/immintrin.h: Include movdirintrin.h.
5149 * config/i386/movdirintrin.h: New file.
5150 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5152 2018-04-19 Richard Biener <rguenther@suse.de>
5155 * cfg.c (clear_bb_flags): When loop state says we have
5156 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5158 2018-04-19 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/84737
5161 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5162 copying restrict info.
5163 (vect_setup_realignment): Use it.
5164 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5165 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5166 the first DR to all generated stores.
5167 (vectorizable_load): Likewise for loads.
5169 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5171 PR tree-optimization/85446
5172 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5173 the integral and pointer types to have the same precision.
5175 * doc/install.texi: Document --disable-cet being the default and
5178 2018-04-18 Martin Liska <mliska@suse.cz>
5180 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5183 2018-04-18 Martin Liska <mliska@suse.cz>
5186 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5189 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5190 arguments if they are comparable.
5192 2018-04-18 Martin Liska <mliska@suse.cz>
5195 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5198 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5201 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5204 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5205 ENDBR after calling __morestack.
5207 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5210 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5211 by using gcc_base_ver to generate a gcc_driver_version, and use
5212 it when generating GCC_DRIVER_NAME.
5213 * configure: Regenerate.
5215 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5218 * config.gcc: Obsolete powerpc*-*-*spe*.
5220 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5223 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5224 (stabstr_D): Change type of unum from unsigned int to
5225 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5228 2018-04-17 Jim Wilson <jimw@sifive.com>
5231 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5232 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5233 Set arg_pointer_offset after using pretend_args_size.
5235 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5237 PR rtl-optimization/85431
5238 * dse.c (record_store): Ignore zero width stores.
5241 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5242 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5243 __builtin_stack_restore rather than after it.
5244 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5245 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5246 argument instead of virtual_dynamic_stack_rtx.
5248 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5250 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5252 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5253 Add note to error message to explain internal mapping of overloaded
5254 built-in function name to non-overloaded built-in function name.
5255 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5258 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5261 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5262 where the inputs overlap with the output.
5264 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5266 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5267 (=v, v) alternative and explicit "memory" attribute.
5268 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5269 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5271 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5272 "sselog1" type instead of "sselog".
5273 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5274 "sselog". Remove explicit "memory" attribute.
5275 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5276 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5278 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5279 "sselog1" type instead of "sselog". Remove explicit "memory"
5281 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5282 use "sselog1" type instead of "sselog". Remove explicit "memory"
5284 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5285 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5287 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5288 "sselog1" type instead of "sselog". Remove explicit "memory"
5290 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5291 use "sselog1" type instead of "sselog". Remove explicit "memory"
5295 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5298 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5301 2018-04-17 Martin Jambor <mjambor@suse.cz>
5304 * ipa-cp.c (create_specialized_node): Call
5305 expand_all_artificial_thunks if necessary.
5307 2018-04-17 Martin Liska <mliska@suse.cz>
5310 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5311 in message, remote space in between '_G' and '('.
5313 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5316 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5317 avx512f_vmcmp<mode>3<round_saeonly_name>,
5318 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5319 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5320 avx512f_rndscale<mode><round_saeonly_name>,
5321 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5322 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5323 Use %<iptr>2 instead of %2 for -masm=intel.
5324 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5325 avx512f_vcvttss2usi<round_saeonly_name>,
5326 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5328 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5329 avx512f_vcvttsd2usi<round_saeonly_name>,
5330 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5331 Use %q1 instead of %1 for -masm=intel.
5332 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5333 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5334 of %3 for -masm=intel.
5335 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5337 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5339 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5341 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5342 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5344 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5346 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5347 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5348 %g1 and one with %0 and %1.
5349 (avx512er_vmrcp28<mode><round_saeonly_name>,
5350 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5352 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5353 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5354 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5355 of %0 and %{%4%} for -masm=intel.
5356 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5357 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5358 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5359 order of %0 and %{%5%}%{z%} for -masm=intel.
5361 2018-04-17 Jan Hubicka <jh@suse.cz>
5364 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5366 2018-04-17 Martin Liska <mliska@suse.cz>
5369 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5370 for target_clone error message. Make default implementation
5371 clone to be a local declaration.
5372 (separate_attrs): Add new argument and check for an empty
5374 (expand_target_clones): Handle it.
5375 (ipa_target_clone): Make redirection just for target_clones
5378 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5379 Tom de Vries <tom@codesourcery.com>
5382 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5383 tiled basic blocks without omp continue statements.
5385 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5388 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5389 vec_extract expression as having side effects to make sure it gets
5392 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5395 * config/i386/i386.c (get_builtin_code_for_version): Check
5398 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5401 * gcc/config.gcc: Support "skylake".
5402 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5404 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5405 (processor_target_table): Add "skylake".
5406 (ix86_option_override_internal): Add "skylake".
5407 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5408 PROCESSOR_CANNONLAKE.
5409 (get_builtin_code_for_version): Fix priority for
5410 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5411 PROCESSOR_SKYLAKE-AVX512.
5412 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5413 (processor_type): Add PROCESSOR_SKYLAKE.
5415 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5416 Jason Merrill <jason@redhat.com>
5419 * convert.c (convert_to_integer_1): Use direct recursion for
5420 enumeral types and types with a precision less than the number
5421 of bits in their mode.
5423 2018-04-16 Julia Koval <julia.koval@intel.com>
5426 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5427 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5429 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5432 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5433 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5434 and -mno-direct-move.
5436 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5439 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5440 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5441 Ensure negative shifts result in {0}.
5443 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5445 PR rtl-optimization/79916
5446 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5447 regs (if any) to define how to gnerate SD moves when LRA is in
5450 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5452 PR rtl-optimization/85393
5453 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5454 * except.c (expand_dw2_landing_pad_for_region): Make static.
5455 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5456 a label and unconditional jump to old_bb, rather than
5457 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5460 PR rtl-optimization/85376
5461 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5462 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5463 instead of a specific value.
5465 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5466 Bin Cheng <bin.cheng@arm.com>
5468 PR tree-optimization/82965
5469 PR tree-optimization/83991
5470 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5471 by_profile_only parameter.
5472 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5473 information if the loop was predicted to iterate too many times.
5474 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5476 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5479 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5482 2018-04-13 Martin Liska <mliska@suse.cz>
5483 Jakub Jelinek <jakub@redhat.com>
5486 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5487 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5488 * builtins.c (expand_builtin_memory_copy_args): Use
5489 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5490 handle dest_addr == pc_rtx.
5492 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5495 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5496 asked to not generate direct moves.
5497 (fix_trunc<mode>si2_stfiwx): Similar.
5498 (fix_trunc<mode>si2_internal): Similar.
5500 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5503 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5504 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5505 lookup if dest in some wider mode is known to be const0_rtx and
5506 if so, record permanent equivalence for it to be ZERO_EXTEND of
5507 the narrower mode destination.
5509 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5511 * lto-streamer-out.c (output_function): Revert 259346.
5512 * omp-expand.c (expand_oacc_for): Likewise.
5514 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5516 PR rtl-optimization/85354
5517 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5518 * sel-sched.c (sel_global_init): ... here.
5520 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5523 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5524 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5525 mode for PE-COFF targets.
5526 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5527 (i386_pe_asm_lto_end): Likewise.
5528 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5529 (TARGET_ASM_LTO_END): Likewise.
5530 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5531 (i386_pe_asm_lto_start): New function.
5532 (i386_pe_asm_lto_end): Likewise.
5534 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5535 Richard Biener <rguenther@suse.de>
5538 * lto-streamer-out.c (output_function): Fix CFG loop state before
5540 * omp-expand.c (expand_oacc_for): Handle calls to internal
5541 functions like regular functions.
5543 2018-04-12 Richard Biener <rguenther@suse.de>
5546 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5547 for the early LTO debug to properly generate references to it
5548 during DIE emission. Do not re-use that for the skeleton for
5550 (dwarf2out_early_finish): Likewise.
5552 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5555 * config/i386/sse.md
5556 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5557 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5558 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5559 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5560 and output is a reg, avoid creating invalid lowpart subreg, but
5561 instead split into a 512-bit move. Don't split if not AVX512VL,
5562 input is xmm16+ reg and output is a mem.
5563 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5564 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5565 xmm16+ reg and output is a mem.
5567 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5569 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5570 also for flag_dwarf2_cfi_asm.
5572 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5574 PR rtl-optimization/85342
5575 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5576 a bool scalar var inside of the loop instead. Don't try to update
5577 recog_data.operand after failed apply_change_group.
5579 2018-04-12 Tom de Vries <tom@codesourcery.com>
5582 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5583 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5584 array with flexible array member as array without given dimension.
5585 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5586 argument for undefined param to true.
5588 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5591 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5592 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5593 from PowerPC section.
5594 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5595 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5596 help text that is too long.
5597 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5598 help text that is too long.
5599 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5600 help text that is too long.
5602 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5604 * config/alpha/alpha.md (stack_probe_internal): Rename
5605 from "probe_stack". Update all callers.
5607 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5609 PR rtl-optimization/84566
5610 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5611 sched_macro_fuse_insns.
5613 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5616 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5617 (compute_block_dependences): ... from here.
5619 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5621 PR tree-optimization/85331
5622 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5623 from int to HOST_WIDE_INT.
5625 2018-04-11 Martin Jambor <mjambor@suse.cz>
5628 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5629 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5630 not the same as the source val.
5631 (cgraph_edge_brings_value_p): New parameter.
5632 (gather_edges_for_value): Pass destination value to
5633 cgraph_edge_brings_value_p.
5634 (perhaps_add_new_callers): Likewise.
5635 (get_info_about_necessary_edges): Likewise and exclude values brought
5636 only by self-recursive edges.
5637 (create_specialized_node): Redirect only clones of self-calling edges.
5638 (+self_recursive_pass_through_p): New function.
5639 (find_more_scalar_values_for_callers_subset): Use it.
5640 (find_aggregate_values_for_callers_subset): Likewise.
5641 (known_aggs_to_agg_replacement_list): Removed.
5642 (decide_whether_version_node): Re-calculate known constants for all
5643 remaining context clones.
5645 2018-04-11 Richard Biener <rguenther@suse.de>
5648 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5649 from early DWARF output.
5650 (dwarf2out_early_finish): Output line info unconditionally into
5651 early DWARF and add reference to it.
5653 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5656 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5657 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5658 other than V2DFmode using iptr mode attribute.
5659 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5661 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5663 PR rtl-optimization/84659
5664 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5666 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5669 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5671 (output_loc_list): Pass address of a dummy size variable even in the
5672 locview handling loop.
5673 (index_location_lists): Add comment on why skip_loc_list_entry can't
5676 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5679 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5682 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5685 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5686 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5687 and -mstring-compare-inline-limit.
5689 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5692 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5693 for stack clash protection in a register whenever we need it to be in
5696 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5698 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5699 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5701 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5704 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5706 (mlong-double-): Ditto.
5707 * config/rs6000/sysv4.opt (msdata=): Ditto.
5708 (mtls-size=): Ditto.
5710 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5712 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5713 erroneous entries for
5714 "vector int vec_ldl (int, long int *)", and
5715 "vector unsigned int vec_ldl (int, unsigned long int *)".
5716 Add comments and entries for
5717 "vector bool char vec_ldl (int, bool char *)",
5718 "vector bool short vec_ldl (int, bool short *)",
5719 "vector bool int vec_ldl (int, bool int *)",
5720 "vector bool long long vec_ldl (int, bool long long *)",
5721 "vector pixel vec_ldl (int, pixel *)",
5722 "vector long long vec_ldl (int, long long *)",
5723 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5724 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5725 type tree bool_long_long_type_node and correct definition of
5726 bool_V2DI_type_node to make reference to this new type tree.
5727 (rs6000_mangle_type): Replace erroneous reference to
5728 bool_long_type_node with bool_long_long_type_node.
5729 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5730 comments to emphasize sign distinctions for char and int types and
5731 replace RS6000_BTI_bool_long constant with
5732 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5733 use of RS6000_BTI_pixel.
5734 (bool_long_type_node): Remove this macro definition.
5735 (bool_long_long_type_node): New macro definition
5737 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5739 PR rtl-optimization/85300
5740 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5741 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5742 simplify_unary_operation fails.
5744 2018-04-10 Martin Liska <mliska@suse.cz>
5746 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5747 cgraph_edge and ipa_ref.
5749 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5753 * config/i386/sse.md
5754 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5755 computation of the VEC_MERGE selector from mask.
5756 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5757 Fix decoding of the VEC_MERGE selector into mask.
5759 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5761 PR tree-optimization/85286
5762 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5764 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5766 * final.c (final_1): Set insn_last_address as well as
5767 insn_current_address.
5769 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5772 * explow.c (emit_stack_probe): Call validize_mem on memory location
5773 before passing it to gen_probe_stack. Create address operand and
5774 legitimize it for the probe_stack_address case.
5776 2018-04-09 Jan Hubicka <jh@suse.cz>
5779 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5780 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5781 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5783 (free_lang_data): Rebuild type inheritance graph.
5785 2018-04-09 Martin Sebor <msebor@redhat.com>
5787 * invoke.texi (-finline-small-functions): Mention other optimization
5789 (-findirect-inlining, -fpartial-inlining): Same.
5790 (-finline-functions-called-once): Same.
5791 (-freorder-blocks-and-partition): Same.
5793 2018-04-09 Jan Hubicka <jh@suse.cz>
5796 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5797 jumps; choose last target that matches the criteria (i.e.
5798 no partition changes for non-crossing jumps).
5799 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5800 support for redirecting crossing jumps to non-crossing.
5802 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5804 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5805 also for naked functions.
5807 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5809 * config/arc/arc.md (add_shift): New pattern.
5810 (add_shift2): Likewise.
5811 (sub_shift): Likewise.
5812 (sub_shift_cmp0_noout): Likewise.
5813 (compare_si_ashiftsi): Likewise.
5814 (xbfu_cmp0_noout): New combine pattern.
5815 (xbfu_cmp0"): Likewise.
5816 (movsi_set_cc_insn): Place the predicable variant first.
5817 (commutative_binary_cmp0_noout): Remove clobber.
5818 (commutative_binary_cmp0): New pattern.
5819 (noncommutative_binary_cmp0): Likewise.
5820 (noncommutative_binary_cmp0_noout): Likewise.
5821 (noncommutative_binary_comparison_result_used): Removed.
5822 (rsub_cmp0): New pattern.
5823 (rsub_cmp0_noout): Likewise.
5824 (extzvsi): Changed, keep only meaningful variants.
5825 (SQH, SEZ): New iterators.
5826 (SQH_postfix): New mode attribute.
5827 (SEZ_prefix): New code attribute.
5828 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5829 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5830 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5832 (noncommutative_operator): Check the availability of barrel
5835 2018-04-09 Richard Biener <rguenther@suse.de>
5837 PR tree-optimization/85284
5838 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5839 Only use the niter constraining form of simple_iv when the exit
5842 2018-04-09 Tom de Vries <tom@codesourcery.com>
5845 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5846 (define_expand "*memory_barrier"): New define_expand.
5847 (define_insn "memory_barrier"): New insn.
5849 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5851 PR rtl-optimization/80463
5852 PR rtl-optimization/83972
5853 PR rtl-optimization/83480
5855 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5856 correct producer for the insn.
5857 (tidy_control_flow): Fixup seqnos in case of debug insns.
5859 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5861 PR rtl-optimization/83913
5863 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5864 different sched-times when merging exprs.
5866 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5868 PR rtl-optimization/83962
5870 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5871 tidy_fallthru_edge and tidy_control_flow.
5873 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5875 PR rtl-optimization/83530
5877 * sel-sched.c (force_next_insn): New global variable.
5878 (remove_insn_for_debug): When force_next_insn is true, also leave only
5879 next insn in the ready list.
5880 (sel_sched_region): When the region wasn't scheduled, make another pass
5881 over it with force_next_insn set to 1.
5883 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5885 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5887 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5888 for interrupt control.
5889 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5890 functions for interrupt control.
5891 * config/nds32/nds32-intrinsic.md: Likewise.
5892 * config/nds32/nds32_intrinsic.h: Likewise.
5893 * config/nds32/nds32.h (nds32_builtins): Likewise.
5895 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5897 * config/nds32/nds32.c (nds32_init_machine_status,
5898 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5899 strict_aligned_p field.
5900 (nds32_expand_to_rtl_hook): New function.
5901 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5902 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5904 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5905 Chung-Ju Wu <jasonwucj@gmail.com>
5907 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5908 * config/nds32/nds32-n7.md: New file.
5909 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5910 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5912 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5913 * config/nds32/nds32.md (pipeline_model): Add n7.
5914 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5915 * config/nds32/pipelines.md: Include n7 settings.
5917 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5918 Chung-Ju Wu <jasonwucj@gmail.com>
5920 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5921 * config/nds32/nds32-e8.md: New file.
5922 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5923 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5925 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5926 * config/nds32/nds32.md (pipeline_model): Add e8.
5927 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5928 * config/nds32/pipelines.md: Include e8 settings.
5930 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5931 Chung-Ju Wu <jasonwucj@gmail.com>
5933 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5934 * config/nds32/nds32-n8.md: New file.
5935 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5936 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5938 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5939 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5940 * config/nds32/nds32.md (pipeline_model): Add n8.
5941 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5942 * config/nds32/pipelines.md: Include n8 settings.
5944 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5945 Chung-Ju Wu <jasonwucj@gmail.com>
5947 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5948 * config/nds32/nds32-n9-2r1w.md: New file.
5949 * config/nds32/nds32-n9-3r2w.md: New file.
5950 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5951 nds32_register_ports): New or modify for cpu n9.
5952 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5954 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5955 * config/nds32/nds32-utils.c: New file.
5956 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5957 TARGET_MUL_SLOW): Define.
5958 * config/nds32/nds32.md (pipeline_model): New attribute.
5959 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5960 New options that support cpu n9.
5961 * config/nds32/pipelines.md: Include n9 settings.
5962 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5964 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5966 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5967 information if necessary.
5968 (output_cond_branch_compare_zero): Likewise.
5969 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5970 (nds32_target_alignment): Refine for alignment.
5971 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5972 (FUNCTION_BOUNDARY): Modify.
5973 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5975 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5977 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5979 * config/nds32/constants.md (unspec_volatile_element): Add values for
5980 TLB operation and data prefetch.
5981 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5982 functions for TLB operation and data prefetch.
5983 * config/nds32/nds32-intrinsic.md: Likewise.
5984 * config/nds32/nds32_intrinsic.h: Likewise.
5985 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5986 (nds32_print_operand): Likewise.
5987 * config/nds32/nds32.h (nds32_builtins): Likewise.
5989 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5990 Andrew Pinski <pinsika@gcc.gnu.org>
5993 * match.pd: Use constant_boolean_node of correct type instead of
5994 boolean_true_node or boolean_false_node for simplifying
5995 pointer comparisons to zero.
5997 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5999 PR tree-optimization/80021
6000 * tree.c (verify_type_variant): Make error call in verify_variant_match
6001 translatable and remove final full stop.
6003 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6005 * config/nds32/constants.md (unspec_volatile_element): Add
6006 UNSPEC_VOLATILE_EH_RETURN.
6007 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6008 nds32_output_stack_pop): Support dwarf exception handling process.
6009 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6010 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6011 exception handling process.
6012 (nds32_compute_stack_frame): Likewise.
6013 (nds32_return_addr_rtx): Likewise.
6014 (nds32_initial_elimination_offset): Likewise.
6015 (nds32_expand_prologue): Likewise.
6016 (nds32_expand_epilogue): Likewise.
6017 (nds32_dynamic_chain_address): New function.
6018 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6020 (DYNAMIC_CHAIN_ADDRESS): Define.
6021 (EH_RETURN_DATA_REGNO): Define.
6022 (EH_RETURN_STACKADJ_RTX): Define.
6023 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6024 patterns for dwarf exception handling.
6026 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6028 * config/nds32/nds32.h: Clean up obsolete macros.
6030 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6032 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6033 Add enum values for particular instructions.
6034 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6035 particular intrinsic functions.
6036 * config/nds32/nds32-intrinsic.md: Likewise.
6037 * config/nds32/nds32_intrinsic.h: Likewise.
6038 * config/nds32/nds32.h (nds32_builtins): Likewise.
6039 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6040 (btst, ave): New patterns for particular instructions.
6042 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6044 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6045 Add enum values for atomic load/store and memory sync.
6046 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6048 * config/nds32/nds32-intrinsic.md: Likewise.
6049 * config/nds32/nds32_intrinsic.h: Likewise.
6050 * config/nds32/nds32.h (nds32_builtins): Likewise.
6052 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6054 PR tree-optimization/85257
6055 * fold-const.c (native_encode_vector): If not all elts could fit
6056 and off is -1, return 0 rather than offset.
6057 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6058 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6059 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6060 adjust buffer in native_interpret_expr call.
6062 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6064 * config/nds32/constants.md (unspec_volatile_element): Add cache
6065 control enum values.
6066 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6067 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6068 * config/nds32/nds32.c (nds32_cctl_names): New.
6069 (nds32_print_operand): Handle cache control register names.
6070 * config/nds32/nds32.h (nds32_builtins): New enum values.
6071 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6073 * config/nds32/nds32.md (type): Add mmu.
6074 * config/nds32/pipelines.md (simple_insn): Add mmu.
6076 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6078 * config/nds32/nds32.md (type): Remove call.
6079 * config/nds32/pipelines.md (simple_insn): Likewise.
6081 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6083 * config/nds32/constants.md (unspec_volatile_element): Add
6084 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6085 UNSPEC_VOLATILE_FMFCFG.
6086 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6087 description for fmfcfg and fmfcsr.
6088 (bdesc_1arg): Add fmtcsr.
6089 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6090 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6091 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6092 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6093 unspec_fmfcfg): New patterns.
6094 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6095 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6096 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6097 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6098 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6099 __nds32__fmfcfg): Define.
6101 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6103 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6104 intrinsic register names.
6105 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6106 intrinsic register enum values and macros.
6108 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6110 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6111 for load/store addressing form.
6112 (nds32_print_operand_address): Likewise.
6114 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6117 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6118 based on LABEL_REF. Remove useless assertion.
6119 (pic_address_needs_scratch): Fix formatting.
6120 (sparc_legitimize_pic_address): Minor tweaks.
6121 (sparc_delegitimize_address): Adjust assertion accordingly.
6122 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6123 into symbolic_operand.
6124 (movsi_high_pic_label_ref): Likewise.
6125 (movsi_lo_sum_pic_label_ref): Likewise.
6126 (movdi_pic_label_ref): Likewise.
6127 (movdi_high_pic_label_ref): Likewise.
6128 (movdi_lo_sum_pic_label_ref): Likewise.
6130 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6132 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6133 custom LIB_SPEC setup.
6135 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6136 Kito Cheng <kito.cheng@gmail.com>
6138 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6139 * config/riscv/freebsd.h: New.
6141 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6143 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6144 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6147 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6148 Kito Cheng <kito.cheng@gmail.com>
6150 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6151 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6152 * config/nds32/nds32-protos.h (nds32_output_call,
6153 nds32_output_return): Declare.
6154 * config/nds32/nds32.md: Refine all the call and return patterns.
6156 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6159 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6160 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6162 PR rtl-optimization/84872
6163 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6164 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6167 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6169 * expr.c (copy_blkmode_to_reg): Revert 254862.
6170 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6172 2018-04-06 Richard Biener <rguenther@suse.de>
6175 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6176 after seeing a component reference with an adjacent field. Treat
6177 refs to arrays at struct end of external decls similar to
6178 refs to unconstrained commons.
6180 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6183 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6184 look through SAVE_EXPRs with non-side-effects argument. Adjust
6186 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6189 2018-04-06 Richard Biener <rguenther@suse.de>
6192 * alias.c (find_base_term): New wrapper around find_base_term
6193 unwinding CSELIB_VAL_PTR changes.
6194 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6197 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6199 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6201 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6202 constant definitions.
6203 ("nop"): lr 0,0 -> nopr r0
6204 ("nop_lr0", "nop_lr1"): New insn definitions.
6206 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6208 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6209 NDS32_V3PUSH_AVAILABLE_P macro.
6211 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6212 Chung-Ju Wu <jasonwucj@gmail.com>
6214 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6215 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6216 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6217 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6218 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6219 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6220 * config/nds32/constraints.md: New constraints and checking for hard
6221 float configuration.
6222 * config/nds32/iterators.md: New mode iterator and attribute for hard
6223 float configuration.
6224 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6226 * config/nds32/nds32-fpu.md: New file.
6227 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6228 deal with hard float code generation.
6229 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6231 (abi_type, float_reg_number): New enum type.
6232 * config/nds32/nds32-predicates.c: New predicates for hard float.
6233 * config/nds32/nds32-protos.h: Declare functions for hard float.
6234 * config/nds32/nds32.c: Implementation for hard float configuration.
6235 * config/nds32/nds32.h: Definitions for hard float configuration.
6236 * config/nds32/nds32.md: Include hard float machine description and
6237 modify patterns for hard float configuration.
6238 * config/nds32/nds32.opt: New options for hard float configuration.
6239 * config/nds32/predicates.md: New predicates for hard float
6242 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6244 * common/config/nds32/nds32-common.c
6245 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6247 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6250 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6251 CONSTRUCTOR_ELT (ctor, ...)->value.
6253 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6256 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6258 2018-04-05 Tom de Vries <tom@codesourcery.com>
6261 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6264 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6265 Kito Cheng <kito.cheng@gmail.com>
6267 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6268 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6269 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6270 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6272 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6273 Kito Cheng <kito.cheng@gmail.com>
6275 * config/nds32/constraints.md (Ufe): New memory constraint.
6276 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6277 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6278 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6280 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6281 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6283 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6285 * config/nds32/nds32.md: Use optimize_size in the condition for
6286 alu-shift instructions.
6288 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6290 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6292 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6294 * config/nds32/nds32.md (negsi2): Refine pattern.
6296 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6297 Chung-Ju Wu <jasonwucj@gmail.com>
6299 * config/nds32/iterators.md (shift_rotate): New code iterator.
6300 (shift): New code attribute.
6301 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6302 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6303 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6304 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6305 bit-wise operations.
6306 (andsi3, *andsi3): Ditto.
6307 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6308 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6309 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6310 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6311 nds32_ior_operand, nds32_xor_operand): New predicates.
6313 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6315 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6316 (addsi3, subsi3): ... this.
6318 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6320 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6322 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6324 * config/nds32/nds32.md: Adjust indention.
6326 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6328 * config/nds32/nds32.md (feature): New attribute.
6330 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6332 * config/nds32/nds32.md (subtype): New attribute.
6334 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6337 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6338 expansion to perform a bitwise AND of the argument followed by a
6339 boolean negation of the result.
6341 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6343 PR rtl-optimization/84878
6344 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6345 the basic block. Assert the use reference is not artificial and that
6346 it has an associated insn.
6348 2018-04-04 Michael Matz <matz@suse.de>
6350 * builtins.c (compute_objsize): Pass correct operand
6351 to array_at_struct_end_p.
6353 2018-04-04 Richard Biener <rguenther@suse.de>
6356 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6357 from contexts for DINFO_LEVEL_TERSE and below.
6359 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6361 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6362 resiter_operand condition.
6363 * config/nds32/nds32.md (*move<mode>): Ditto.
6365 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6366 Monk Chiang <sh.chiang04@gmail.com>
6368 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6370 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6372 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6374 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6375 Kito Cheng <kito.cheng@gmail.com>
6377 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6378 nds32_cond_code_str, output_cond_branch,
6379 output_cond_branch_compare_zero, nds32_expand_cbranch,
6380 nds32_expand_cstore, nds32_expand_movcc,
6381 nds32_output_cbranchsi4_equality_zero,
6382 nds32_output_cbranchsi4_equality_reg,
6383 nds32_output_cbranchsi4_equality_reg_or_const_int,
6384 nds32_output_cbranchsi4_greater_less_zero: New functions.
6385 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6386 nds32_expand_cstore, nds32_expand_movcc,
6387 nds32_output_cbranchsi4_equality_zero,
6388 nds32_output_cbranchsi4_equality_reg,
6389 nds32_output_cbranchsi4_equality_reg_or_const_int,
6390 nds32_output_cbranchsi4_greater_less_zero): Declare.
6391 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6392 nds32_rimm11s_operand): New predicates.
6393 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6394 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6397 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6399 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6400 * config/nds32/nds32.md: Ditto.
6401 * config/nds32/pipelines.md: Ditto.
6403 2018-04-04 Richard Biener <rguenther@suse.de>
6405 PR tree-optimization/85168
6406 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6407 propagating abnormals.
6409 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6411 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6413 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6414 Kito Cheng <kito.cheng@gmail.com>
6416 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6417 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6418 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6419 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6420 * config/nds32/nds32.md (sibcall_internal): New.
6421 (sibcall_register): Remove.
6422 (sibcall_immediate): Remove.
6423 (sibcall_value_internal): New.
6424 (sibcall_value_register): Remove.
6425 (sibcall_value_immediate): Remove.
6426 * config/nds32/predicates.md (nds32_general_register_operand): New.
6427 (nds32_call_address_operand): New.
6429 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6431 PR rtl-optimization/85167
6432 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6433 bb_defs if *split_p, instead preinitialize it to NULL.
6435 PR tree-optimization/85156
6436 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6437 evaluating the argument multiple times.
6439 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6441 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6443 (_mm_cvtpd_ps): Likewise.
6444 (_mm_cvttpd_epi32): Likewise.
6445 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6446 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6447 vector, pixel, and bool following altivec.h include.
6449 2018-04-03 Martin Sebor <msebor@redhat.com>
6451 * doc/extend.texi (Common Function Attributes): Clarify.
6452 (const attribute): Likewise.
6453 (pure attribute): Likewise.
6455 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6458 * config/i386/i386.c (ix86_expand_vector_set): Use
6459 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6461 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6463 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6464 instructions when changing rounding bits to preserve precision bits
6465 in the x87 control word.
6467 2018-04-03 Martin Liska <mliska@suse.cz>
6469 PR tree-optimization/82491
6470 * rtl.h (strip_offset_and_add): Replace += suboffset with
6471 poly_uint64 () + suboffset.
6473 2018-03-29 Martin Liska <mliska@suse.cz>
6474 Martin Jambor <mjambor@suse.cz>
6477 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6478 param_type is not an integral or pointer type.
6480 2018-04-03 Richard Biener <rguenther@suse.de>
6482 * sese.h (recompute_all_dominators): Remove.
6484 2018-04-02 Martin Sebor <msebor@redhat.com>
6486 * doc/invoke.texi (-Wrestrict): Fix typos.
6488 2018-04-02 Jim Wilson <jimw@sifive.com>
6490 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6491 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6492 (<optab>di3, <optab>si3_extend): Likewise.
6493 (<optab>si3_mask, <optab>si3_mask_1): New.
6494 (<optab>di3_mask, <optab>di3_mask_1): New.
6495 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6496 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6497 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6499 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6501 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6504 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6506 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6507 (nds32_canonicalize_comparison): New function.
6509 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6510 Kito Cheng <kito.cheng@gmail.com>
6511 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6513 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6514 * config/nds32/constants.md (unspec_volatile_element): Add
6515 UNSPEC_VOLATILE_RELAX_GROUP.
6516 * config/nds32/nds32-relax-opt.c: New file.
6517 * config/nds32/nds32-predicates.c
6518 (nds32_symbol_load_store_p): New function.
6519 * config/nds32/nds32-protos.h
6520 (nds32_symbol_load_store_p): Declare function.
6521 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6522 * config/nds32/nds32.c
6523 (nds32_register_pass): New function to register pass.
6524 (nds32_register_passes): New function to register passes.
6525 * config/nds32/nds32.md (relax_group): New pattern.
6526 * config/nds32/nds32.opt (mrelax-hint): New option.
6527 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6529 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6531 * config/nds32/t-nds32: Modify files dependency.
6533 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6535 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6536 (PROFILE_HOOK): Define its implementation.
6538 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6540 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6541 type and 32-bit size.
6543 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6546 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6547 (V_128_256): New mode iterator.
6548 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6549 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6550 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6552 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6553 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6555 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6558 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6559 NaN inputs correctly.
6561 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6564 * config/rs6000/vsx.md (??r): New mode attribute.
6565 (*vsx_mov<mode>_64bit): Use it.
6566 (*vsx_mov<mode>_32bit): Likewise.
6568 2018-03-30 Martin Sebor <msebor@redhat.com>
6570 PR tree-optimization/84818
6571 * builtins.c (check_access): Use warning_n.
6573 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6576 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6578 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6581 2018-03-30 Julia Koval <julia.koval@intel.com>
6584 * x86-tune.def (movx, partial_reg_dependency): Enable for
6587 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6590 * lra-constraints.c (process_alt_operands): Move setting
6591 this_alternative_matches below.
6593 2018-03-29 Martin Liska <mliska@suse.cz>
6596 * doc/invoke.texi: Document how LTO works with debug info.
6597 Describe auto-load support of binutils. Mention 'x86-64'
6598 as valid option value of -march option.
6600 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6602 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6605 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6606 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6607 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6610 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6613 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6614 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6615 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6616 for RS6000_BTM_POWERPC64.
6617 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6618 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6619 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6624 2018-03-28 Carl Love <cel@us.ibm.com>
6627 2017-09-27 Carl Love <cel@us.ibm.com>
6629 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6630 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6631 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6634 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6636 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6637 instead of __vector bool.
6638 (_mm_max_pu8): Likewise.
6639 (_mm_min_pi16): Likewise.
6641 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6644 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6645 (DIVWEUO): Likewise.
6647 (DIVDEUO): Likewise.
6648 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6649 DIVWEUO and DIVDEUO.
6650 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6651 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6652 (div_extend): Likewise.
6653 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6655 (__builtin_divweuo): Likewise.
6656 (__builtin_divdeo): Likewise.
6657 (__builtin_divdeuo): Likewise.
6659 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6662 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6663 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6665 PR tree-optimization/82004
6666 * gimple-match-head.c (optimize_pow_to_exp): New function.
6667 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6668 Don't fold to exp if optimize_pow_to_exp is false.
6670 2018-03-28 Martin Liska <mliska@suse.cz>
6673 * calls.c (initialize_argument_information): Fix trailing space.
6674 * common.opt: Fix typo and provide better explanation for
6675 -fsanitize-coverage option.
6676 * config/i386/i386.opt: Fix typo.
6678 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6679 Martin Liska <mliska@suse.cz>
6682 * gimplify.c (asan_poison_variable): Don't do the check for
6683 gimplify_omp_ctxp here.
6684 (gimplify_decl_expr): Do it here.
6685 (gimplify_target_expr): Likewise.
6687 2018-03-28 Martin Liska <mliska@suse.cz>
6690 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6691 chkp_type_bounds_count if MPX is not enabled.
6693 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6695 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6697 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6700 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6701 function to create the function decl for complex long double
6702 multiply and divide for -mabi=ieeelongdouble.
6703 (init_float128_ieee): Call it.
6705 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6708 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6709 -fcf-protection=branch -mibt.
6710 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6712 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6715 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6716 arm_word_relocations.
6718 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6721 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6722 extern array declarations.
6724 2018-03-27 Richard Biener <rguenther@suse.de>
6727 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6728 explicit single_use checks.
6730 2018-03-27 Richard Biener <rguenther@suse.de>
6732 PR tree-optimization/85082
6733 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6736 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6738 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6739 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6740 Turn on fasynchronous-unwind-tables and funwind-tables.
6742 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6745 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6746 (*bmi_blsr_<mode>_ccz): Ditto.
6748 2018-03-26 Tom de Vries <tom@codesourcery.com>
6750 PR tree-optimization/85063
6751 * omp-general.c (offloading_function_p): New function. Factor out
6753 * omp-offload.c (pass_omp_target_link::gate): ... here.
6754 * omp-general.h (offloading_function_p): Declare.
6755 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6756 with attribute omp declare target for offloading functions.
6758 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6760 PR tree-optimization/84005
6761 * tree-data-ref.h (get_base_for_alignment): Declare.
6762 * tree-data-ref.c (get_base_for_alignment_1): New function.
6763 (get_base_for_alignment): Likewise.
6764 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6765 get_base_for_alignment to find a suitable base object, instead
6766 of always using drb->base_address.
6768 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6771 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6772 known size by default.
6774 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6777 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6778 and non BLKmode operands.
6780 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6783 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6784 Clean up attributes.
6786 2018-03-23 Richard Biener <rguenther@suse.de>
6789 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6790 we are going to emit early debug for LTO.
6792 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6795 * function.c (match_asm_constraints_1): Don't optimize if input
6796 doesn't satisfy general_operand predicate for output's mode.
6799 * alias.c (write_dependence_p): Don't require for x_canonicalized
6800 non-VOIDmode if x has VOIDmode.
6803 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6804 just don't try to optimize it rather than assert it never happens.
6806 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6808 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6809 macro expansions for definition of ST_INTERNAL_<mode> and
6810 LD_INTERNAL_<mode> builtins.
6811 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6813 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6815 (altivec_expand_st_builtin): Likewise.
6816 (altivec_expand_builtin): Remove calls to deleted functions.
6817 (rs6000_address_for_altivec): Delete this function.
6818 * config/rs6000/vector.md: Remove expands for
6819 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6821 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6824 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6825 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6826 re-computing once computed.
6827 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6828 (arm_init_machine_status): Initialize
6829 machine->static_chain_stack_bytes.
6831 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6834 * doc/extend.texi: Add four new prototypes for vec_ld.
6835 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6836 definitions for more logical presentation.
6837 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6838 entries for V1TI variants of __builtin_altivec_ld builtin.
6839 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6840 handling of V1TI variant of LVX icode pattern.
6841 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6842 (rs6000_gimple_fold_builtin): Likewise.
6843 (altivec_init_builtins): Add code to define
6844 __builtin_altivec_lvx_v1ti function.
6846 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6849 * function.c (match_asm_constraints_1): Don't do the optimization
6850 if input isn't a REG, SUBREG, MEM or constant.
6852 2018-03-22 Tom de Vries <tom@codesourcery.com>
6854 PR tree-optimization/84956
6855 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6856 bb_has_abnormal_pred.
6858 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6861 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6862 DECL_INITIAL (decl) to decl at the end.
6863 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6866 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6868 * doc/extend.texi (__builtin_tgmath): Document when complex
6869 integer types are treated as _Complex _Float64.
6871 2018-03-21 Tom de Vries <tom@codesourcery.com>
6873 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6875 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6877 PR tree-optimization/84960
6878 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6879 if it is ENTRY block, move them into single succ of ENTRY in that case.
6881 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6883 PR tree-optimization/84811
6884 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6885 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6886 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6887 (known_subrange_p): Update accordingly. Cast each value involved
6888 in the size comparison, rather than casting the result of the
6891 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6893 PR tree-optimization/84982
6894 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6895 by flipping the least significant bit rather than all bits from
6896 bitpos to bitpos + bitsize - 1.
6898 2018-03-21 Nathan Sidwell <nathan@acm.org>
6900 * doc/extend.texi (Deprecated Features): Remove mention of
6901 long-deleted deprecations.
6903 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6906 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6907 * configure: Regenerate.
6909 2018-03-21 Tom de Vries <tom@codesourcery.com>
6911 PR tree-optimization/83126
6912 * tree-parloops.c (num_phis): New function.
6913 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6915 2018-03-21 Nathan Sidwell <nathan@acm.org>
6917 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6918 mention anon-struct/union members and trailing attributes.
6920 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6922 PR tree-optimization/84969
6923 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6924 builtin memset partitions if they set different rhs values.
6926 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6928 PR rtl-optimization/84989
6929 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6930 VEC_DUPLICATE with scalar result mode.
6932 2018-03-21 Martin Liska <mliska@suse.cz>
6935 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6936 not intended return statement.
6938 2018-03-21 Martin Liska <mliska@suse.cz>
6941 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6942 (chkp_find_bound_slots_1): Limit number of iterations.
6944 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6947 * Minor grammar fixes for x86 options.
6949 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6952 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6953 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6955 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6958 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6959 (altivec_lvx_<mode>_1op): Likewise.
6960 (altivec_stvx_<mode>_2op): Likewise.
6961 (altivec_stvx_<mode>_1op): Likewise.
6962 (altivec_lvx_<VM2:mode>): New define_expand.
6963 (altivec_stvx_<VM2:mode>): Likewise.
6964 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6965 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6966 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6967 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6968 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6969 (rs6000_gen_lvx): Likewise.
6970 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6971 (altivec_expand_stv_builtin): Likewise.
6972 (altivec_expand_builtin): Likewise.
6973 * config/rs6000/vector.md: Likewise.
6975 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6978 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6981 2018-03-20 Richard Biener <rguenther@suse.de>
6984 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6985 sign-conversions as zero, fall back to standard scalar_stmt
6988 2018-03-20 Martin Liska <mliska@suse.cz>
6991 * predict.c (rebuild_frequencies): Handle case when we have
6992 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6994 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6997 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6998 flag_section_anchors.
6999 * varasm.c (use_blocks_for_decl_p): Remove hack for
7000 dw2_force_const_mem.
7003 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7005 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7006 be created, use lowpart_subreg of operands[0] rather than operands[0]
7008 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7009 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7010 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7011 and n constraint instead of aarch64_shift_imm_di and Usd.
7012 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7013 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7015 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7018 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7019 to favor GPR over NEON registers.
7020 (<shift>di3_neon): Likewise.
7022 2018-03-20 Tom de Vries <tom@codesourcery.com>
7025 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7026 (nvptx_process_pars): Emit bar.sync asap and alap.
7028 2018-03-20 Tom de Vries <tom@codesourcery.com>
7031 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7032 seen_label if seen_label is already set.
7034 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7037 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7038 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7039 Use 1U instead of 1. Formatting fixes.
7042 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7043 instead of TREE_TYPE (s1) for the return value.
7045 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7047 PR tree-optimization/84946
7048 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7049 bitsize + bitsize in poly_uint64 rather than poly_int64.
7052 * dwarf2asm.c: Include fold-const.c.
7053 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7054 of decl rather than decl itself.
7056 PR rtl-optimization/84643
7057 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7059 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7062 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7063 calling assemble_variable.
7065 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7068 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7069 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7071 2018-03-19 Jim Wilson <jimw@sifive.com>
7074 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7075 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7076 (riscv_first_stack_step): Likewise.
7077 (riscv_option_override): Use STACK_BOUNDARY instead of
7079 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7081 (BIGGEST_ALIGNMENT): Set to 128.
7082 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7083 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7086 2018-03-19 Richard Biener <rguenther@suse.de>
7088 PR tree-optimization/84933
7089 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7090 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7092 2018-03-19 Richard Biener <rguenther@suse.de>
7094 PR tree-optimization/84859
7095 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7096 (cond_if_else_store_replacement): Perform sinking operation on
7097 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7098 Generalize what a BB with a single eligible store is.
7100 2018-03-19 Richard Biener <rguenther@suse.de>
7102 PR tree-optimization/84929
7103 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7104 chrec_is_positive against non-chrec arg.
7106 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7109 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7111 2018-03-18 Martin Liska <mliska@suse.cz>
7113 PR rtl-optimization/84635
7114 * regrename.c (build_def_use): Use matches_mode only when
7117 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7119 PR tree-optimization/84913
7120 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7121 vectorize chains of COND_EXPRs.
7123 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7125 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7127 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7129 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7131 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7133 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7135 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7136 Kito Cheng <kito.cheng@gmail.com>
7138 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7139 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7140 (nds32_adjust_reg_alloc_order): New function.
7141 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7143 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7145 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7146 nds32_print_operand, nds32_print_operand_address): Use
7147 HOST_WIDE_INT_PRINT_DEC instead.
7149 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7151 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7153 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7156 * config/i386/i386.c (initial_ix86_tune_features,
7157 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7159 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7160 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7161 rather than 1u << ix86_tune. Formatting fix.
7162 (ix86_option_override_internal): Change ix86_arch_mask from
7163 unsigned int to unsigned HOST_WIDE_INT, initialize to
7164 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7165 (ix86_function_specific_restore): Likewise.
7167 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7170 * postreload.c (reload_combine_recognize_pattern): Perform
7171 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7172 truncate_int_for_mode the result for the destination's mode.
7175 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7176 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7178 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7181 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7184 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7185 regno_allocno_class_array and sorted_pseudos.
7186 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7187 insns where regno is used.
7189 2018-03-16 Martin Liska <mliska@suse.cz>
7192 * multiple_target.c (create_dispatcher_calls): Redirect
7193 reference in the symbol table.
7195 2018-03-16 Martin Liska <mliska@suse.cz>
7198 * multiple_target.c (create_dispatcher_calls): Redirect also
7201 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7205 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7206 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7207 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7209 2018-03-16 Julia Koval <julia.koval@intel.com>
7211 * doc/invoke.texi (Skylake Server): Add CLWB.
7212 Cannonlake): Remove CLWB.
7214 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7216 PR tree-optimization/84841
7217 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7219 (FLOAT_ONE_CONST_TYPE): Define.
7220 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7221 (sort_by_operand_rank): Put entries with higher constant_type last
7222 rather than first to match comments.
7224 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7226 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7229 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7232 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7233 check and use address of target always.
7235 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7238 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7239 (indirect_thunk_bnd_needed): Likewise.
7240 (indirect_thunks_used): Likewise.
7241 (indirect_thunks_bnd_used): Likewise.
7242 (indirect_return_needed): New.
7243 (indirect_return_bnd_needed): Likewise.
7244 (output_indirect_thunk_function): Add a bool argument for
7246 (output_indirect_thunk_function): Don't generate alias for
7247 function return thunk.
7248 (ix86_code_end): Call output_indirect_thunk_function to generate
7249 function return thunks.
7250 (ix86_output_function_return): Set indirect_return_bnd_needed
7251 and indirect_return_needed instead of indirect_thunk_bnd_needed
7252 and indirect_thunk_needed.
7254 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7256 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7257 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7258 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7260 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7261 Paul Hua <paul.hua.gm@gmail.com>
7264 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7266 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7268 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7269 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7272 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7275 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7276 instead of GET_MODE_SIZE when comparing Units.
7278 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7281 * varasm.c (hash_section): Return an unchangeble hash value
7282 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7283 Return !aarch64_can_use_per_function_literal_pools_p ().
7285 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7288 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7289 rather than address of cmode as last argument to prepare_cmp_insn.
7291 2018-03-15 Julia Koval <julia.koval@intel.com>
7293 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7294 F_AVX512VNNI, F_AVX512BITALG): New.
7296 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7299 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7300 insn for floating-point loads and stores.
7302 2018-03-14 Carl Love <cel@us.ibm.com>
7304 * config/rs6000/rs6000-c.c: Add macro definitions for
7305 ALTIVEC_BUILTIN_VEC_PERMXOR.
7306 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7307 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7308 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7309 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7311 * config/doc/extend.texi: Add prototypes for vec_permxor.
7313 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7316 * diagnostic-show-locus.c (class layout_point): Convert m_line
7317 from int to linenum_type.
7318 (line_span::comparator): Use linenum "compare" function when
7319 comparing line numbers.
7320 (test_line_span): New function.
7321 (layout_range::contains_point): Convert param "row" from int to
7323 (layout_range::intersects_line_p): Likewise.
7324 (layout::will_show_line_p): Likewise.
7325 (layout::print_source_line): Likewise.
7326 (layout::should_print_annotation_line_p): Likewise.
7327 (layout::print_annotation_line): Likewise.
7328 (layout::print_leading_fixits): Likewise.
7329 (layout::annotation_line_showed_range_p): Likewise.
7330 (struct line_corrections): Likewise for field m_row.
7331 (line_corrections::line_corrections): Likewise for param "row".
7332 (layout::print_trailing_fixits): Likewise.
7333 (layout::get_state_at_point): Likewise.
7334 (layout::get_x_bound_for_row): Likewise.
7335 (layout::print_line): Likewise.
7336 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7337 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7338 * input.c (selftest::test_linenum_comparisons): New function.
7339 (selftest::input_c_tests): Call it.
7340 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7341 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7342 * selftest.h (ASSERT_GT): New macro.
7343 (ASSERT_GT_AT): New macro.
7344 (ASSERT_LT): New macro.
7345 (ASSERT_LT_AT): New macro.
7347 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7349 PR rtl-optimization/84780
7350 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7352 2018-03-14 Martin Liska <mliska@suse.cz>
7354 * tree.c (record_node_allocation_statistics): Use
7355 get_stats_node_kind.
7356 (get_stats_node_kind): New function extracted from
7357 record_node_allocation_statistics.
7358 (free_node): Use get_stats_node_kind.
7360 2018-03-14 Richard Biener <rguenther@suse.de>
7362 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7363 that the value-set of ANTIC_IN doesn't grow.
7366 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7368 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7369 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7371 2018-03-14 Julia Koval <julia.koval@intel.com>
7373 * config.gcc (icelake-client, icelake-server): New.
7375 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7376 (initial_ix86_arch_features): Ditto.
7377 (PTA_SKYLAKE): Add SGX.
7378 (PTA_ICELAKE): Remove.
7379 (PTA_ICELAKE_CLIENT): New.
7380 (PTA_ICELAKE_SERVER): New.
7381 (ix86_option_override_internal): Split up icelake on icelake client and
7383 (get_builtin_code_for_version): Ditto.
7384 (fold_builtin_cpu): Ditto.
7385 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7386 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7387 * config/i386/i386.h (processor_type): Ditto.
7388 * doc/invoke.texi: Ditto.
7390 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7393 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7394 INTEGER_CST offset, add it together with bitpos / 8 and
7395 sign extend based on POINTER_SIZE.
7399 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7402 * config/i386/constraints.md (Yc): New register constraint.
7403 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7404 Use Yc constraint for alternative 2 of operand 0. Remove
7405 preferred_for_speed attribute.
7407 2018-03-14 Richard Biener <rguenther@suse.de>
7409 PR tree-optimization/84830
7410 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7411 with the old one to avoid oscillations.
7413 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7416 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7418 (assign_by_spills): Return a flag of reload assignment failure.
7419 Do not process the reload assignment failures. Do not spill other
7420 reload pseudos if they has the same reg class. Update n if
7422 (lra_assign): Add a return arg. Set up from the result of
7423 assign_by_spills call.
7424 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7425 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7426 usage_insns if it is not NULL.
7427 (spill_hard_reg_in_range): New function.
7428 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7429 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7430 function prototypes.
7431 (lra_assign): Change prototype.
7432 * lra.c (lra): Add code to deal with fails by splitting hard reg
7435 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7437 * config/riscv/riscv.opt (mrelax): New option.
7438 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7439 "norelax" when riscv_mrelax is disabled.
7440 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7442 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7445 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7446 reassociation for int modes.
7448 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7450 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7451 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7453 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7454 * config/aarch64/aarch64-sve.md
7455 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7456 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7457 (*extend<mode><Vwide>2): Rename to...
7458 (aarch64_sve_extend<mode><Vwide>2): ...this.
7459 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7460 renaming the old pattern to...
7461 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7463 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7464 define_expand, renaming the old pattern to...
7465 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7466 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7467 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7468 account when deciding which SVE instruction the optab should use.
7469 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7471 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7473 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7474 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7475 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7476 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7477 (tlsdesc_small_<mode>): Turn a define_expand and use
7478 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7479 (tlsdesc_small_advsimd_<mode>): ...this.
7480 (tlsdesc_small_sve_<mode>): New pattern.
7482 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7484 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7485 (UNSPEC_UMUL_HIGHPART): New constants.
7486 (MUL_HIGHPART): New int iteraor.
7487 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7488 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7490 (*<su>mul<mode>3_highpart): New define_insn.
7492 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7495 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7498 2018-03-13 Martin Liska <mliska@suse.cz>
7501 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7503 (sem_item_optimizer::~sem_item_optimizer): Release it.
7504 (sem_item_optimizer::merge_classes): Register variable aliases.
7505 (sem_item_optimizer::fixup_pt_set): New function.
7506 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7507 * ipa-icf.h: Declare new variables and functions.
7509 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7512 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7513 integer_pow2p@2 and test integer_pow2p in condition.
7514 (A < 0 ? C : 0): Similarly for @1.
7517 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7518 characters starting at p contain '\0' character, don't look beyond
7522 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7523 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7526 * reg-stack.c (change_stack): Change update_end var from int to
7527 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7528 also call set_block_for_insn on the newly added insns and rescan.
7531 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7532 on the last operand.
7535 * tree.c (stabilize_reference_1): Return save_expr (e) for
7536 STATEMENT_LIST even if it doesn't have side-effects.
7538 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7540 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7542 2018-03-12 Renlin Li <renlin.li@arm.com>
7544 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7545 aarch64_output_scalar_simd_mov_immediate.
7547 2018-03-12 Martin Sebor <msebor@redhat.com>
7549 PR tree-optimization/83456
7550 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7551 for perfectly overlapping calls to memcpy.
7552 (gimple_fold_builtin_memory_chk): Same.
7553 (gimple_fold_builtin_strcpy): Handle no-warning.
7554 (gimple_fold_builtin_stxcpy_chk): Same.
7555 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7557 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7559 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7560 parameter. Use it for SFmode.
7561 (rs6000_function_arg_advance_1): Adjust.
7562 (rs6000_function_arg): Adjust.
7563 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7565 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7567 PR rtl-optimization/84169
7568 PR rtl-optimization/84780
7569 * combine.c (can_combine_p): Check for a 2-insn combination whether
7570 the destination register is used between the two insns, too.
7572 2018-03-12 Richard Biener <rguenther@suse.de>
7574 PR tree-optimization/84803
7575 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7576 for refs DR analysis didn't process.
7578 2018-03-12 Richard Biener <rguenther@suse.de>
7580 PR tree-optimization/84777
7581 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7582 force-vectorize loops ignore whether we are optimizing for size.
7584 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7586 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7587 (TARGET_MD_ASM_ADJUST): Define.
7589 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7590 Kito Cheng <kito.cheng@gmail.com>
7591 Chung-Ju Wu <jasonwucj@gmail.com>
7593 * config/nds32/nds32.c (nds32_compute_stack_frame,
7594 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7595 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7596 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7597 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7598 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7599 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7600 * config/nds32/nds32.md (prologue, epilogue): Use macro
7601 NDS32_V3PUSH_AVAILABLE_P to do checking.
7603 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7606 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7607 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7608 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7609 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7610 addition of most attributes on !orig_type_die or the attribute not
7611 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7613 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7614 Chung-Ju Wu <jasonwucj@gmail.com>
7616 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7618 * config/nds32/nds32.opt (mvh): New option.
7620 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7621 Chung-Ju Wu <jasonwucj@gmail.com>
7623 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7625 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7626 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7629 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7630 Chung-Ju Wu <jasonwucj@gmail.com>
7632 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7634 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7635 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7637 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7638 Kito Cheng <kito.cheng@gmail.com>
7639 Chung-Ju Wu <jasonwucj@gmail.com>
7641 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7642 UNSPEC_FFMISM and UNSPEC_FLMISM.
7643 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7644 for ffb, ffmism and flmism.
7645 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7646 (unspec_ffmism): Ditto.
7647 (unspec_flmism): Ditto.
7648 (nds32_expand_builtin_impl): Check if string extension is available.
7649 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7650 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7652 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7655 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7658 * lra-assigns.c (assign_by_spills): Return a flag of reload
7659 assignment failure. Do not process the reload assignment
7660 failures. Do not spill other reload pseudos if they has the same
7662 (lra_assign): Add a return arg. Set up from the result of
7663 assign_by_spills call.
7664 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7665 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7666 usage_insns if it is not NULL.
7667 (spill_hard_reg_in_range): New function.
7668 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7669 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7670 function prototypes.
7671 (lra_assign): Change prototype.
7672 * lra.c (lra): Add code to deal with fails by splitting hard reg
7675 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7678 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7680 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7683 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7684 (dw_val_node): Add val_symbolic_view.
7685 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7686 (symview_upper_bound): New.
7687 (new_line_info_table): Initialize symviews_since_reset.
7688 (dwarf2out_source_line): Count symviews_since_reset and set
7689 symview_upper_bound.
7690 (dw_val_equal_p): Handle symview.
7691 (add_AT_symview): New.
7692 (print_dw_val): Handle symview.
7693 (attr_checksum, attr_checksum_ordered): Likewise.
7694 (same_dw_val_p, size_of_die): Likewise.
7695 (value_format, output_die): Likewise.
7696 (add_high_low_attributes): Use add_AT_symview for entry_view.
7697 (dwarf2out_finish): Reset symview_upper_bound, clear
7700 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7703 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7704 Add strict argument and use it.
7705 (rs6000_split_multireg_move): Update for new strict argument.
7706 (mem_operand_gpr): Disallow all non-offsettable addresses.
7707 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7709 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7712 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7713 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7714 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7717 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7718 decl, use remap_type if we want to use the type.
7720 2018-03-09 Martin Sebor <msebor@redhat.com>
7722 PR tree-optimization/84526
7723 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7725 (builtin_access::generic_overlap): Be prepared to handle non-array
7728 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7730 PR rtl-optimization/84682
7731 * lra-constraints.c (process_address_1): Check is_address flag
7732 for address constraints.
7733 (process_alt_operands): Likewise.
7734 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7735 preprocess_constraints.
7736 * recog.h (preprocess_constraints): Add oploc parameter.
7738 * recog.c (preprocess_constraints): Test address_operand for
7739 CT_ADDRESS constraints.
7741 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7744 * lra-assigns.c (assign_by_spills): Return a flag of reload
7745 assignment failure. Do not process the reload assignment
7746 failures. Do not spill other reload pseudos if they has the same
7748 (lra_assign): Add a return arg. Set up from the result of
7749 assign_by_spills call.
7750 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7751 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7752 usage_insns if it is not NULL.
7753 (spill_hard_reg_in_range): New function.
7754 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7755 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7756 function prototypes.
7757 (lra_assign): Change prototype.
7758 * lra.c (lra): Add code to deal with fails by splitting hard reg
7761 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7764 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7765 Accept complain bool parameter. Only emit errors if it is true.
7766 (arm_parse_cpu_option_name): Likewise.
7767 (arm_target_thumb_only): Adjust callers of the above.
7768 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7769 prototype to take a default true bool parameter.
7770 (arm_parse_arch_option_name): Likewise.
7772 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7773 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7777 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7778 * configure: Regenerate.
7779 * configure.ac ("linker --version-script option"): New.
7780 ("linker soname option"): New.
7782 2018-03-09 Richard Biener <rguenther@suse.de>
7784 PR tree-optimization/84775
7785 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7786 immediate uses of predicate stmts and mark them modified.
7789 PR tree-optimization/84178
7790 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7792 (version_loop_for_if_conversion): Delay update_ssa call.
7793 (tree_if_conversion): Delay update_ssa until after predicate
7796 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7799 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7800 when the function accesses prior frames.
7802 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7805 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7806 gen_llsym, otherwise call maybe_gen_llsym.
7809 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7810 has ',' character inside of it.
7812 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7815 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7816 as clobbering CC_REGNUM.
7818 2018-03-08 Richard Biener <rguenther@suse.de>
7821 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7822 (follow_copies_to_constant): Do not follow SSA names registered
7825 2018-03-08 Richard Biener <rguenther@suse.de>
7827 PR tree-optimization/84178
7828 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7830 (version_loop_for_if_conversion): Delay update_ssa call.
7831 (tree_if_conversion): Delay update_ssa until after predicate
7834 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7836 PR tree-optimization/84178
7837 * tree-if-conv.c (release_bb_predicate): Remove the
7838 the assertion that the stmts have NULL use_ops.
7839 Discard the statements, asserting that they haven't
7840 yet been added to a BB.
7842 2018-03-08 Richard Biener <rguenther@suse.de>
7844 PR tree-optimization/84746
7845 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7846 (phi_translate): Pass in destination ANTIC_OUT set.
7847 (phi_translate_1): Likewise. For a simplified result lookup
7848 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7849 (phi_translate_set): Adjust.
7850 (do_pre_regular_insertion): Likewise.
7851 (do_pre_partial_partial_insertion): Likewise.
7853 2018-03-08 Martin Liska <mliska@suse.cz>
7855 PR gcov-profile/84735
7856 * doc/gcov.texi: Document usage of profile files.
7857 * gcov-io.h: Document changes in the format.
7859 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7863 * dwarf2out.c (struct dw_line_info_table): Update comments for
7865 (FORCE_RESET_NEXT_VIEW): New.
7866 (FORCE_RESETTING_VIEW_P): New.
7867 (RESETTING_VIEW_P): Check for -1 too.
7868 (ZERO_VIEW_P): Likewise.
7869 (new_line_info_table): Force-reset next view.
7870 (dwarf2out_begin_function): Likewise.
7871 (dwarf2out_source_line): Simplify zero_view_p initialization.
7872 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7873 view directly. Omit view when omitting .loc at line 0.
7875 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7877 PR tree-optimization/84740
7878 * tree-switch-conversion.c (process_switch): Call build_constructors
7879 only if info.phi_count is non-zero.
7881 PR tree-optimization/84739
7882 * tree-tailcall.c (find_tail_calls): Check call arguments against
7883 DECL_ARGUMENTS (current_function_decl) rather than
7884 DECL_ARGUMENTS (func) when checking for tail recursion.
7886 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7888 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7889 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7890 Volker Reichelt's entry and add entries for people that perform
7891 GCC fuzzy testing and report numerous bugs.
7893 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7896 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7897 readonly data in sdata, if that is disabled.
7898 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7899 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7900 -mreadonly-in-sdata option.
7902 2018-03-07 Martin Sebor <msebor@redhat.com>
7904 PR tree-optimization/84468
7905 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7906 basic block when looking for nul assignment.
7908 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7911 * except.h (output_function_exception_table): Adjust prototype.
7912 * except.c (output_function_exception_table): Remove FNNAME parameter
7913 and add SECTION parameter. Ouput one part of the table at a time.
7914 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7915 the first part of the exception table and emit unwind directives.
7916 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7917 (i386_pe_seh_cold_init): Likewise.
7918 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7919 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7920 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7921 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7922 * config/i386/winnt.c: Include except.h.
7923 (struct seh_frame_state): Add reg_offset, after_prologue and
7924 in_cold_section fields.
7925 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7926 (i386_pe_seh_cold_init): New function.
7927 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7928 to seh->in_cold_section.
7929 (seh_emit_push): Record the offset of the push.
7930 (seh_emit_save): Record the offet of the save.
7931 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7932 Test seh->after_prologue to disregard the epilogue.
7933 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7934 (i386_pe_end_cold_function): New function.
7936 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7939 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7940 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7943 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7945 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7946 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7949 * multiple_target.c: Include tree-inline.h and intl.h.
7950 (expand_target_clones): Diagnose and fail if node->definition and
7951 !tree_versionable_function_p (node->decl).
7953 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7955 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7957 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7958 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7959 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7961 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7964 * combine.c (try_combine): Use reg_or_subregno instead of handling
7965 just paradoxical SUBREGs and REGs.
7967 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7969 * config/arc/arc.c (arc_finalize_pic): Remove function.
7970 (arc_must_save_register): We use single base PIC register, remove
7971 checks to save/restore the PIC register.
7972 (arc_expand_prologue): Likewise.
7973 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7975 (arc_verify_short): Likewise.
7976 (arc_attr_type): Likewise.
7977 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7978 (walk_stores): Likewise.
7979 (arc_address_cost): Make it static.
7980 (arc_verify_short): Likewise.
7981 (branch_dest): Likewise.
7982 (arc_attr_type): Likewise.
7983 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7984 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7985 (arc_final_prescan_insn): Remove inserting the nops due to
7986 hardware hazards. It is done in reorg step.
7987 (insn_length_variant_t): Remove.
7988 (insn_length_parameters_t): Likewise.
7989 (arc_insn_length_parameters): Likewise.
7990 (arc_get_insn_variants): Likewise.
7991 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7993 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7996 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7999 PR tree-optimization/84687
8000 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8002 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8004 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8006 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8007 Rename to ppc_speculation_barrier.
8008 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8009 __builtin_ppc_speculation_barrier.
8011 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8014 * combine.c (combine_simplify_rtx): Don't try to simplify if
8015 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8018 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8020 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8021 to 32 bytes when compiling for POWER9.
8023 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8026 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8027 regparm >= 3 with no arg reg available also for calls with
8028 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8031 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8033 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8035 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8038 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8040 2018-03-05 Richard Biener <rguenther@suse.de>
8042 PR tree-optimization/84486
8043 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8044 When inserting a __builtin_assume_aligned call set the LHS
8045 SSA name alignment info accordingly.
8047 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8049 PR tree-optimization/84114
8050 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8051 Avoid reassociation of FLOAT_MODE addition.
8053 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8055 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8056 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8057 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8058 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8059 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8060 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8061 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8063 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8064 __builtin_ia32_wbinvd): New builtins.
8065 (SPECIAL_ARGS2): New.
8066 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8067 (SPECIAL_ARGS2): New.
8068 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8069 (ix86_valid_target_attribute_inner_p): Ditto.
8070 (ix86_init_mmx_sse_builtins): Add special_args2.
8071 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8072 TARGET_WBNOINVD_P): New.
8073 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8074 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8075 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8076 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8077 * config/i386/pconfigintrin.h: New file.
8078 * config/i386/wbnoinvdintrin.h: Ditto.
8079 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8081 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8083 2018-03-05 Richard Biener <rguenther@suse.de>
8085 PR tree-optimization/84670
8086 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8088 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8089 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8090 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8091 doesn't grow if all successors (recursively) were visited at least
8094 2018-03-05 Richard Biener <rguenther@suse.de>
8096 PR tree-optimization/84650
8097 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8098 if executed in the loop pipeline.
8100 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8102 * doc/configfiles.texi (Configuration Files): Move info about
8103 conditionalizing $target-protos.h to...
8104 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8105 differs from $target-protos.h.
8107 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8108 Chung-Ju Wu <jasonwucj@gmail.com>
8110 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8111 * config/nds32/nds32-multiple.md (setmemsi): Define.
8112 * config/nds32/nds32-memory-manipulation.c
8113 (nds32_gen_dup_4_byte_to_word_value): New.
8114 (emit_setmem_word_loop): New.
8115 (emit_setmem_byte_loop): New.
8116 (nds32_expand_setmem_loop): New.
8117 (nds32_expand_setmem_loop_v3m): New.
8118 (nds32_expand_setmem_unroll): New.
8119 (nds32_expand_setmem): New.
8121 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8122 Chung-Ju Wu <jasonwucj@gmail.com>
8124 * config/nds32/nds32-memory-manipulation.c
8125 (nds32_emit_load_store): New.
8126 (nds32_emit_post_inc_load_store): New.
8127 (nds32_emit_mem_move): New.
8128 (nds32_emit_mem_move_block): New.
8129 (nds32_expand_movmemsi_loop_unknown_size): New.
8130 (nds32_expand_movmemsi_loop_known_size): New.
8131 (nds32_expand_movmemsi_loop): New.
8132 (nds32_expand_movmemsi_unroll): New.
8133 (nds32_expand_movmemqi): Rename ...
8134 (nds32_expand_movmemsi): ... to this.
8135 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8136 (movmemsi): ... to this.
8137 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8138 (nds32_expand_movmemsi): ... to this.
8140 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8141 Monk Chiang <sh.chiang04@gmail.com>
8142 Chung-Ju Wu <jasonwucj@gmail.com>
8144 * config/nds32/nds32-protos.h
8145 (nds32_expand_load_multiple): New arguments.
8146 (nds32_expand_store_multiple): Ditto.
8147 (nds32_valid_multiple_load_store): Rename ...
8148 (nds32_valid_multiple_load_store_p): ... to this.
8149 * config/nds32/nds32-memory-manipulation.c
8150 (nds32_expand_load_multiple): Refine implementation.
8151 (nds32_expand_store_multiple): Ditto.
8152 * config/nds32/nds32-multiple.md
8153 (load_multiple): Update nds32_expand_load_multiple interface.
8154 (store_multiple): Update nds32_expand_store_multiple interface.
8155 * config/nds32/nds32-predicates.c
8156 (nds32_valid_multiple_load_store): Rename ...
8157 (nds32_valid_multiple_load_store_p): ... to this and refine
8159 * config/nds32/predicates.md
8160 (nds32_load_multiple_and_update_address_operation): New predicate.
8161 (nds32_store_multiple_and_update_address_operation): New predicate.
8163 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8164 Chung-Ju Wu <jasonwucj@gmail.com>
8166 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8167 (combo): New attribute.
8168 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8170 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8172 * config/nds32/nds32.opt: Change -mcmodel= default value.
8174 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8175 Monk Chiang <sh.chiang04@gmail.com>
8176 Chung-Ju Wu <jasonwucj@gmail.com>
8178 * config/nds32/constants.md (unspec_element): New enum.
8179 * config/nds32/constraints.md (Umw): New constraint.
8180 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8181 * config/nds32/nds32-intrinsic.md: Likewise.
8182 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8183 (nds32_valid_smw_lwm_base_p): New.
8184 (nds32_output_smw_single_word): New.
8185 (nds32_output_lmw_single_word): New.
8186 (nds32_expand_unaligned_load): New.
8187 (nds32_expand_unaligned_store): New.
8188 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8189 (nds32_output_smw_single_word): Declare.
8190 (nds32_output_lmw_single_word): Declare.
8191 (nds32_expand_unaligned_load): Declare.
8192 (nds32_expand_unaligned_store): Declare.
8193 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8194 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8195 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8196 NDS32_BUILTIN_UASTORE_DW.
8197 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8200 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8201 Kito Cheng <kito.cheng@gmail.com>
8202 Chung-Ju Wu <jasonwucj@gmail.com>
8204 * config/nds32/nds32-intrinsic.c
8205 (nds32_expand_builtin_null_ftype_reg): Delete.
8206 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8207 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8208 (nds32_read_argument): New.
8209 (nds32_legitimize_target): Ditto.
8210 (nds32_legitimize_argument): Ditto.
8211 (nds32_check_constant_argument): Ditto.
8212 (nds32_expand_unop_builtin): Ditto.
8213 (nds32_expand_unopimm_builtin): Ditto.
8214 (nds32_expand_binop_builtin): Ditto.
8215 (nds32_builtin_decl_impl): Ditto.
8216 (builtin_description): Ditto.
8217 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8218 (nds32_init_builtins_impl): Ditto.
8219 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8220 (nds32_builtin_decl): New.
8221 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8222 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8224 2018-03-02 Jeff Law <law@redhat.com>
8226 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8227 (redundant_insn, fill_simple_delay_slots): Likewise.
8228 (fill_slots_from_thread): Likewise.
8229 * resource.c (mark_referenced_resources): Likewise.
8230 (mark_set_resources, find_dead_or_set_registers): Likewise.
8232 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8234 * substring-locations.h (format_warning_va): Formatting fix for
8236 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8238 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8239 * substring-locations.c: Include intl.h.
8240 (format_warning_va): Turned into small wrapper around
8241 format_warning_n_va, renamed to ...
8242 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8243 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8245 (format_warning_at_substring_n): New function.
8246 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8247 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8248 format_warning_at_substring with just a shorter name instead of
8249 const function pointer.
8250 (fmtwarn_n): New function.
8251 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8252 appropriate, get rid of all the fmtstr temporaries, move conditionals
8253 with G_() wrapped string literals directly into fmtwarn arguments,
8254 cast dir.len to (int), formatting fixes.
8256 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8258 * doc/invoke.texi: Remove "Cilk Plus" references.
8260 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8261 Richard Biener <rguenther@suse.de>
8264 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8265 for error or warning attributes if CALL_FROM_THUNK_P is set.
8268 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8271 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8272 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8275 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8276 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8278 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8279 instead of -1U in last predictors element's probability member.
8281 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8284 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8285 arguments if they are comparable.
8287 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8289 PR tree-optimization/84634
8290 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8291 masks and masked_loop_p with a single loop_masks, making sure it's
8292 null for bb vectorization.
8294 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8296 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8297 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8298 loop->force_vectorize to check whether there is no alias.
8300 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8303 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8305 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8307 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8308 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8309 instead of a loop around prev_real_insn.
8310 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8314 * config/i386/i386.c (ix86_print_operand): Use conditional
8315 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8318 2018-03-02 Richard Biener <rguenther@suse.de>
8320 PR tree-optimization/84427
8321 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8322 (bitmap_set_subtract_values): Rewrite to handle multiple
8325 (prune_clobbered_mems): Likewise.
8326 (phi_translate): Take edge instead of pred/phiblock.
8327 (phi_translate_1): Likewise.
8328 (phi_translate_set): Likewise. Insert all translated
8329 exprs for a value into the set, keeping possibly multiple
8330 expressions per value.
8331 (compute_antic_aux): Adjust for phi_translate changes.
8332 When intersecting union the expressions and prune those
8333 not in the final value set, keeping possibly multiple
8334 expressions per value. Do not use value-insertion
8335 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8336 all expressions. Add verification that the value-sets
8337 only shrink during iteration.
8338 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8339 (do_pre_regular_insertion): Likewise.
8340 (do_pre_partial_partial_insertion): Likewise.
8342 2018-03-02 Richard Biener <rguenther@suse.de>
8345 * config/darwin.c (saved_debug_info_level): New static global.
8346 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8347 (darwin_asm_lto_end): Restore debug info generation settings.
8349 2018-03-01 Martin Liska <mliska@suse.cz>
8352 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8355 2018-03-01 Richard Biener <rguenther@suse.de>
8358 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8359 type annotation with LTO when debug was disabled at compile-time.
8361 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8363 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8365 (mips_final_postscan_insn): Likewise.
8367 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8369 PR rtl-optimization/84528
8370 * alias.c (init_alias_target): Add commentary.
8371 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8372 a unique base value if the frame pointer is not eliminated
8373 to the stack pointer.
8375 2018-03-01 Tom de Vries <tom@codesourcery.com>
8377 PR rtl-optimization/83327
8378 * lra-int.h (hard_regs_spilled_into): Declare.
8379 * lra.c (hard_regs_spilled_into): Define.
8380 (init_reg_info): Init hard_regs_spilled_into.
8381 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8382 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8383 (process_bb_lives): Handle hard_regs_spilled_into.
8384 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8385 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8387 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8389 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8390 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8391 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8392 * config/rs6000/aix72.h: New file.
8394 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8396 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8397 instead of warning_at with conditional singular and plural messages
8401 * stor-layout.c (update_alignment_for_field): For
8402 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8403 && !DECL_PACKED (field), do the alignment update, just use
8404 only desired_align instead of MAX (type_align, desired_align)
8406 (place_field): Don't do known_align < desired_align handling
8407 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8408 is non-NULL, instead do it after rli->prev_field handling and
8409 only if not within a bitfield word. For DECL_PACKED (field)
8410 use type_align of BITS_PER_UNIT.
8412 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8414 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8415 superfluous parentheses and trailing spaces.
8417 2018-02-28 Richard Biener <rguenther@suse.de>
8419 PR tree-optimization/84584
8420 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8421 SCoPs with fake exit edge.
8423 2018-02-28 Martin Liska <mliska@suse.cz>
8426 * timevar.c (timer::print): Fix format to properly print 100%
8429 2018-02-28 Richard Biener <rguenther@suse.de>
8432 * genmatch.c (capture_info::walk_match): Do not mark
8433 captured expressions without operands as expr_p given
8434 they act more like predicates and should be subject to
8435 "lost tail" side-effect preserving.
8437 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8439 PR rtl-optimization/81611
8440 * auto-inc-dec.c (attempt_change): Move dead note from
8441 mem_insn if it's the next use of regno
8442 (find_address): Take address use of reg holding
8443 non-incremented value. Add parm to limit search to the named
8445 (merge_in_block): Attempt to use a mem insn that is the next
8446 use of the original regno.
8448 2018-02-27 Martin Sebor <msebor@redhat.com>
8451 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8452 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8454 2018-02-27 Martin Sebor <msebor@redhat.com>
8456 PR translation/84207
8457 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8458 n argument to unsigned HOST_WIDE_INT.
8459 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8460 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8461 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8462 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8464 2018-02-27 Richard Biener <rguenther@suse.de>
8466 PR tree-optimization/84512
8467 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8468 Do not use the estimate returned from record_stmt_cost for
8469 the scalar iteration cost but sum properly using add_stmt_cost.
8471 2018-02-27 Richard Biener <rguenther@suse.de>
8473 PR tree-optimization/84466
8474 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8475 Adjust last change to less strictly validate use operands.
8477 2018-02-27 Martin Liska <mliska@suse.cz>
8479 PR gcov-profile/84548
8480 * gcov.c (process_file): Allow partial overlap and consider it
8481 also as group functions.
8482 (output_lines): Properly calculate range of lines for a group.
8484 2018-02-27 Martin Liska <mliska@suse.cz>
8486 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8487 'ggc' suffixes. Change first column width.
8488 (timer::print): Fix formatting of the column.
8490 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8492 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8493 preserve inline entry blocks for the sake of debug inline
8494 entry point markers alone.
8495 (remove_unused_locals): Suggest in comments a better place to
8496 force the preservation of inline entry blocks that are
8497 otherwise unused, but do not preserve them.
8499 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8501 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8503 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8506 * config/i386/constraints.md (Bs): Replace
8507 ix86_indirect_branch_register with
8508 TARGET_INDIRECT_BRANCH_REGISTER.
8510 * config/i386/i386.md (indirect_jump): Likewise.
8511 (tablejump): Likewise.
8512 (*sibcall_memory): Likewise.
8513 (*sibcall_value_memory): Likewise.
8514 Peepholes of indirect call and jump via memory: Likewise.
8515 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8516 (*sibcall_value_GOT_32): Likewise.
8517 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8518 (GOT_memory_operand): Likewise.
8519 (call_insn_operand): Likewise.
8520 (sibcall_insn_operand): Likewise.
8521 (GOT32_symbol_operand): Likewise.
8522 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8524 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8526 PR rtl-optimization/83496
8527 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8528 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8529 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8530 redundant insn, if any.
8531 (relax_delay_slots): Likewise.
8532 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8534 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8536 PR tree-optimization/83965
8537 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8538 that grouped statements are part of a reduction chain. Return
8539 true if the statement is not marked as a reduction itself but
8541 (vect_recog_dot_prod_pattern): Don't check whether the statement
8542 is part of a group here.
8543 (vect_recog_sad_pattern): Likewise.
8544 (vect_recog_widen_sum_pattern): Likewise.
8546 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8549 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8551 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8554 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8556 (ix86_output_indirect_function_return): New prototype.
8557 (ix86_split_simple_return_pop_internal): Likewise.
8558 * config/i386/i386.c (indirect_return_via_cx): New.
8559 (indirect_return_via_cx_bnd): Likewise.
8560 (indirect_thunk_name): Handle return va CX_REG.
8561 (output_indirect_thunk_function): Create alias for
8562 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8563 (ix86_output_indirect_jmp): Remove the bool argument.
8564 (ix86_output_indirect_function_return): New function.
8565 (ix86_split_simple_return_pop_internal): Likewise.
8566 * config/i386/i386.md (*indirect_jump): Don't pass false
8567 to ix86_output_indirect_jmp.
8568 (*tablejump_1): Likewise.
8569 (simple_return_pop_internal): Change it to define_insn_and_split.
8570 Call ix86_split_simple_return_pop_internal to split it for
8572 (simple_return_indirect_internal): Call
8573 ix86_output_indirect_function_return instead of
8574 ix86_output_indirect_jmp.
8576 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8579 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8580 memset and value initialization afterwards.
8582 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8584 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8586 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8589 * common/config/aarch64/aarch64-common.c
8590 (aarch_option_optimization_table[]): Switch
8591 off fomit-frame-pointer
8593 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8594 Chung-Ju Wu <jasonwucj@gmail.com>
8596 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8598 (store_multiple): Ditto.
8600 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8602 * config.gcc: Add --with-cpu support for nds32 target.
8603 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8604 * config/nds32/nds32.opt: Add -mcpu= option.
8606 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8608 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8609 isel=yes): Warn for these deprecated options.
8611 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8613 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8614 ISA_2_5_MASKS_EMBEDDED.
8616 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8618 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8619 p->max as pointers rather than using iterative_hash_expr.
8621 2018-02-23 Carl Love <cel@us.ibm.com>
8623 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8624 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8626 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8627 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8628 P8V_BUILTIN_VEC_VUNSIGNED2.
8630 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8633 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8634 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8636 * lra-constraints.c (curr_insn_transform): Set up
8637 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8639 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8640 * lra-eliminations.c (spill_pseudos): Ditto.
8641 (process_insn_for_elimination): Ditto.
8642 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8643 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8644 LRA_NON_CLOBBERED_ALT.
8646 2018-02-22 Martin Sebor <msebor@redhat.com>
8648 PR tree-optimization/84480
8649 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8650 to maybe_diag_stxncpy_trunc. Call it.
8651 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8652 from gimple_fold_builtin_strcpy. Print inlining stack.
8653 (handle_builtin_stxncpy): Print inlining stack.
8654 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8656 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8659 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8660 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8661 and -fcheck-pointer-bounds are used together.
8662 (indirect_thunk_prefix): New enum.
8663 (indirect_thunk_need_prefix): New function.
8664 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8665 "_nt" instead of "_bnd" for NOTRACK prefix.
8666 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8667 (output_indirect_thunk_function): Likewise.
8669 (ix86_code_end): Update output_indirect_thunk_function calls.
8670 (ix86_output_indirect_branch_via_reg): Replace
8671 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8672 (ix86_output_indirect_branch_via_push): Likewise.
8673 (ix86_output_function_return): Likewise.
8674 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8675 incompatible with -fcf-protection=branch and
8676 -fcheck-pointer-bounds.
8678 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8681 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8682 Change gcc_assert call to output_operand_lossage.
8684 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8686 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8688 2018-02-22 DJ Delorie <dj@redhat.com>
8689 Sebastian Perta <sebastian.perta@renesas.com>
8690 Oleg Endo <olegendo@gcc.gnu.org>
8692 * config/rx/rx.c (rx_rtx_costs): New function.
8693 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8695 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8697 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8699 2018-02-22 Martin Liska <mliska@suse.cz>
8702 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8703 Add "native" as a possible value.
8705 2018-02-22 Martin Liska <mliska@suse.cz>
8708 * config/i386/i386.c (ix86_option_override_internal):
8709 Add "native" as a possible value for -march and -mtune.
8711 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8714 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8715 to all type variants.
8717 PR tree-optimization/84503
8718 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8719 width as info->bitpos + info->bitsize - start.
8720 (merged_store_group::merge_overlapping): Simplify width computation.
8721 (check_no_overlap): New function.
8722 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8723 start + width and last_order of the group, fail if check_no_overlap
8725 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8726 to group if check_no_overlap fails.
8728 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8730 * config/rs6000/altivec.md: Delete contraint arguments to
8731 define_expand, define_split, and define_peephole2, and in
8732 define_insn_and_split if always unused.
8733 * config/rs6000/darwin.md: Ditto.
8734 * config/rs6000/dfp.md: Ditto.
8735 * config/rs6000/rs6000.md: Ditto.
8736 * config/rs6000/sync.md: Ditto.
8737 * config/rs6000/vector.md: Ditto.
8738 * config/rs6000/vsx.md: Ditto.
8740 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8742 * config/rs6000/altivec.md: Write output control strings as braced
8743 blocks instead of double-quoted strings.
8744 * config/rs6000/darwin.md: Ditto.
8745 * config/rs6000/rs6000.md: Ditto.
8746 * config/rs6000/vector.md: Ditto.
8747 * config/rs6000/vsx.md: Ditto.
8749 2018-02-21 Jason Merrill <jason@redhat.com>
8751 PR c++/84314 - ICE with templates and fastcall attribute.
8752 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8754 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8756 * ipa-cp.c (determine_versionability): Fix comment typos.
8758 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8761 * ipa-cp.c (determine_versionability): Do not version functions caling
8764 2018-02-21 Martin Liska <mliska@suse.cz>
8767 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8768 Add "native" as a possible value.
8769 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8770 the macro when native cpu detection is available.
8772 2018-02-21 Martin Liska <mliska@suse.cz>
8775 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8776 Add "native" as a possible value.
8777 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8778 when native cpu detection is available.
8780 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8781 Martin Sebor <msebor@redhat.com>
8783 PR tree-optimization/84478
8784 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8786 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8787 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8788 support which is conservatively correct, for 2 only stay conservative
8789 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8790 argument to the 2 argument get_range_strlen, adjust 6 arg
8791 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8793 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8794 (gimple_fold_builtin_strlen): Pass true as last argument to
8797 2018-02-20 Martin Sebor <msebor@redhat.com>
8800 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8801 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8802 (builtin_memref::builtin_memref): Factor out parts into
8803 set_base_and_offset and call it.
8805 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8808 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8809 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8810 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8811 search at the associated MODE_INT.
8813 2018-02-20 Jeff Law <law@redhat.com>
8816 PR tree-optimization/81592
8818 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8819 for range data rather than using global data.
8820 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8821 range data rather than using global data.
8822 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8823 pass it to children as needed.
8824 (struct directive::fmtresult): Similarly.
8825 (struct directive::set_width): Similarly.
8826 (struct directive::set_precision): Similarly.
8827 (format_integer, format_directive, parse_directive): Similarly.
8828 (format_none): Accept unnamed vr_values parameter.
8829 (format_percent, format_floating, format_character): Similarly.
8830 (format_string, format_plain): Similarly.
8831 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8832 the EVRP range analyzer for range data rather than using global data.
8833 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8834 gimple-ssa-evrp-analyze.h
8835 (class sprintf_dom_walker): Add after_dom_children member function.
8836 Add evrp_range_analyzer member.
8837 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8838 range analyzer as needed.
8839 (sprintf_dom_walker::after_dom_children): New member function.
8840 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8842 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8843 (evrp_range_analyzer::pop_to_marker): Likewise.
8845 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8847 PR tree-optimization/84419
8848 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8849 with the required type if its current type is compatible but
8852 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8855 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8856 after vectorization.
8858 2018-02-20 Martin Liska <mliska@suse.cz>
8861 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8862 possible values if we don't have a hint.
8864 2018-02-20 Martin Liska <mliska@suse.cz>
8868 * final.c (shorten_branches): Build align_tab array with one
8870 * opts.c (finish_options): Add alignment option limit check.
8871 (MAX_CODE_ALIGN): Likewise.
8872 (MAX_CODE_ALIGN_VALUE): Likewise.
8873 * doc/invoke.texi: Document maximum allowed option value for
8874 all -falign-* options.
8876 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8879 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8880 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8881 * var-tracking.c (emit_note_insn_var_location): Remove all references
8882 to NOTE_INSN_CALL_ARG_LOCATION.
8883 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8884 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8885 Use copy_rtx_if_shared.
8886 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8887 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8888 (dwarf2out_var_location): Remove handling of
8889 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8891 * final.c (final_scan_insn): Remove all references to
8892 NOTE_INSN_CALL_ARG_LOCATION.
8893 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8894 before dumping final insns.
8895 * except.c (emit_note_eh_region_end): Remove all references to
8896 NOTE_INSN_CALL_ARG_LOCATION.
8897 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8898 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8899 * config/arc/arc.c (hwloop_optimize): Likewise.
8900 * config/arm/arm.c (create_fix_barrier): Likewise.
8901 * config/s390/s390.c (s390_chunkify_start): Likewise.
8902 * config/sh/sh.c (find_barrier): Likewise.
8903 * config/i386/i386.c (rest_of_insert_endbranch,
8904 ix86_seh_fixup_eh_fallthru): Likewise.
8905 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8906 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8907 * config/frv/frv.c (frv_function_prologue): Likewise.
8908 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8910 (note_outside_basic_block_p): Remove all references to
8911 NOTE_INSN_CALL_ARG_LOCATION.
8912 * gengtype.c (adjust_field_rtx_def): Likewise.
8913 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8915 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8916 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8919 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8922 PR tree-optimization/84452
8923 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8924 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8927 2018-02-19 Martin Liska <mliska@suse.cz>
8930 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8932 2018-02-19 Martin Liska <mliska@suse.cz>
8933 Richard Sandiford <richard.sandiford@linaro.org>
8935 PR tree-optimization/82491
8936 * gimple-fold.c (get_base_constructor): Make earlier bail out
8939 2018-02-19 Carl Love <cel@us.ibm.com>
8941 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8942 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8944 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8945 P8V_BUILTIN_VEC_NEG.
8947 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8949 * config/rl78/rl78.md (movdf): New define expand.
8951 2018-02-19 Martin Liska <mliska@suse.cz>
8954 * doc/invoke.texi: Fix typo.
8955 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8957 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8959 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8960 handle rs6000_single_float and rs6000_double_float specially for
8963 2018-02-16 Jeff Law <law@redhat.com>
8965 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8968 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8971 * ipa-inline.c (inline_small_functions): Fix a typo.
8973 2018-02-16 Nathan Sidwell <nathan@acm.org>
8975 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8977 2018-02-16 Carl Love <cel@us.ibm.com>
8979 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8980 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8981 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8982 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8983 expansion to P8V_BUILTIN_VEC_FLOAT2.
8985 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8987 PR rtl-optimization/70023
8988 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8989 src_regno into account.
8991 2018-02-16 Carl Love <cel@us.ibm.com>
8993 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8994 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8995 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8996 * config/rs6000/rs6000.c: Remove case statements for
8997 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8998 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8999 and P9V_BUILTIN_VEC_VINSERT4B.
9000 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9001 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9002 * config/rs6000/vsx.md:
9003 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9006 2018-02-16 Carl Love <cel@us.ibm.com>
9008 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9010 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9012 * config/rs6000/rs6000-c.c: Add the definitions for
9013 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9014 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9015 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9016 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9017 definition for insert4b and define insn *insert3b_internal.
9018 * doc/extend.texi: Add documentation for vec_extract4b.
9020 2018-02-16 Nathan Sidwell <nathan@acm.org>
9022 * doc/extend.texi (Backwards Compatibility): Mention friend
9023 injection. Note for-scope is deprecated.
9024 * doc/invoke.texi (-ffriend-injection): Deprecate.
9026 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9028 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9029 that moved to I2, also allow destinations that are a paradoxical
9030 subreg (instead of a normal reg).
9032 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9035 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9038 2018-02-16 Richard Biener <rguenther@suse.de>
9040 PR tree-optimization/84037
9041 PR tree-optimization/84016
9043 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9044 Adjust vec_construct for the fact we need additional higher latency
9045 128bit inserts for AVX256 and AVX512 vector builds.
9046 (ix86_add_stmt_cost): Scale vector construction cost for
9049 2018-02-16 Richard Biener <rguenther@suse.de>
9051 PR tree-optimization/84417
9052 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9053 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9054 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9056 2018-02-16 Martin Liska <mliska@suse.cz>
9059 * internal-fn.def (ASAN_CHECK): Set proper flags.
9060 (ASAN_MARK): Likewise.
9062 2018-02-16 Julia Koval <julia.koval@intel.com>
9064 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9065 from PTA_CANNONLAKE.
9067 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9070 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9071 Use ++iter rather than iter++ for std::list iterators.
9072 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9073 defer deleting them until all nodes in the forest are processed. Do
9074 free even leaf nodes. Change to_process into auto_vec.
9077 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9078 * vec.h (vec_default_construct): Use memset instead of placement new
9079 if BROKEN_VALUE_INITIALIZATION is defined.
9080 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9081 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9084 PR rtl-optimization/83723
9085 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9086 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9087 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9089 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9091 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9093 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9095 PR rtl-optimization/81443
9096 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9097 from inner REGs to paradoxical SUBREGs.
9099 2018-02-16 Richard Biener <rguenther@suse.de>
9101 PR tree-optimization/84399
9102 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9103 For operands we can analyze at their definition make sure we can
9104 analyze them at each use as well.
9106 2018-02-16 Richard Biener <rguenther@suse.de>
9108 PR tree-optimization/84190
9109 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9110 volatile accesses if the decl isn't volatile.
9112 2018-02-15 Jason Merrill <jason@redhat.com>
9114 PR c++/84314 - ICE with templates and fastcall attribute.
9115 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9116 TYPE_CANONICAL on an existing type.
9118 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9120 PR tree-optimization/84383
9121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9122 dstoff nor call operand_equal_p if dstbase is NULL.
9124 PR tree-optimization/84334
9125 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9126 also a CONSTANT_CLASS_P, punt.
9128 2018-02-14 Jim Wilson <jimw@sifive.com>
9130 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9131 first SMALL_OPERAND check. New local min_second_step. Move assert
9132 to where locals are set. Add TARGET_RVC support.
9133 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9135 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9137 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9139 2018-02-14 Martin Sebor <msebor@redhat.com>
9141 PR tree-optimization/83698
9142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9143 arrays constrain the offset range to their bounds.
9144 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9145 (builtin_access::overlap): Avoid setting the size of overlap if it's
9147 (maybe_diag_overlap): Also consider arrays when deciding what values
9148 of offsets to include in diagnostics.
9150 2018-02-14 Martin Sebor <msebor@redhat.com>
9153 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9154 that correspond to the kind of a declaration.
9156 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9159 * config/pa/pa.md: Load address of PIC label using the linkage table
9160 if the label is nonlocal.
9162 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9164 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9165 warning message if user requests -maltivec=be.
9166 * doc/invoke.texi: Document deprecation of -maltivec=be.
9168 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9171 * config/rs6000/rs6000-c.c: Update definitions for
9172 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9173 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9175 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9178 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9179 add _get_ssp intrinsics. Remove argument from
9180 __builtin_ia32_rdssp[d|q].
9181 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9182 * config/i386/i386-builtin.def: Remove argument from
9183 __builtin_ia32_rdssp[d|q].
9184 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9185 ix86_expand_special_args_builtin for _rdssp[d|q].
9186 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9187 Clear register before usage.
9188 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9189 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9191 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9193 PR tree-optimization/84357
9194 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9195 operand 1 of an ARRAY_REF too.
9197 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9200 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9201 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9203 (set_of_reg): New struct.
9204 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9205 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9206 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9208 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9209 Split into bitclr, bitset, bitinvert patterns if appropriate.
9210 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9211 use rx_fuse_in_memory_bitop.
9212 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9213 to named insn, correct maximum insn length.
9215 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9218 * machmode.def: Define a complex mode for PARTIAL_INT.
9219 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9221 * doc/rtl.texi: Document CSPImode.
9222 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9224 (msp430_hard_regno_nregs_with_padding): Likewise.
9226 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9229 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9231 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9233 PR rtl-optimization/84169
9234 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9235 we generated a parallel as new i3 and we split that to new i2 and i3
9236 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9237 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9238 those to i2, not i1. Partially rewrite this scan code.
9240 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9243 * stor-layout.c (place_field): For variable length fields, adjust
9244 offset_align afterwards not just based on the field's alignment,
9245 but also on the size.
9248 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9249 of exps and logs in the use_exp2 case.
9251 2018-02-13 Jeff Law <law@redhat.com>
9253 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9256 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9259 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9263 * common.opt (gas-loc-support, gas-locview-support): New.
9264 (ginline-points, ginternal-reset-location-views): New.
9265 * doc/invoke.texi: Document them. Use @itemx where intended.
9266 (gvariable-location-views): Adjust.
9267 * target.def (reset_location_view): New.
9268 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9269 (TARGET_RESET_LOCATION_VIEW): New.
9270 * doc/tm.texi: Rebuilt.
9271 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9272 (dwarf2out_default_as_locview_support): New.
9273 (output_asm_line_debug_info): Use option variables.
9274 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9275 (output_loc_list): Likewise.
9276 (add_high_low_attributes): Check option variables.
9277 Don't output entry view attribute in strict mode.
9278 (gen_inlined_subroutine_die): Check option variables.
9279 (dwarf2out_inline_entry): Likewise.
9280 (init_sections_and_labels): Likewise.
9281 (dwarf2out_early_finish): Likewise.
9282 (maybe_reset_location_view): New, from...
9283 (dwarf2out_var_location): ... here. Call it.
9284 * debug.h (dwarf2out_default_as_loc_support): Declare.
9285 (dwarf2out_default_as_locview_support): Declare.
9286 * hooks.c (hook_int_rtx_insn_0): New.
9287 * hooks.h (hook_int_rtx_insn_0): Declare.
9288 * toplev.c (process_options): Take -gas-loc-support and
9289 -gas-locview-support from dwarf2out. Enable
9290 -gvariable-location-views by default only with locview
9291 assembler support. Enable -ginternal-reset-location-views by
9292 default only if the target defines the corresponding hook.
9293 Enable -ginline-points by default if location views are
9294 enabled; force it disabled if statement frontiers are
9296 * tree-inline.c (expand_call_inline): Check option variables.
9297 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9299 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9301 PR tree-optimization/84321
9302 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9303 handling. Also check whether the anti-range contains any values
9304 that satisfy the mask; switch to a VR_RANGE if not.
9306 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9309 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9311 2018-02-13 Martin Jambor <mjambor@suse.cz>
9314 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9315 of call statements, also set location of a load to a temporary.
9317 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9319 * config/rl78/rl78.c (add_vector_labels): New function.
9320 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9321 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9322 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9323 which checks that no arguments are passed.
9324 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9325 * doc/extend.texi: Documentation for the new attribute.
9327 2018-02-13 Andreas Schwab <schwab@suse.de>
9329 * config/riscv/linux.h (CPP_SPEC): Define.
9331 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9334 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9335 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9336 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9337 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9338 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9339 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9340 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9341 temporarily for AES and PCLMUL builtins.
9343 PR tree-optimization/84339
9344 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9345 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9349 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9350 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9351 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9353 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9355 * omp-simd-clone.h: New file.
9356 * omp-simd-clone.c: Include omp-simd-clone.h.
9357 (expand_simd_clones): No longer static.
9358 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9359 cgraph.h and omp-simd-clone.h.
9360 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9361 (vect_recog_widen_shift_pattern): Formatting fix.
9362 (vect_pattern_recog_1): Don't check optab for calls.
9365 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9366 operands[2] into a REG before using gen_lowpart on it.
9368 2018-02-12 Jeff Law <law@redhat.com>
9371 * config/sh/sh.c (find_barrier): Consider a sibling call
9374 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9375 successfully back substituting a reg.
9377 2018-02-12 Richard Biener <rguenther@suse.de>
9379 PR tree-optimization/84037
9380 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9381 parameter, move visited init to caller.
9382 (vect_slp_analyze_operations): Separate cost from validity
9383 check, initialize visited once for all instances.
9384 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9386 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9388 (vect_model_promotion_demotion_cost): Likewise.
9389 (vectorizable_bswap): Guard cost modeling with !slp_node
9390 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9392 (vectorizable_call): Likewise.
9393 (vectorizable_conversion): Likewise.
9394 (vectorizable_assignment): Likewise.
9395 (vectorizable_shift): Likewise.
9396 (vectorizable_operation): Likewise.
9397 (vectorizable_store): Likewise.
9398 (vectorizable_load): Likewise.
9399 (vectorizable_condition): Likewise.
9400 (vectorizable_comparison): Likewise.
9402 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9405 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9406 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9407 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9409 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9412 * params.def (inline-min-speedup): Increase from 8 to 15.
9413 (max-inline-insns-auto): Decrease from 40 to 30.
9414 * ipa-split.c (consider_split): Add some buffer for function to
9415 be considered inlining candidate.
9416 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9419 2018-02-12 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/84037
9422 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9423 matched stmts if we cannot swap the non-matched ones.
9425 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9427 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9428 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9429 _mm_maskz_scalef_round_ss): New intrinsics.
9430 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9431 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9432 __builtin_ia32_scalefss_round): Remove.
9433 (__builtin_ia32_scalefsd_mask_round,
9434 __builtin_ia32_scalefss_mask_round): New intrinsics.
9435 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9436 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9437 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9438 "<round_constraint>")): Changed to ...
9439 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9440 "<round_scalar_constraint>")): ... this.
9441 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9442 %0, %1, %2<round_op3>}"): Changed to ...
9443 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9444 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9445 %2<round_scalar_mask_op3>}"): ... this.
9446 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9448 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9450 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9451 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9452 (_mm_maskz_sqrt_round_ss): New intrinsics.
9453 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9454 (__builtin_ia32_sqrtsd_mask_round)
9455 (__builtin_ia32_sqrtss_mask_round): New builtins.
9456 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9457 (__builtin_ia32_sqrtss_round): Remove.
9458 (__builtin_ia32_sqrtsd_mask_round)
9459 (__builtin_ia32_sqrtss_mask_round): New builtins.
9460 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9461 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9462 ((match_operand:VF_128 1 "vector_operand"
9463 "xBm,<round_constraint>")): Changed to ...
9464 ((match_operand:VF_128 1 "vector_operand"
9465 "xBm,<round_scalar_constraint>")): ... this.
9466 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9467 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9468 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9469 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9470 %<iptr>1<round_scalar_mask_op3>}): ... this.
9471 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9472 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9474 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9477 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9478 Cast vec_cmpeq result to correct type.
9479 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9480 Cast vec_cmpgt result to correct type.
9482 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9484 * final.c (final_scan_insn_1): Renamed from...
9485 (final_scan_insn): ... this. New wrapper, to recover
9486 seen from the outermost call in recursive ones.
9487 * config/sparc/sparc.c (output_return): Drop seen from call.
9488 (output_sibcall): Likewise.
9489 * config/visium/visium.c (output_branch): Likewise.
9491 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9493 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9496 2018-02-10 Alan Modra <amodra@gmail.com>
9499 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9500 Specify LR as an input.
9502 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9505 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9506 remove_member_access_dummy_vars): New functions.
9507 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9508 lower_omp_1, execute_lower_omp): Use them.
9510 PR rtl-optimization/84308
9511 * shrink-wrap.c (spread_components): Release todo vector.
9513 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9515 PR rtl-optimization/57193
9516 * ira-color.c (struct allocno_color_data): Add member
9517 conflict_allocno_hard_prefs.
9518 (update_conflict_allocno_hard_prefs): New.
9519 (bucket_allocno_compare_func): Add a preference based on
9520 conflict_allocno_hard_prefs.
9521 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9522 (color_allocnos): Remove a dead code. Initiate
9523 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9525 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9528 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9529 constraint from =wa to wa. Avoid a subreg on the output operand,
9530 instead use a pseudo and subreg it in a move.
9531 (p9_xxbrd_<mode>): Changed to ...
9532 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9533 (p9_xxbrd_v2df): New expander.
9534 (p9_xxbrw_<mode>): Changed to ...
9535 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9536 (p9_xxbrw_v4sf): New expander.
9538 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9540 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9542 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9545 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9546 multiply in 32-bit mode.
9547 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9548 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9551 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9553 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9554 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9555 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9556 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9558 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9561 * dwarf2out.c (is_trivial_indirect_ref): New function.
9562 (dwarf2out_late_global_decl): Do not generate a location
9563 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9564 and that are not defined in the current unit.
9566 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9568 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9569 instead of a libcall for UNORDERED.
9571 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9574 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9575 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9577 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9580 * config/s390/s390.c (s390_set_current_function): Invoke
9581 s390_indirect_branch_settings also if fndecl didn't change.
9583 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9585 * config/rs6000/rs6000.md (blockage): Set length to zero.
9587 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9589 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9591 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9594 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9595 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9599 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9600 PARALLEL incoming that failed vt_get_decl_and_offset check.
9603 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9604 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9606 (get_variable_section): For decls in named .bss* sections pass true as
9607 second argument to bss_initializer_p.
9609 2018-02-09 Marek Polacek <polacek@redhat.com>
9610 Jakub Jelinek <jakub@redhat.com>
9613 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9614 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9615 Sync some changes from cxx_fold_indirect_ref.
9617 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9619 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9621 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9622 (BLOCK_INLINE_ENTRY_LABEL): New.
9623 (dwarf2out_var_location): Disregard inline entry markers.
9624 (inline_entry_data): New struct.
9625 (inline_entry_data_hasher): New hashtable type.
9626 (inline_entry_data_hasher::hash): New.
9627 (inline_entry_data_hasher::equal): New.
9628 (inline_entry_data_table): New variable.
9629 (add_high_low_attributes): Add DW_AT_entry_pc and
9630 DW_AT_GNU_entry_view attributes if a pending entry is found
9631 in inline_entry_data_table. Add old entry_pc attribute only
9632 if debug nonbinding markers are disabled.
9633 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9634 markers are enabled.
9635 (block_within_block_p, dwarf2out_inline_entry): New.
9636 (dwarf2out_finish): Check that no entries remained in
9637 inline_entry_data_table.
9638 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9639 (final_scan_insn, notice_source_line): Likewise.
9640 (rest_of_clean_state): Skip inline entry markers.
9641 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9643 * gimple.c (gimple_build_debug_inline_entry): New.
9644 * gimple.h (enum gimple_debug_subcode): Add
9645 GIMPLE_DEBUG_INLINE_ENTRY.
9646 (gimple_build_debug_inline_entry): Declare.
9647 (gimple_debug_inline_entry_p): New.
9648 (gimple_debug_nonbind_marker_p): Adjust.
9649 * insn-notes.def (INLINE_ENTRY): New.
9650 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9651 inline entry marker notes.
9652 (print_insn): Likewise.
9653 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9654 (INSN_DEBUG_MARKER_KIND): Likewise.
9655 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9656 * tree-inline.c (expand_call_inline): Build and insert
9657 debug_inline_entry stmt.
9658 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9659 inline entry blocks early, if nonbind markers are enabled.
9660 (dump_scope_block): Dump fragment info.
9661 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9662 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9663 (gimple_build_debug_inline_entry): New.
9664 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9665 Enable/disable inline entry points too.
9666 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9667 (DEBUG_INSN): Describe inline entry markers.
9669 * common.opt (gvariable-location-views): New.
9670 (gvariable-location-views=incompat5): New.
9671 * config.in: Rebuilt.
9672 * configure: Rebuilt.
9673 * configure.ac: Test assembler for view support.
9674 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9675 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9676 * dwarf2out.c (var_loc_view): New typedef.
9677 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9678 (dwarf2out_locviews_in_attribute): New.
9679 (dwarf2out_locviews_in_loclist): New.
9680 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9681 (enum dw_line_info_opcode): Add LI_adv_address.
9682 (struct dw_line_info_table): Add view.
9683 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9684 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9685 (zero_view_p): New variable.
9686 (ZERO_VIEW_P): New macro.
9687 (output_asm_line_debug_info): New.
9688 (struct var_loc_node): Add view.
9689 (add_AT_view_list, AT_loc_list): New.
9690 (add_var_loc_to_decl): Add view param. Test it against last.
9691 (new_loc_list): Add view params. Record them.
9692 (AT_loc_list_ptr): Handle loc and view lists.
9693 (view_list_to_loc_list_val_node): New.
9694 (print_dw_val): Handle dw_val_class_view_list.
9695 (size_of_die): Likewise.
9696 (value_format): Likewise.
9697 (loc_list_has_views): New.
9698 (gen_llsym): Set vl_symbol too.
9699 (maybe_gen_llsym, skip_loc_list_entry): New.
9700 (dwarf2out_maybe_output_loclist_view_pair): New.
9701 (output_loc_list): Output view list or entries too.
9702 (output_view_list_offset): New.
9703 (output_die): Handle dw_val_class_view_list.
9704 (output_dwarf_version): New.
9705 (output_compilation_unit_header): Use it.
9706 (output_skeleton_debug_sections): Likewise.
9707 (output_rnglists, output_line_info): Likewise.
9708 (output_pubnames, output_aranges): Update version comments.
9709 (output_one_line_info_table): Output view numbers in asm comments.
9710 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9711 Call maybe_gen_llsym.
9712 (loc_list_from_tree_1): Adjust.
9713 (add_AT_location_description): Create view list attribute if
9714 needed, check it's absent otherwise.
9715 (convert_cfa_to_fb_loc_list): Adjust.
9716 (maybe_emit_file): Call output_asm_line_debug_info for test.
9717 (dwarf2out_var_location): Reset views as needed. Precompute
9718 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9719 attribute. Set view.
9720 (new_line_info_table): Reset next view.
9721 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9722 (dwarf2out_source_line): Likewise. Output view resets and labels to
9723 the assembler, or select appropriate line info opcodes.
9724 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9725 (optimize_string_length): Catch it. Adjust.
9726 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9727 dw_val_class_view_list, and remove it if no longer needed.
9728 (hash_loc_list): Hash view numbers.
9729 (loc_list_hasher::equal): Compare them.
9730 (optimize_location_lists): Check whether a view list symbol is
9731 needed, and whether the locview attribute is present, and
9732 whether they match. Remove the locview attribute if no longer
9734 (index_location_lists): Call skip_loc_list_entry for test.
9735 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9736 Use output_dwarf_version.
9737 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9738 (struct dw_val_node): Add val_view_list.
9739 * final.c (SEEN_NEXT_VIEW): New.
9740 (set_next_view_needed): New.
9741 (clear_next_view_needed): New.
9742 (maybe_output_next_view): New.
9743 (final_start_function): Rename to...
9744 (final_start_function_1): ... this. Take pointer to FIRST,
9745 add SEEN parameter. Emit param bindings in the initial view.
9746 (final_start_function): Reintroduce SEEN-less interface.
9747 (final): Rename to...
9748 (final_1): ... this. Take SEEN parameter. Output final pending
9749 next view at the end.
9750 (final): Reintroduce seen-less interface.
9751 (final_scan_insn): Output pending next view before switching
9752 sections or ending a block. Mark the next view as needed when
9753 outputting variable locations. Notify debug backend of section
9754 changes, and of location view changes.
9755 (rest_of_handle_final): Adjust.
9756 * toplev.c (process_options): Autodetect value for debug variable
9757 location views option. Warn on incompat5 without -gdwarf-5.
9758 * doc/invoke.texi (gvariable-location-views): New.
9759 (gvariable-location-views=incompat5): New.
9760 (gno-variable-location-views): New.
9762 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9764 PR tree-optimization/84136
9765 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9766 that the result of find_edge is non-NULL.
9768 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9771 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9772 storing integer register in SImode. Fix cost of 256 and 512
9773 byte aligned SSE register store.
9775 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9777 * config/i386/i386.c (ix86_multiplication_cost): Fix
9778 multiplication cost for TARGET_AVX512DQ.
9780 2018-02-08 Marek Polacek <polacek@redhat.com>
9782 PR tree-optimization/84238
9783 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9786 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9788 PR tree-optimization/84265
9789 * tree-vect-stmts.c (vectorizable_store): Don't treat
9790 VMAT_CONTIGUOUS accesses as grouped.
9791 (vectorizable_load): Likewise.
9793 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9795 PR tree-optimization/81635
9796 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9797 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9798 (test_round_for_mask): New functions.
9799 (wide_int_cc_tests): Call test_round_for_mask.
9800 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9801 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9802 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9803 range returned by get_range_info.
9805 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9808 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9809 * symtab.c: Include builtins.h
9810 (symtab_node::output_to_lto_symbol_table_p): Move here
9811 from lto-streamer-out.c:output_symbol_p.
9812 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9813 (output_symbol_p): Move all logic to symtab.c
9814 (produce_symtab): Update.
9816 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9818 * config/s390/s390-opts.h (enum indirect_branch): Define.
9819 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9820 (s390_indirect_branch_via_thunk)
9821 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9822 (enum s390_indirect_branch_type): Define.
9823 * config/s390/s390.c (struct s390_frame_layout, struct
9824 machine_function): Remove.
9825 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9826 (indirect_branch_table_label_no, indirect_branch_table_name):
9828 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9829 (enum s390_indirect_branch_option): Define.
9830 (s390_return_addr_from_memory): New function.
9831 (s390_handle_string_attribute): New function.
9832 (s390_attribute_table): Add new attribute handler.
9833 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9834 (s390_indirect_branch_via_thunk): New function.
9835 (s390_indirect_branch_via_inline_thunk): New function.
9836 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9837 sibling call optimization for non z10 compiles.
9838 (s390_emit_call): Force indirect branch target to be a single
9839 register. Add r1 clobber for non-z10 compiles.
9840 (s390_emit_epilogue): Emit return jump via return_use expander.
9841 (s390_reorg): Handle JUMP_INSNs as execute targets.
9842 (s390_option_override_internal): Perform validity checks for the
9843 new command line options.
9844 (s390_indirect_branch_attrvalue): New function.
9845 (s390_indirect_branch_settings): New function.
9846 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9847 (s390_output_indirect_thunk_function): New function.
9848 (s390_code_end): Implement target hook.
9849 (s390_case_values_threshold): Implement target hook.
9850 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9852 * config/s390/s390.h (struct s390_frame_layout)
9853 (struct machine_function): Move here from s390.c.
9854 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9855 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9856 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9857 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9858 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9859 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9860 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9861 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9862 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9863 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9864 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9865 (mnemonic attribute): Add values which aren't recognized
9867 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9868 pattern for branch conversion. Fix mnemonic attribute.
9869 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9870 indirect branch via thunk if requested.
9871 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9872 ("*indirect_jump"): Disable for branch conversion using out of
9874 ("indirect_jump_via_thunk<mode>_z10")
9875 ("indirect_jump_via_thunk<mode>")
9876 ("indirect_jump_via_inlinethunk<mode>_z10")
9877 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9878 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9879 ("casesi_jump_via_inlinethunk<mode>_z10")
9880 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9881 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9882 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9883 ("*indirect2_jump"): Disable for branch conversion.
9884 ("casesi_jump"): Turn into expander and expand patterns for branch
9886 ("return_use"): New expander.
9887 ("*return"): Emit return via thunk and rename it to ...
9888 ("*return<mode>"): ... this one.
9889 * config/s390/s390.opt: Add new options and and enum for the
9892 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9894 * lra-constraints.c (match_reload): Unconditionally use
9895 gen_lowpart_SUBREG, rather than selecting between that
9896 and equivalent gen_rtx_SUBREG code.
9898 2018-02-08 Richard Biener <rguenther@suse.de>
9900 PR tree-optimization/84233
9901 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9902 changed flag instead of boguously re-using phi_inserted.
9904 2018-02-08 Martin Jambor <mjambor@suse.cz>
9906 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9907 static local variables.
9909 2018-02-08 Richard Biener <rguenther@suse.de>
9911 PR tree-optimization/84278
9912 * tree-vect-stmts.c (vectorizable_store): When looking for
9913 smaller vector types to perform grouped strided loads/stores
9914 make sure the mode is supported by the target.
9915 (vectorizable_load): Likewise.
9917 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9919 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9920 Increase LDP/STP opportunities by adding adjacent callee-saves.
9922 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9924 PR rtl-optimization/84068
9925 PR rtl-optimization/83459
9926 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9928 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9930 PR tree-optimization/84224
9931 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9932 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9935 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9938 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9939 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9940 count, remove one USE.
9942 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9944 * doc/install.texi (Configuration): Document the
9945 --with-long-double-format={ibm,ieee} PowerPC configuration
9949 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9950 Convert from define_expand to be define_insn_and_split. Rework
9951 float/double/_Float128 conversions to QI/HI/SImode to work with
9952 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9953 conversions to QI/HImode types did a store and then a load to
9954 truncate the value. For conversions to VSX registers, don't split
9955 the insn, instead emit the code directly. Use the code iterator
9956 any_fix to combine signed and unsigned conversions.
9957 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9958 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9959 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9960 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9961 (fix_<mode>di2_hw): Likewise.
9962 (fixuns_<mode>di2_hw): Likewise.
9963 (fix_<mode>si2_hw): Likewise.
9964 (fixuns_<mode>si2_hw): Likewise.
9965 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9966 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9967 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9968 fix<uns>_trunc<SFDF:mode>si2_p8.
9969 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9970 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9971 (fix<uns>_<mode>_mem): Likewise.
9972 (fctiw<u>z_<mode>_mem): Likewise.
9973 (fix<uns>_<mode>_mem): Likewise.
9974 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9975 the register allocator from doing a direct move to the GPRs to do
9976 a store, and instead use the ISA 3.0 store byte/half-word from
9977 vector register instruction. For IEEE 128-bit floating point,
9978 also optimize stores of 32-bit ints.
9979 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9981 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9983 * genextract.c (push_pathstr_operand): New function to support
9985 (walk_rtx): Call push_pathstr_operand.
9986 (print_path): Support [a-zA-Z].
9988 2018-02-07 Richard Biener <rguenther@suse.de>
9990 PR tree-optimization/84037
9991 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9992 (cse_and_gimplify_to_preheader): Declare.
9993 (vect_get_place_in_interleaving_chain): Likewise.
9994 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9996 (_loop_vec_info::~_loop_vec_info): Delete it.
9997 (cse_and_gimplify_to_preheader): New function.
9998 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9999 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10000 (vectorizable_load): Likewise. For grouped stores always base
10001 the IV on the first element.
10002 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10003 condition before gimplifying.
10005 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10007 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10008 *DIV_EXPR and *MOD_EXPR.
10010 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10013 * config/i386/i386.c (ix86_option_override_internal): Mask out
10014 the CF_SET bit when checking -fcf-protection.
10016 2018-02-07 Tom de Vries <tom@codesourcery.com>
10019 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10022 2018-02-07 Richard Biener <rguenther@suse.de>
10024 PR tree-optimization/84204
10025 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10028 PR tree-optimization/84205
10029 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10030 special-case isl_ast_op_zdiv_r.
10032 PR tree-optimization/84223
10033 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10034 Only add conditions from within the region.
10035 (gather_bbs::after_dom_children): Adjust.
10037 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10040 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10041 * config/avr/avr.md: Only post-reload split REG-REG moves if
10042 either register is GENERAL_REG_P.
10044 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10046 PR tree-optimization/84235
10047 * tree-ssa-scopedtables.c
10048 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10049 if the subtraction is performed in floating point type where NaNs are
10050 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10051 build 1. Formatting fix.
10053 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10056 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10057 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10058 and skip it regardless of bb boundaries. Use CALL_P macro,
10059 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10060 unnecessarily, formatting fix.
10062 2018-02-06 Michael Collison <michael.collison@arm.com>
10064 * config/arm/thumb2.md:
10065 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10066 (*thumb_mov_notscc): Ditto.
10068 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10071 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10072 unsigned_fix, not "s".
10074 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10076 * configure.ac (gcc_fn_eh_frame_ro): New function.
10077 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10078 correct .eh_frame permissions.
10079 * configure: Regenerate.
10081 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10083 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10084 irrelevant options.
10086 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10088 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10089 Display warning message for -mno-speculate-indirect-jumps.
10091 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10093 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10095 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10097 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10099 PR tree-optimization/84225
10100 * tree-eh.c (find_trapping_overflow): Only call
10101 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10103 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10106 * config/i386/i386.c: Reimplement the check of possible options
10107 -mibt/-mshstk conbination. Change error messages.
10108 * doc/invoke.texi: Fix a typo: remove extra '='.
10110 2018-02-06 Marek Polacek <polacek@redhat.com>
10112 PR tree-optimization/84228
10113 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10115 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10118 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10119 emitted arch directives.
10120 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10121 __ARM_FEATURE_COPROC before changing architectures.
10123 2018-02-06 Richard Biener <rguenther@suse.de>
10125 * config/i386/i386.c (print_reg): Fix typo.
10126 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10128 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10130 * configure: Regenerate.
10132 2018-02-05 Martin Sebor <msebor@redhat.com>
10134 PR tree-optimization/83369
10135 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10138 2018-02-05 Martin Liska <mliska@suse.cz>
10140 * doc/invoke.texi: Cherry-pick upstream r323995.
10142 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10144 * ira.c (ira_init_register_move_cost): Adjust comment.
10146 2018-02-05 Martin Liska <mliska@suse.cz>
10148 PR gcov-profile/84137
10149 * doc/gcov.texi: Fix typo in documentation.
10151 2018-02-05 Martin Liska <mliska@suse.cz>
10153 PR gcov-profile/83879
10154 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10155 using dlopen functionality.
10157 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10159 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10160 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10161 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10162 _mm_maskz_range_round_ss): New intrinsics.
10163 (__builtin_ia32_rangesd128_round)
10164 (__builtin_ia32_rangess128_round): Remove.
10165 (__builtin_ia32_rangesd128_mask_round,
10166 __builtin_ia32_rangess128_mask_round): New builtins.
10167 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10168 __builtin_ia32_rangess128_round): Remove.
10169 (__builtin_ia32_rangesd128_mask_round,
10170 __builtin_ia32_rangess128_mask_round): New builtins.
10171 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10172 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10173 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10174 "<round_saeonly_constraint>")): Changed to ...
10175 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10176 "<round_saeonly_scalar_constraint>")): ... this.
10177 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10178 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10179 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10180 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10181 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10183 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10185 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10187 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10188 Remove all values except native, 8540 and 8548.
10190 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10192 * config/i386/i386.c (ix86_output_function_return): Pass
10193 INVALID_REGNUM, instead of -1, as invalid register number to
10194 indirect_thunk_name and output_indirect_thunk.
10196 2018-02-02 Julia Koval <julia.koval@intel.com>
10198 * config.gcc: Add -march=icelake.
10199 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10200 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10201 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10202 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10203 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10204 (processor_target_table): Add icelake.
10205 (ix86_option_override_internal): Handle new PTAs.
10206 (get_builtin_code_for_version): Handle icelake.
10207 (M_INTEL_COREI7_ICELAKE): New.
10208 (fold_builtin_cpu): Handle icelake.
10209 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10210 * doc/invoke.texi: Add -march=icelake.
10212 2018-02-02 Julia Koval <julia.koval@intel.com>
10214 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10215 to wide_int_bitmask.
10216 * wide-int-bitmask.h: New.
10218 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10221 * config/i386/i386.md: Replace Pmode with word_mode in
10222 builtin_setjmp_setup and builtin_longjmp to support x32.
10224 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10228 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10230 (rs6000_supported_cpu_names): New static variable.
10231 (linux_cpu_translation_table): Likewise.
10232 (elf_platform) <cpu>: Define new static variable and use it.
10233 Translate kernel AT_PLATFORM name to canonical name if needed.
10234 Error if platform name is unknown.
10236 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10239 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10241 2018-02-01 Jeff Law <law@redhat.com>
10244 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10245 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10246 the scratch if RELEASE_VIA_POP is false.
10247 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10248 If we have to save a temporary register, decrement SIZE appropriately.
10249 Pass new arguments to release_scratch_register_on_entry.
10250 (ix86_adjust_stack_and_probe): Likewise.
10251 (ix86_emit_probe_stack_range): Pass new arguments to
10252 release_scratch_register_on_entry.
10254 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10256 PR rtl-optimization/84157
10257 * combine.c (change_zero_ext): Use REG_P predicate in
10258 front of HARD_REGISTER_P predicate.
10260 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10262 * config/avr/avr.c (avr_option_override): Move disabling of
10263 -fdelete-null-pointer-checks to...
10264 * common/config/avr/avr-common.c (avr_option_optimization_table):
10267 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10269 PR tree-optimization/81635
10270 * tree-data-ref.c (split_constant_offset_1): For types that
10271 wrap on overflow, try to use range info to prove that wrapping
10274 2018-02-01 Renlin Li <renlin.li@arm.com>
10277 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10278 TAILCALL_ADDR_REGS.
10279 (aarch64_register_move_cost): Likewise.
10280 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10281 TAILCALL_ADDR_REGS.
10282 (REG_CLASS_NAMES): Likewise.
10283 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10284 TAILCALL_ADDR_REGS. Remove IP registers.
10285 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10287 2018-02-01 Richard Biener <rguenther@suse.de>
10289 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10290 for specifying RPO order and allow NULL for that.
10291 * domwalk.c (dom_walker::dom_walker): Likewise.
10292 (dom_walker::walk): Handle NULL RPO order.
10293 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10295 (rewrite_update_dom_walker): Likewise.
10296 (mark_def_dom_walker): Likewise.
10298 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10300 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10301 (aarch64_maybe_expand_sve_subreg_move): Declare.
10302 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10303 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10305 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10306 that are semantically a reverse operation.
10307 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10308 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10309 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10311 (aarch64_can_change_mode_class): For big-endian, forbid changes
10312 between two SVE modes if they have different element sizes.
10314 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10316 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10317 the TImode handling for big-endian targets.
10319 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10321 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10322 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10324 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10325 Remove BSWAP handing for big-endian targets and use the form of
10326 LD1RQ appropariate for the mode.
10328 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10330 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10331 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10332 duplicated element.
10334 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10337 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10338 check for operands that need to go through aarch64_sve_reload_be.
10340 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10342 PR tree-optimization/81661
10343 PR tree-optimization/84117
10344 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10345 * tree-eh.c: Include gimplify.h.
10346 (find_trapping_overflow, replace_trapping_overflow,
10347 rewrite_to_non_trapping_overflow): New functions.
10348 * tree-vect-loop.c: Include tree-eh.h.
10349 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10350 * tree-data-ref.c: Include tree-eh.h.
10351 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10353 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10355 PR rtl-optimization/84123
10356 * combine.c (change_zero_ext): Check if hard register satisfies
10357 can_change_dest_mode before calling gen_lowpart_SUBREG.
10359 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10362 * ira.c (ira_init_register_move_cost): Remove assert.
10364 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10366 PR rtl-optimization/84071
10367 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10368 * doc/tm.texi: Regenerate.
10370 2018-01-31 Richard Biener <rguenther@suse.de>
10372 PR tree-optimization/84132
10373 * tree-data-ref.c (analyze_miv_subscript): Properly
10374 check whether evolution_function_is_affine_multivariate_p
10375 before calling gcd_of_steps_may_divide_p.
10377 2018-01-31 Julia Koval <julia.koval@intel.com>
10380 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10381 * config/i386/i386.md (rdpid_rex64) New.
10382 (rdpid): Make 32bit only.
10384 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10387 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10388 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10390 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10393 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10395 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10397 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10399 PR rtl-optimization/84071
10400 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10401 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10403 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10405 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10406 (arc_attribute_table): Add 'aux' attribute.
10407 (arc_in_small_data_p): Consider aux like variables.
10408 (arc_is_aux_reg_p): New function.
10409 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10410 (arc_get_aux_arg): New function.
10411 (prepare_move_operands): Handle aux-register access.
10412 (arc_handle_aux_attribute): New function.
10413 * doc/extend.texi (ARC Variable attributes): Add subsection.
10415 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10417 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10418 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10419 (arc_attribute_table): Add 'uncached' attribute.
10420 (arc_print_operand): Print '.di' flag for uncached memory
10422 (arc_in_small_data_p): Do not consider for small data the uncached
10424 (arc_is_uncached_mem_p): New function.
10425 * config/arc/predicates.md (compact_store_memory_operand): Check
10426 for uncached memory accesses.
10427 (nonvol_nonimm_operand): Likewise.
10428 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10430 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10433 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10434 falign-loops=): Add Optimization flag.
10436 2018-01-30 Jeff Law <law@redhat.com>
10439 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10440 INT_REGISTERS_SAVED. Check it prior to calling
10441 get_scratch_register_on_entry.
10442 (ix86_adjust_stack_and_probe): Similarly.
10443 (ix86_emit_probe_stack_range): Similarly.
10444 (ix86_expand_prologue): Corresponding changes.
10446 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10449 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10450 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10452 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10455 * lra-constraints.c (curr_insn_transform): Process AND in the
10458 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10460 PR rtl-optimization/83986
10461 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10462 dependence against last_pending_memory_flush in addition to
10463 pending_jump_insns.
10465 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10467 PR tree-optimization/81611
10468 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10471 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10474 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10477 2018-01-30 Richard Biener <rguenther@suse.de>
10478 Jakub Jelinek <jakub@redhat.com>
10480 PR tree-optimization/84111
10481 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10482 inner loops added during recursion, as they don't have up-to-date
10485 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10488 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10489 (can_inline_edge_by_limits_p): ... here.
10490 (can_early_inline_edge_p, check_callers,
10491 update_caller_keys, update_callee_keys, recursive_inlining,
10492 add_new_edges_to_heap, speculation_useful_p,
10493 inline_small_functions,
10494 inline_small_functions, flatten_function,
10495 inline_to_all_callers_1): Update.
10497 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10499 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10502 2018-01-30 Richard Biener <rguenther@suse.de>
10504 PR tree-optimization/83008
10505 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10506 invariant and constant vector uses in stmts when they need
10507 more than one stmt.
10509 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10512 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10513 * configure: Regenerate.
10515 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10517 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10519 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10520 Use gen_rtx_REG rather than gen_lowpart.
10522 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10524 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10525 rather than 0 when creating partial subregs.
10527 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10529 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10532 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10535 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10536 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10537 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10538 flags. This restores the settings used before the 2017-07-24.
10539 Turning off pre increment/decrement/modify allows IVOPTS to
10540 optimize DF/SF loops where the index is an int.
10542 2018-01-29 Richard Biener <rguenther@suse.de>
10543 Kelvin Nilsen <kelvin@gcc.gnu.org>
10546 * tree-vect-stmts.c (vectorizable_call): Don't call
10547 targetm.vectorize_builtin_md_vectorized_function if callee is
10550 2018-01-22 Carl Love <cel@us.ibm.com>
10552 * doc/extend.tex: Fix typo in second arg in
10553 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10555 2018-01-29 Richard Biener <rguenther@suse.de>
10557 PR tree-optimization/84086
10558 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10559 (flush_ssaname_freelist): When SSA names were released reset
10560 the SCEV hash table.
10562 2018-01-29 Richard Biener <rguenther@suse.de>
10564 PR tree-optimization/84057
10565 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10566 removed paths when removing edges.
10568 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10570 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10571 -mfunction-return=@var{choice}.
10573 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10575 PR diagnostic/84034
10576 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10577 Handle CR like TAB.
10578 (layout::print_source_line): Likewise.
10579 (test_get_line_width_without_trailing_whitespace): Add test cases.
10581 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10583 PR middle-end/84040
10584 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10587 2018-01-26 Jim Wilson <jimw@sifive.com>
10589 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10591 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10594 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10596 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10597 and CMP + SUB-immediate -> SUBS.
10599 2018-01-26 Martin Sebor <msebor@redhat.com>
10601 PR tree-optimization/83896
10602 * tree-ssa-strlen.c (get_string_len): Rename...
10603 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10604 Avoid assuming length is constant.
10605 (handle_char_store): Use HOST_WIDE_INT for string length.
10607 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10610 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10611 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10613 2018-01-26 Richard Biener <rguenther@suse.de>
10615 PR rtl-optimization/84003
10616 * dse.c (record_store): Only record redundant stores when
10617 the earlier store aliases at least all accesses the later one does.
10619 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10621 PR rtl-optimization/83985
10622 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10623 REG_CFA_RESTORE insns.
10624 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10625 REG_CFA_RESTORE insns here.
10628 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10629 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10631 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10633 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10634 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10635 (arc_init): Likewise.
10636 (arc_override_options): Likewise.
10637 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10639 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10641 * config/arc/arc.h (TARGET_DBNZ): Define.
10642 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10643 properly set the tune attribute.
10644 (dbnz): Use TARGET_DBNZ guard.
10645 * config/arc/arc.opt (mtune): Add core3 option.
10647 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10649 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10650 recognize new pic like addresses.
10651 (arc_delegitimize_address): Clean up.
10653 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10655 * config/arc/arc-arches.def: Option mrf16 valid for all
10657 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10658 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10659 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10660 * config/arc/arc-tables.opt: Regenerate.
10661 * config/arc/arc.c (arc_conditional_register_usage): Handle
10662 reduced register file case.
10663 (arc_file_start): Set must have build attributes.
10664 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10665 mrf16 option value.
10666 * config/arc/arc.opt (mrf16): Add new option.
10667 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10668 * config/arc/genmultilib.awk: Handle new mrf16 option.
10669 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10670 * config/arc/t-multilib: Regenerate.
10671 * doc/invoke.texi (ARC Options): Document mrf16 option.
10673 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10675 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10676 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10677 (arc_attribute_table): Add 'secure_call' attribute.
10678 (arc_print_operand): Print secure call operand.
10679 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10681 (arc_is_secure_call_p): New function. * config/arc/arc.md
10682 (call_i): Add support for sjli instruction.
10683 (call_value_i): Likewise.
10684 * config/arc/constraints.md (Csc): New constraint.
10686 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10687 John Eric Martin <John.Martin@emmicro-us.com>
10689 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10690 * config/arc/arc.c (_arc_jli_section): New struct.
10691 (arc_jli_section): New type.
10692 (rc_jli_sections): New static variable.
10693 (arc_handle_jli_attribute): New function.
10694 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10695 (arc_file_end): New function.
10696 (TARGET_ASM_FILE_END): Define.
10697 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10698 (arc_add_jli_section): New function.
10699 (jli_call_scan): Likewise.
10700 (arc_reorg): Call jli_call_scan.
10701 (arc_output_addsi): Remove 'S' from printing asm operand.
10702 (arc_is_jli_call_p): New function.
10703 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10705 (movhi_insn): Likewise.
10706 (movsi_insn): Likewise.
10707 (movsi_set_cc_insn): Likewise.
10708 (loadqi_update): Likewise.
10709 (load_zeroextendqisi_update): Likewise.
10710 (load_signextendqisi_update): Likewise.
10711 (loadhi_update): Likewise.
10712 (load_zeroextendhisi_update): Likewise.
10713 (load_signextendhisi_update): Likewise.
10714 (loadsi_update): Likewise.
10715 (loadsf_update): Likewise.
10716 (movsicc_insn): Likewise.
10717 (bset_insn): Likewise.
10718 (bxor_insn): Likewise.
10719 (bclr_insn): Likewise.
10720 (bmsk_insn): Likewise.
10721 (bicsi3_insn): Likewise.
10722 (cmpsi_cc_c_insn): Likewise.
10723 (movsi_ne): Likewise.
10724 (movsi_cond_exec): Likewise.
10725 (clrsbsi2): Likewise.
10726 (norm_f): Likewise.
10736 (call_i): Remove 'S' asm letter, add jli instruction.
10737 (call_value_i): Likewise.
10738 * config/arc/arc.op (mjli-always): New option.
10739 * config/arc/constraints.md (Cji): New constraint.
10740 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10742 (subsf3_fpx): Likewise.
10743 (mulsf3_fpx): Likewise.
10744 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10746 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10747 function attrbutes.
10748 * doc/invoke.texi (ARC): Document mjli-always option.
10750 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10752 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10753 avoid addition with 0 and use incw and decw where possible.
10755 2018-01-26 Richard Biener <rguenther@suse.de>
10757 PR tree-optimization/81082
10758 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10759 association if it requires casting to unsigned.
10760 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10761 from fold_plusminus_mult_expr to catch important cases late when
10762 range info is available.
10764 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10766 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10767 * configure.ac (hidden_linkonce): New test.
10768 * configure: Regenerate.
10769 * config.in: Regenerate.
10771 2018-01-26 Julia Koval <julia.koval@intel.com>
10773 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10774 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10775 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10776 _mm_mask_bitshuffle_epi64_mask): Fix type.
10777 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10778 USI_FTYPE_V4DI_V4DI_USI): Remove.
10779 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10780 __builtin_ia32_vpshufbitqmb256_mask,
10781 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10782 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10783 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10785 2018-01-26 Alan Modra <amodra@gmail.com>
10788 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10789 UNSPEC_VBPERMQ. Sort other unspecs.
10791 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10793 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10795 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10797 PR middle-end/83055
10798 * predict.c (drop_profile): Do not push/pop cfun; update also
10800 (handle_missing_profiles): Fix logic looking for zero profiles.
10802 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10804 PR middle-end/83977
10805 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10806 on functions with #pragma omp declare simd or functions with simd
10808 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10809 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10810 Remove trailing \n from warning_at calls.
10812 2018-01-25 Tom de Vries <tom@codesourcery.com>
10815 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10816 for neutered workers.
10818 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10821 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10822 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10824 2018-01-24 Jeff Law <law@redhat.com>
10827 * i386.c (get_probe_interval): Move to earlier point.
10828 (ix86_compute_frame_layout): If -fstack-clash-protection and
10829 the frame is larger than the probe interval, then use pushes
10830 to save registers rather than reg->mem moves.
10831 (ix86_expand_prologue): Remove conditional for int_registers_saved
10834 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10837 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10838 min/max for never referenced object.
10840 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10842 PR middle-end/83977
10843 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10845 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10846 attributes from DECL_ATTRIBUTES (decl) without affecting
10847 DECL_ATTRIBUTES (current_function_decl).
10848 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10849 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10851 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10853 PR tree-optimization/83979
10854 * fold-const.c (fold_comparison): Use constant_boolean_node
10855 instead of boolean_{true,false}_node.
10857 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10859 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10862 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10864 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10865 Simplify the clause that sets the length attribute.
10866 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10867 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10868 clause that sets the length attribute.
10869 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10871 2018-01-24 Tom de Vries <tom@codesourcery.com>
10874 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10875 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10876 Add strict parameter.
10877 (prevent_branch_around_nothing): Insert dummy insn between branch to
10878 label and label with no ptx insn inbetween.
10879 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10881 2018-01-24 Tom de Vries <tom@codesourcery.com>
10884 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10885 for neutered threads in warp.
10886 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10888 2018-01-24 Richard Biener <rguenther@suse.de>
10890 PR tree-optimization/83176
10891 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10894 2018-01-24 Richard Biener <rguenther@suse.de>
10896 PR tree-optimization/82819
10897 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10898 code generating pluses that are no-ops in the target precision.
10900 2018-01-24 Richard Biener <rguenther@suse.de>
10902 PR middle-end/84000
10903 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10905 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10907 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10908 to merge probabilities.
10909 * predict.c (probably_never_executed): Also mark as cold functions
10910 with global 0 profile and guessed local profile.
10911 * profile-count.c (profile_probability::combine_with_count): New
10913 * profile-count.h (profile_probability::operator*,
10914 profile_probability::operator*=, profile_probability::operator/,
10915 profile_probability::operator/=): Reduce precision to adjusted
10916 and set value to guessed on contradictory divisions.
10917 (profile_probability::combine_with_freq): Remove.
10918 (profile_probability::combine_wiht_count): Declare.
10919 (profile_count::force_nonzero):: Set to adjusted.
10920 (profile_count::probability_in):: Set quality to adjusted.
10921 * tree-ssa-tail-merge.c (replace_block_by): Use
10922 combine_with_count.
10924 2018-01-23 Andrew Waterman <andrew@sifive.com>
10925 Jim Wilson <jimw@sifive.com>
10927 * config/riscv/riscv.c (riscv_stack_boundary): New.
10928 (riscv_option_override): Set riscv_stack_boundary. Handle
10929 riscv_preferred_stack_boundary_arg.
10930 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10931 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10932 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10933 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10934 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10935 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10937 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10940 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10941 of struct ix86_frame.
10942 (ix86_expand_epilogue): Likewise. Add a local variable for
10943 the reg_save_offset field in struct ix86_frame.
10945 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10947 PR tree-optimization/82604
10948 * tree-loop-distribution.c (enum partition_kind): New enum item
10949 PKIND_PARTIAL_MEMSET.
10950 (partition_builtin_p): Support above new enum item.
10951 (generate_code_for_partition): Ditto.
10952 (compute_access_range): Differentiate cases that equality can be
10953 proven at all loops, the innermost loops or no loops.
10954 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10955 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10956 (finalize_partitions, distribute_loop): Don't fuse partition of
10957 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10958 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10959 parloop is enabled.
10961 2018-01-23 Martin Liska <mliska@suse.cz>
10963 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10964 order to ignore the predictor.
10965 (PRED_POLYMORPHIC_CALL): Likewise.
10966 (PRED_RECURSIVE_CALL): Likewise.
10968 2018-01-23 Martin Liska <mliska@suse.cz>
10970 * tree-profile.c (tree_profiling): Print function header to
10971 aware reader which function we are working on.
10972 * value-prof.c (gimple_find_values_to_profile): Do not print
10973 not interesting value histograms.
10975 2018-01-23 Martin Liska <mliska@suse.cz>
10977 * profile-count.h (enum profile_quality): Add
10978 profile_uninitialized as the first value. Do not number values
10979 as they are zero based.
10980 (profile_count::verify): Update sanity check.
10981 (profile_probability::verify): Likewise.
10983 2018-01-23 Nathan Sidwell <nathan@acm.org>
10985 * doc/invoke.texi (ffor-scope): Deprecate.
10987 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10989 PR tree-optimization/83510
10990 * domwalk.c (set_all_edges_as_executable): New function.
10991 (dom_walker::dom_walker): Convert bool param
10992 "skip_unreachable_blocks" to enum reachability. Move setup of
10993 edge flags to set_all_edges_as_executable and only do it when
10994 reachability is REACHABLE_BLOCKS.
10995 * domwalk.h (enum dom_walker::reachability): New enum.
10996 (dom_walker::dom_walker): Convert bool param
10997 "skip_unreachable_blocks" to enum reachability.
10998 (set_all_edges_as_executable): New decl.
10999 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11000 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11002 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11003 but converting true to REACHABLE_BLOCKS.
11004 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11006 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11007 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11008 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11010 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11011 if check_all_array_refs will be called.
11013 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11015 * tree.c (selftest::test_location_wrappers): Add more test
11018 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11020 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11021 (selftest::test_bit_in_range): Likewise.
11023 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11026 * doc/sourcebuild.texi (vect_float): Say that the selector
11027 only describes the situation when -funsafe-math-optimizations is on.
11028 (vect_float_strict): Document.
11030 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11032 PR tree-optimization/83965
11033 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11034 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11035 instead of checking only for a reduction.
11036 (vect_recog_widen_sum_pattern): Likewise.
11038 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11040 * predict.c (probably_never_executed): Only use precise profile info.
11041 (compute_function_frequency): Skip after inlining hack since we now
11042 have quality checking.
11044 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11046 * profile-count.h (profile_probability::very_unlikely,
11047 profile_probability::unlikely, profile_probability::even): Set
11048 precision to guessed.
11050 2018-01-23 Richard Biener <rguenther@suse.de>
11052 PR tree-optimization/83963
11053 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11054 Properly terminate dominator walk when crossing the exit edge not
11055 when visiting its source block.
11057 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11060 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11061 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11063 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11065 PR tree-optimization/83957
11066 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11067 semicolon after for body surrounded by braces.
11069 PR tree-optimization/83081
11070 * profile-count.h (profile_probability::split): New method.
11071 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11072 Use profile_probability::split.
11073 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11074 when splitting a single conditional jump into 2.
11076 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11078 PR tree-optimization/69452
11079 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11082 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11084 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11085 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11086 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11088 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11090 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11092 * config/rl78/rl78.md (movdi): New define_expand.
11093 * config/rl78/rl78.c (rl78_split_movdi): New function.
11095 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11098 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11100 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11101 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11102 128-bit to produce an UNSPEC move to get the double word with the
11103 signbit and then a shift directly to do signbit.
11104 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11105 implementation with a new version that just does either a direct
11106 move or a regular move. Move memory interface to separate insns.
11107 Move insns so they are next to the expander.
11108 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11109 with signbit move. Split big and little endian case.
11110 (signbit<mode>2_dm_mem_le): Likewise.
11111 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11112 (signbit<mode>2_dm2): Likewise.
11114 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11116 * config/rl78/rl78.md (anddi3): New define_expand.
11118 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11120 * config/rl78/rl78.md (umindi3): New define_expand.
11122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11124 * config/rl78/rl78.md (smindi3): New define_expand.
11126 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11128 * config/rl78/rl78.md (smaxdi3): New define_expand.
11130 2018-01-22 Carl Love <cel@us.ibm.com>
11132 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11133 LVX_V1TI): Add macro expansion.
11134 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11135 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11136 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11137 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11138 Change check to determine if the instruction is a byte reversing
11139 entry. Fix typo in comment.
11140 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11141 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11142 Add def_builtin calls for new builtins.
11143 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11144 Add define_insn expansion.
11146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11148 * config/rl78/rl78.md (umaxdi3): New define_expand.
11150 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11152 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11153 for non-QImode registers.
11155 2018-01-22 Richard Biener <rguenther@suse.de>
11157 PR tree-optimization/83963
11158 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11159 including the loop exit block.
11160 (scop_detection::merge_sese): Likewise.
11161 (scop_detection::add_scop): Do it here instead.
11163 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11165 * doc/sourcebuild.texi (arm_softfloat): Document.
11167 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11170 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11171 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11172 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11174 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11175 David Edelsohn <dje.gcc@gmail.com>
11178 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11179 Change "crset eq" to "crset 2".
11180 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11181 (*call_indirect_aix<mode>_nospec): Likewise.
11182 (*call_value_indirect_aix<mode>_nospec): Likewise.
11183 (*call_indirect_elfv2<mode>_nospec): Likewise.
11184 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11185 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11186 change assembly output from . to $.
11187 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11188 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11189 (*tablejump<mode>_internal1_nospec): Likewise.
11191 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11194 * config/sh/sh_optimize_sett_clrt.cc:
11195 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11197 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11199 PR tree-optimization/83940
11200 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11201 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11202 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11203 use it to pass back the definition type.
11204 (vect_check_store_rhs): Likewise rhs_dt_out.
11205 (vect_build_gather_load_calls): Add a mask_dt argument and use
11206 it instead of a call to vect_is_simple_use.
11207 (vectorizable_store): Update calls to vect_check_load_store_mask
11208 and vect_check_store_rhs. Use the dt returned by the latter instead
11209 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11210 instead of calls to vect_is_simple_use. Pass the scalar rather
11211 than the vector operand to vect_is_simple_use when handling
11212 second and subsequent copies of an rhs value.
11213 (vectorizable_load): Update calls to vect_check_load_store_mask
11214 and vect_build_gather_load_calls. Use the cached mask_dt and
11215 gs_info.offset_dt instead of calls to vect_is_simple_use.
11217 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11219 PR middle-end/83945
11220 * tree-emutls.c: Include gimplify.h.
11221 (lower_emutls_2): New function.
11222 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11223 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11224 it before further processing.
11227 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11228 UINTVAL (trueop1) instead of INTVAL (op1).
11230 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11234 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11235 INCOMING_FRAME_SP_OFFSET if not defined.
11236 (scan_trace): Add ENTRY argument. If true and
11237 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11238 emit a note to adjust the CFA offset.
11239 (create_cfi_notes): Adjust scan_trace callers.
11240 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11241 INCOMING_FRAME_SP_OFFSET in the CIE.
11242 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11243 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11245 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11246 * doc/tm.texi: Regenerated.
11248 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11250 PR rtl-optimization/83147
11251 * lra-constraints.c (remove_inheritance_pseudos): Use
11252 lra_substitute_pseudo_within_insn.
11254 2018-01-19 Tom de Vries <tom@codesourcery.com>
11255 Cesar Philippidis <cesar@codesourcery.com>
11258 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11260 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11263 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11264 spaces for function labels.
11266 2018-01-19 Martin Liska <mliska@suse.cz>
11268 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11269 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11270 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11271 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11272 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11273 (PRED_CONST_RETURN): Change from 69 to 65.
11274 (PRED_NULL_RETURN): Change from 91 to 71.
11275 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11276 (PRED_LOOP_GUARD): Change from 66 to 73.
11278 2018-01-19 Martin Liska <mliska@suse.cz>
11280 * predict.c (predict_insn_def): Add new assert.
11281 (struct branch_predictor): Change type to signed integer.
11282 (test_prediction_value_range): Amend test to cover
11283 PROB_UNINITIALIZED.
11284 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11285 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11286 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11287 (PRED_LOOP_IV_COMPARE): Likewise.
11288 * predict.h (PROB_UNINITIALIZED): Define new constant.
11290 2018-01-19 Martin Liska <mliska@suse.cz>
11292 * predict.c (dump_prediction): Add new format for
11293 analyze_brprob.py script which is enabled with -details
11295 * profile-count.h (precise_p): New function.
11297 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11299 PR tree-optimization/83922
11300 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11301 there are no statements that need masking.
11302 (vect_active_double_reduction_p): New function.
11303 (vect_analyze_loop_operations): Use it when handling phis that
11304 are not in the loop header.
11306 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11308 PR tree-optimization/83914
11309 * tree-vect-loop.c (vectorizable_induction): Don't convert
11310 init_expr or apply the peeling adjustment for inductions
11311 that are nested within the vectorized loop.
11313 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11315 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11318 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11322 * function.h (gimplify_parameters): Add gimple_seq * argument.
11323 * function.c: Include gimple.h and options.h.
11324 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11325 for the added local temporaries if needed.
11326 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11327 if there are any parameter cleanups, wrap whole body into a
11328 try/finally with the cleanups.
11330 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11333 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11334 Use GET_MODE_CLASS for scalar floating point.
11336 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11340 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11341 Fix call of call_cgraph_insertion_hooks.
11343 2018-01-18 Martin Sebor <msebor@redhat.com>
11345 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11347 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11350 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11353 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11356 * common.opt: (-ffile-prefix-map): New option.
11357 * opts.c (common_handle_option): Defer it.
11358 * opts-global.c (handle_common_deferred_options): Handle it.
11359 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11360 * file-prefix-map.h: New file.
11361 (remap_debug_filename, add_debug_prefix_map): ...here.
11362 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11363 * final.c (debug_prefix_map, add_debug_prefix_map
11364 remap_debug_filename): Move to...
11365 * file-prefix-map.c: New file.
11366 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11367 generalize, get rid of alloca(), use strrchr() instead of strchr().
11368 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11369 Implement in terms of add_prefix_map().
11370 (remap_macro_filename, remap_debug_filename): Implement in term of
11372 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11373 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11374 * dbxout.c: Include file-prefix-map.h.
11375 * varasm.c: Likewise.
11376 * vmsdbgout.c: Likewise.
11377 * xcoffout.c: Likewise.
11378 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11379 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11380 * doc/invoke.texi (-ffile-prefix-map): Document.
11381 (-fdebug-prefix-map): Update description.
11383 2018-01-18 Martin Liska <mliska@suse.cz>
11385 * config/i386/i386.c (indirect_thunk_name): Document that also
11387 (output_indirect_thunk): Document why both instructions
11388 (pause and lfence) are generated.
11390 2018-01-18 Richard Biener <rguenther@suse.de>
11392 PR tree-optimization/83887
11393 * graphite-scop-detection.c
11394 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11395 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11396 (scop_detection::merge_sese): Re-implement with a flood-fill
11397 algorithm that properly finds a SESE region if it exists.
11399 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11402 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11403 pointer_diff optimizations use view_convert instead of convert.
11405 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11407 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11408 Generate different code for -mno-speculate-indirect-jumps.
11409 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11410 (*call_indirect_aix<mode>): Disable for
11411 -mno-speculate-indirect-jumps.
11412 (*call_indirect_aix<mode>_nospec): New define_insn.
11413 (*call_value_indirect_aix<mode>): Disable for
11414 -mno-speculate-indirect-jumps.
11415 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11416 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11417 -mno-speculate-indirect-jumps.
11418 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11420 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11422 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11423 long double type, set the flags for noting the default long double
11424 type, even if we don't pass or return a long double type.
11426 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11429 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11432 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11434 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11435 support for merge[hl].
11436 (fold_mergehl_helper): New helper function.
11437 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11438 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11439 (altivec_vmrglw_direct): Add xxmrglw insn.
11441 2018-01-17 Andrew Waterman <andrew@sifive.com>
11443 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11444 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11446 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11449 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11450 call the lto_location_cache before reading the
11451 DECL_SOURCE_LOCATION of the types.
11453 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11454 Richard Sandiford <richard.sandiford@linaro.org>
11456 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11457 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11458 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11459 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11460 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11462 * config/aarch64/constraints.md (aarch64_movti_operand):
11464 * config/aarch64/predicates.md (Uti): Add new constraint.
11466 2018-01-17 Carl Love <cel@us.ibm.com>
11468 * config/rs6000/vsx.md (define_expand xl_len_r,
11469 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11470 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11472 (define_expand, define_insn): Move the shift left from the
11473 define_insn to the define_expand for lxvl and stxvl instructions.
11474 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11475 and XL_LEN_R definitions to PURE.
11477 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11479 * config/i386/i386.c (indirect_thunk_name): Declare regno
11480 as unsigned int. Compare regno with INVALID_REGNUM.
11481 (output_indirect_thunk): Ditto.
11482 (output_indirect_thunk_function): Ditto.
11483 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11484 in the call to output_indirect_thunk_function.
11486 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11488 PR middle-end/83884
11489 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11490 rather than the size of inner_type to determine the stack slot size
11491 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11493 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11496 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11499 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11501 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11502 endian Linux systems to optionally enable multilibs for selecting
11503 the long double type if the user configured an explicit type.
11504 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11505 have no long double multilibs if not defined.
11506 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11507 warn if the user used -mabi={ieee,ibm}longdouble and we built
11508 multilibs for long double.
11509 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11510 appropriate multilib option.
11511 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11513 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11514 for building long double multilibs.
11515 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11517 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11519 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11522 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11524 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11527 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11530 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11533 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11535 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11536 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11538 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11540 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11541 different rtl trees depending on TARGET_64BIT.
11542 (rs6000_gen_lvx): Likewise.
11544 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11546 * config/visium/visium.md (nop): Tweak comment.
11547 (hazard_nop): Likewise.
11549 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11551 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11552 -mspeculate-indirect-jumps.
11553 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11554 for -mno-speculate-indirect-jumps.
11555 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11556 (*call_value_indirect_elfv2<mode>): Disable for
11557 -mno-speculate-indirect-jumps.
11558 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11559 (indirect_jump): Emit different RTL for
11560 -mno-speculate-indirect-jumps.
11561 (*indirect_jump<mode>): Disable for
11562 -mno-speculate-indirect-jumps.
11563 (*indirect_jump<mode>_nospec): New define_insn.
11564 (tablejump): Emit different RTL for
11565 -mno-speculate-indirect-jumps.
11566 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11567 (tablejumpsi_nospec): New define_expand.
11568 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11569 (tablejumpdi_nospec): New define_expand.
11570 (*tablejump<mode>_internal1): Disable for
11571 -mno-speculate-indirect-jumps.
11572 (*tablejump<mode>_internal1_nospec): New define_insn.
11573 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11576 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11578 * caller-save.c (insert_save): Drop unnecessary parameter. All
11581 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11582 Richard Biener <rguenth@suse.de>
11585 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11586 return early, inline manually is_gimple_sizepos. Make sure if we
11587 call gimplify_expr we don't end up with a gimple constant.
11588 * tree.c (variably_modified_type_p): Don't return true for
11589 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11590 * gimplify.h (is_gimple_sizepos): Remove.
11592 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11594 PR tree-optimization/83857
11595 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11596 vectorizable_live_operation for pure SLP statements.
11597 (vectorizable_live_operation): Handle PHIs.
11599 2018-01-16 Richard Biener <rguenther@suse.de>
11601 PR tree-optimization/83867
11602 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11603 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11605 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11608 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11609 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11610 If off is not INTEGER_CST, issue a may not be aligned warning
11611 rather than isn't aligned. Use isn%'t rather than isn't.
11612 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11614 <case MULT_EXPR>: Improve the case when bottom and one of the
11615 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11616 operand, in that case check if the other operand is multiple of
11617 bottom divided by the INTEGER_CST operand.
11619 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11622 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11623 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11624 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11625 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11626 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11627 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11628 (pa_function_arg_size): New function.
11630 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11632 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11633 in a separate statement.
11635 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11637 PR tree-optimization/83847
11638 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11639 group gathers and scatters.
11641 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11643 PR rtl-optimization/86620
11644 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11646 PR rtl-optimization/83213
11647 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11648 to last if both are JUMP_INSNs.
11650 PR tree-optimization/83843
11651 * gimple-ssa-store-merging.c
11652 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11653 store_immediate_info for bswap/nop orig_stores.
11655 2018-01-15 Andrew Waterman <andrew@sifive.com>
11657 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11659 <UDIV>: Increase cost if !TARGET_DIV.
11661 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11663 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11664 (define_attr "cr_logical_3op"): New.
11665 (cceq_ior_compare): Adjust.
11666 (cceq_ior_compare_complement): Adjust.
11667 (*cceq_rev_compare): Adjust.
11668 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11669 (is_cracked_insn): Adjust.
11670 (insn_must_be_first_in_group): Adjust.
11671 * config/rs6000/40x.md: Adjust.
11672 * config/rs6000/440.md: Adjust.
11673 * config/rs6000/476.md: Adjust.
11674 * config/rs6000/601.md: Adjust.
11675 * config/rs6000/603.md: Adjust.
11676 * config/rs6000/6xx.md: Adjust.
11677 * config/rs6000/7450.md: Adjust.
11678 * config/rs6000/7xx.md: Adjust.
11679 * config/rs6000/8540.md: Adjust.
11680 * config/rs6000/cell.md: Adjust.
11681 * config/rs6000/e300c2c3.md: Adjust.
11682 * config/rs6000/e500mc.md: Adjust.
11683 * config/rs6000/e500mc64.md: Adjust.
11684 * config/rs6000/e5500.md: Adjust.
11685 * config/rs6000/e6500.md: Adjust.
11686 * config/rs6000/mpc.md: Adjust.
11687 * config/rs6000/power4.md: Adjust.
11688 * config/rs6000/power5.md: Adjust.
11689 * config/rs6000/power6.md: Adjust.
11690 * config/rs6000/power7.md: Adjust.
11691 * config/rs6000/power8.md: Adjust.
11692 * config/rs6000/power9.md: Adjust.
11693 * config/rs6000/rs64.md: Adjust.
11694 * config/rs6000/titan.md: Adjust.
11696 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11698 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11699 ix86_indirect_branch_register logic.
11701 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11703 * config/i386/constraints.md (Bs): Update
11704 ix86_indirect_branch_register check. Don't check
11705 ix86_indirect_branch_register with GOT_memory_operand.
11707 * config/i386/predicates.md (GOT_memory_operand): Don't check
11708 ix86_indirect_branch_register here.
11709 (GOT32_symbol_operand): Likewise.
11711 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11713 * config/i386/predicates.md (constant_call_address_operand):
11714 Rewrite ix86_indirect_branch_register logic.
11715 (sibcall_insn_operand): Likewise.
11717 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11719 * config/i386/constraints.md (Bs): Replace
11720 ix86_indirect_branch_thunk_register with
11721 ix86_indirect_branch_register.
11723 * config/i386/i386.md (indirect_jump): Likewise.
11724 (tablejump): Likewise.
11725 (*sibcall_memory): Likewise.
11726 (*sibcall_value_memory): Likewise.
11727 Peepholes of indirect call and jump via memory: Likewise.
11728 * config/i386/i386.opt: Likewise.
11729 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11730 (GOT_memory_operand): Likewise.
11731 (call_insn_operand): Likewise.
11732 (sibcall_insn_operand): Likewise.
11733 (GOT32_symbol_operand): Likewise.
11735 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11737 PR middle-end/83837
11738 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11739 type rather than type addr's type points to.
11740 (expand_omp_atomic_mutex): Likewise.
11741 (expand_omp_atomic): Likewise.
11743 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11746 * config/i386/i386.c (output_indirect_thunk_function): Use
11747 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11748 for __x86_return_thunk.
11750 2018-01-15 Richard Biener <rguenther@suse.de>
11752 PR middle-end/83850
11753 * expmed.c (extract_bit_field_1): Fix typo.
11755 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11758 * config/arm/iterators.md (VF): New mode iterator.
11759 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11760 Remove integer-related logic from pattern.
11761 (neon_vabd<mode>_3): Likewise.
11763 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11765 PR middle-end/82694
11766 * common.opt (fstrict-overflow): No longer an alias.
11767 (fwrapv-pointer): New option.
11768 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11769 also for pointer types based on flag_wrapv_pointer.
11770 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11771 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11772 opts->x_flag_wrapv got set.
11773 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11774 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11775 POINTER_TYPE_OVERFLOW_UNDEFINED.
11776 * match.pd: Likewise in address comparison pattern.
11777 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11779 2018-01-15 Richard Biener <rguenther@suse.de>
11782 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11783 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11784 Reset type names to their identifier if their TYPE_DECL doesn't
11785 have linkage (and thus is used for ODR and devirt).
11786 (save_debug_info_for_decl): Remove.
11787 (save_debug_info_for_type): Likewise.
11788 (add_tree_to_fld_list): Adjust.
11789 * tree-pretty-print.c (dump_generic_node): Make dumping of
11790 type names more robust.
11792 2018-01-15 Richard Biener <rguenther@suse.de>
11794 * BASE-VER: Bump to 8.0.1.
11796 2018-01-14 Martin Sebor <msebor@redhat.com>
11799 * builtins.c (check_access): Avoid warning when the no-warning bit
11802 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11804 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11805 * ira-color (allocno_hard_regs_compare): Likewise.
11807 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11810 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11811 Use .pushsection/.popsection.
11813 2018-01-14 Martin Sebor <msebor@redhat.com>
11816 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11818 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11820 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11821 entry from extra_headers.
11822 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11823 extra_headers, make the list bitwise identical to the i?86-*-* one.
11825 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11827 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11828 -mcmodel=large with -mindirect-branch=thunk,
11829 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11830 -mfunction-return=thunk-extern.
11831 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11832 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11833 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11835 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11837 * config/i386/i386.c (print_reg): Print the name of the full
11838 integer register without '%'.
11839 (ix86_print_operand): Handle 'V'.
11840 * doc/extend.texi: Document 'V' modifier.
11842 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11844 * config/i386/constraints.md (Bs): Disallow memory operand for
11845 -mindirect-branch-register.
11847 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11848 (GOT_memory_operand): Likewise.
11849 (call_insn_operand): Likewise.
11850 (sibcall_insn_operand): Likewise.
11851 (GOT32_symbol_operand): Likewise.
11852 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11853 for -mindirect-branch-register.
11854 (tablejump): Likewise.
11855 (*sibcall_memory): Likewise.
11856 (*sibcall_value_memory): Likewise.
11857 Disallow peepholes of indirect call and jump via memory for
11858 -mindirect-branch-register.
11859 (*call_pop): Replace m with Bw.
11860 (*call_value_pop): Likewise.
11861 (*sibcall_pop_memory): Replace m with Bs.
11862 * config/i386/i386.opt (mindirect-branch-register): New option.
11863 * doc/invoke.texi: Document -mindirect-branch-register option.
11865 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11867 * config/i386/i386-protos.h (ix86_output_function_return): New.
11868 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11869 set function_return_type.
11870 (indirect_thunk_name): Add ret_p to indicate thunk for function
11872 (output_indirect_thunk_function): Pass false to
11873 indirect_thunk_name.
11874 (ix86_output_indirect_branch_via_reg): Likewise.
11875 (ix86_output_indirect_branch_via_push): Likewise.
11876 (output_indirect_thunk_function): Create alias for function
11877 return thunk if regno < 0.
11878 (ix86_output_function_return): New function.
11879 (ix86_handle_fndecl_attribute): Handle function_return.
11880 (ix86_attribute_table): Add function_return.
11881 * config/i386/i386.h (machine_function): Add
11882 function_return_type.
11883 * config/i386/i386.md (simple_return_internal): Use
11884 ix86_output_function_return.
11885 (simple_return_internal_long): Likewise.
11886 * config/i386/i386.opt (mfunction-return=): New option.
11887 (indirect_branch): Mention -mfunction-return=.
11888 * doc/extend.texi: Document function_return function attribute.
11889 * doc/invoke.texi: Document -mfunction-return= option.
11891 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11893 * config/i386/i386-opts.h (indirect_branch): New.
11894 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11895 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11896 with local indirect jump when converting indirect call and jump.
11897 (ix86_set_indirect_branch_type): New.
11898 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11899 (indirectlabelno): New.
11900 (indirect_thunk_needed): Likewise.
11901 (indirect_thunk_bnd_needed): Likewise.
11902 (indirect_thunks_used): Likewise.
11903 (indirect_thunks_bnd_used): Likewise.
11904 (INDIRECT_LABEL): Likewise.
11905 (indirect_thunk_name): Likewise.
11906 (output_indirect_thunk): Likewise.
11907 (output_indirect_thunk_function): Likewise.
11908 (ix86_output_indirect_branch_via_reg): Likewise.
11909 (ix86_output_indirect_branch_via_push): Likewise.
11910 (ix86_output_indirect_branch): Likewise.
11911 (ix86_output_indirect_jmp): Likewise.
11912 (ix86_code_end): Call output_indirect_thunk_function if needed.
11913 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11915 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11916 (ix86_attribute_table): Add indirect_branch.
11917 * config/i386/i386.h (machine_function): Add indirect_branch_type
11918 and has_local_indirect_jump.
11919 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11921 (tablejump): Likewise.
11922 (*indirect_jump): Use ix86_output_indirect_jmp.
11923 (*tablejump_1): Likewise.
11924 (simple_return_indirect_internal): Likewise.
11925 * config/i386/i386.opt (mindirect-branch=): New option.
11926 (indirect_branch): New.
11929 (thunk-inline): Likewise.
11930 (thunk-extern): Likewise.
11931 * doc/extend.texi: Document indirect_branch function attribute.
11932 * doc/invoke.texi: Document -mindirect-branch= option.
11934 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11937 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11939 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11941 * ipa-inline.c (want_inline_small_function_p): Return false if
11942 inlining has already failed with CIF_FINAL_ERROR.
11943 (update_caller_keys): Call want_inline_small_function_p before
11945 (update_callee_keys): Likewise.
11947 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11949 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11951 (rs6000_quadword_masked_address_p): Likewise.
11952 (quad_aligned_load_p): Likewise.
11953 (quad_aligned_store_p): Likewise.
11954 (const_load_sequence_p): Add comment to describe the outer-most loop.
11955 (mimic_memory_attributes_and_flags): New function.
11956 (rs6000_gen_stvx): Likewise.
11957 (replace_swapped_aligned_store): Likewise.
11958 (rs6000_gen_lvx): Likewise.
11959 (replace_swapped_aligned_load): Likewise.
11960 (replace_swapped_load_constant): Capitalize argument name in
11961 comment describing this function.
11962 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11963 and stores that access quad-word aligned addresses and replace
11964 with stvx or lvx instructions when appropriate.
11965 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11966 New function prototype.
11967 (rs6000_quadword_masked_address_p): Likewise.
11968 (rs6000_gen_lvx): Likewise.
11969 (rs6000_gen_stvx): Likewise.
11970 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11971 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11972 when memory address is aligned.
11973 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11974 this split to select lvx instruction when memory address is aligned.
11975 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11976 instruction when memory address is aligned.
11977 (*vsx_le_perm_load_v16qi): Likewise.
11978 (four unnamed splitters): Modify to select the stvx instruction
11979 when memory is aligned.
11981 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11983 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11984 which appears in the queue multiple times.
11986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11987 Alan Hayward <alan.hayward@arm.com>
11988 David Sherwood <david.sherwood@arm.com>
11990 * tree-vectorizer.h (vec_lower_bound): New structure.
11991 (_loop_vec_info): Add check_nonzero and lower_bounds.
11992 (LOOP_VINFO_CHECK_NONZERO): New macro.
11993 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11994 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11995 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11996 fields. Make seg_len the distance travelled, not including the
11998 (dr_direction_indicator): Declare.
11999 (dr_zero_step_indicator): Likewise.
12000 (dr_known_forward_stride_p): Likewise.
12001 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12003 (runtime_alias_check_p): Allow runtime alias checks with
12005 (operator ==): Compare access_size and align.
12006 (prune_runtime_alias_test_list): Rework for new distinction between
12007 the access_size and seg_len.
12008 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12010 (get_segment_min_max): New function.
12011 (create_intersect_range_checks): Use it.
12012 (dr_step_indicator): New function.
12013 (dr_direction_indicator): Likewise.
12014 (dr_zero_step_indicator): Likewise.
12015 (dr_known_forward_stride_p): Likewise.
12016 * tree-loop-distribution.c (data_ref_segment_size): Return
12017 DR_STEP * (niters - 1).
12018 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12020 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12021 (vect_preserves_scalar_order_p): New function, split out from...
12022 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12023 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12024 (vect_vfa_access_size): New function.
12025 (vect_vfa_align): Likewise.
12026 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12027 (dump_lower_bound): New function.
12028 (vect_check_lower_bound): Likewise.
12029 (vect_small_gap_p): Likewise.
12030 (vectorizable_with_step_bound_p): Likewise.
12031 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12032 depencies if the vectorization factor is 1. Convert the checks
12033 for nonzero steps into checks on the bounds of DR_STEP. Try using
12034 a bunds check for variable steps if the minimum required step is
12035 relatively small. Update calls to the dr_with_seg_len
12036 constructor and to vect_compile_time_alias.
12037 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12039 (vect_loop_versioning): Call it.
12040 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12042 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12045 Alan Hayward <alan.hayward@arm.com>
12046 David Sherwood <david.sherwood@arm.com>
12048 * doc/sourcebuild.texi (vect_scatter_store): Document.
12049 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12051 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12053 * genopinit.c (main): Add supports_vec_scatter_store and
12054 supports_vec_scatter_store_cached to target_optabs.
12055 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12056 IFN_MASK_SCATTER_STORE.
12057 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12059 * internal-fn.h (internal_store_fn_p): Declare.
12060 (internal_fn_stored_value_index): Likewise.
12061 * internal-fn.c (scatter_store_direct): New macro.
12062 (expand_scatter_store_optab_fn): New function.
12063 (direct_scatter_store_optab_supported_p): New macro.
12064 (internal_store_fn_p): New function.
12065 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12066 IFN_MASK_SCATTER_STORE.
12067 (internal_fn_mask_index): Likewise.
12068 (internal_fn_stored_value_index): New function.
12069 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12070 for scatter stores.
12071 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12072 * optabs-query.c (supports_vec_scatter_store_p): New function.
12073 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12074 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12075 true for scatter stores.
12076 (vect_gather_scatter_fn_p): Handle scatter stores too.
12077 (vect_check_gather_scatter): Consider using scatter stores if
12078 supports_vec_scatter_store_p.
12079 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12080 scatter stores too.
12081 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12082 internal_fn_stored_value_index.
12083 (check_load_store_masking): Handle scatter stores too.
12084 (vect_get_store_rhs): Make public.
12085 (vectorizable_call): Use internal_store_fn_p.
12086 (vectorizable_store): Handle scatter store internal functions.
12087 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12088 when deciding whether the end of the group has been reached.
12089 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12090 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12091 (mask_scatter_store<mode>): New insns.
12093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12094 Alan Hayward <alan.hayward@arm.com>
12095 David Sherwood <david.sherwood@arm.com>
12097 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12098 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12099 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12101 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12102 Use vect_truncate_gather_scatter_offset if we can't treat the
12103 operation as a normal gather load or scatter store.
12104 (get_group_load_store_type): Take the gather_scatter_info
12105 as argument. Try using a gather load or scatter store for
12106 single-element groups.
12107 (get_load_store_type): Update calls to get_group_load_store_type
12108 and vect_use_strided_gather_scatters_p.
12110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12111 Alan Hayward <alan.hayward@arm.com>
12112 David Sherwood <david.sherwood@arm.com>
12114 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12115 optional tree argument.
12116 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12118 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12119 but continue to use the current value as a fallback.
12120 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12121 to compare the updates.
12122 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12123 (get_load_store_type): Use it when handling a strided access.
12124 (vect_get_strided_load_store_ops): New function.
12125 (vect_get_data_ptr_increment): Likewise.
12126 (vectorizable_load): Handle strided gather loads. Always pass
12127 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12130 Alan Hayward <alan.hayward@arm.com>
12131 David Sherwood <david.sherwood@arm.com>
12133 * doc/md.texi (gather_load@var{m}): Document.
12134 (mask_gather_load@var{m}): Likewise.
12135 * genopinit.c (main): Add supports_vec_gather_load and
12136 supports_vec_gather_load_cached to target_optabs.
12137 * optabs-tree.c (init_tree_optimization_optabs): Use
12138 ggc_cleared_alloc to allocate target_optabs.
12139 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12140 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12142 * internal-fn.h (internal_load_fn_p): Declare.
12143 (internal_gather_scatter_fn_p): Likewise.
12144 (internal_fn_mask_index): Likewise.
12145 (internal_gather_scatter_fn_supported_p): Likewise.
12146 * internal-fn.c (gather_load_direct): New macro.
12147 (expand_gather_load_optab_fn): New function.
12148 (direct_gather_load_optab_supported_p): New macro.
12149 (direct_internal_fn_optab): New function.
12150 (internal_load_fn_p): Likewise.
12151 (internal_gather_scatter_fn_p): Likewise.
12152 (internal_fn_mask_index): Likewise.
12153 (internal_gather_scatter_fn_supported_p): Likewise.
12154 * optabs-query.c (supports_at_least_one_mode_p): New function.
12155 (supports_vec_gather_load_p): Likewise.
12156 * optabs-query.h (supports_vec_gather_load_p): Declare.
12157 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12158 and memory_type field.
12159 (NUM_PATTERNS): Bump to 15.
12160 * tree-vect-data-refs.c: Include internal-fn.h.
12161 (vect_gather_scatter_fn_p): New function.
12162 (vect_describe_gather_scatter_call): Likewise.
12163 (vect_check_gather_scatter): Try using internal functions for
12164 gather loads. Recognize existing calls to a gather load function.
12165 (vect_analyze_data_refs): Consider using gather loads if
12166 supports_vec_gather_load_p.
12167 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12168 (vect_get_gather_scatter_offset_type): Likewise.
12169 (vect_convert_mask_for_vectype): Likewise.
12170 (vect_add_conversion_to_patterm): Likewise.
12171 (vect_try_gather_scatter_pattern): Likewise.
12172 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12173 (vect_vect_recog_func_ptrs): Add it.
12174 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12175 internal_fn_mask_index and internal_gather_scatter_fn_p.
12176 (check_load_store_masking): Take the gather_scatter_info as an
12177 argument and handle gather loads.
12178 (vect_get_gather_scatter_ops): New function.
12179 (vectorizable_call): Check internal_load_fn_p.
12180 (vectorizable_load): Likewise. Handle gather load internal
12182 (vectorizable_store): Update call to check_load_store_masking.
12183 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12184 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12185 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12186 (aarch64_gather_scale_operand_d): New predicates.
12187 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12188 (mask_gather_load<mode>): New insns.
12190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12191 Alan Hayward <alan.hayward@arm.com>
12192 David Sherwood <david.sherwood@arm.com>
12194 * optabs.def (fold_left_plus_optab): New optab.
12195 * doc/md.texi (fold_left_plus_@var{m}): Document.
12196 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12197 * internal-fn.c (fold_left_direct): Define.
12198 (expand_fold_left_optab_fn): Likewise.
12199 (direct_fold_left_optab_supported_p): Likewise.
12200 * fold-const-call.c (fold_const_fold_left): New function.
12201 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12202 * tree-parloops.c (valid_reduction_p): New function.
12203 (gather_scalar_reductions): Use it.
12204 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12205 (vect_finish_replace_stmt): Declare.
12206 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12207 (needs_fold_left_reduction_p): New function, split out from...
12208 (vect_is_simple_reduction): ...here. Accept reductions that
12209 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12210 (vect_force_simple_reduction): Also store the reduction type in
12211 the assignment's STMT_VINFO_REDUC_TYPE.
12212 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12213 (merge_with_identity): New function.
12214 (vect_expand_fold_left): Likewise.
12215 (vectorize_fold_left_reduction): Likewise.
12216 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12217 scalar phi in place for it. Check for target support and reject
12218 cases that would reassociate the operation. Defer the transform
12219 phase to vectorize_fold_left_reduction.
12220 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12221 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12222 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12226 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12227 call to ifc_temp_var.
12229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * target.def (legitimize_address_displacement): Take the original
12234 offset as a poly_int.
12235 * targhooks.h (default_legitimize_address_displacement): Update
12237 * targhooks.c (default_legitimize_address_displacement): Likewise.
12238 * doc/tm.texi: Regenerate.
12239 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12240 as an argument, moving assert of ad->disp == ad->disp_term to...
12241 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12242 Try calling targetm.legitimize_address_displacement before expanding
12243 the address rather than afterwards, and adjust for the new interface.
12244 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12245 Match the new hook interface. Handle SVE addresses.
12246 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12247 new hook interface.
12249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12251 * Makefile.in (OBJS): Add early-remat.o.
12252 * target.def (select_early_remat_modes): New hook.
12253 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12254 * doc/tm.texi: Regenerate.
12255 * targhooks.h (default_select_early_remat_modes): Declare.
12256 * targhooks.c (default_select_early_remat_modes): New function.
12257 * timevar.def (TV_EARLY_REMAT): New timevar.
12258 * passes.def (pass_early_remat): New pass.
12259 * tree-pass.h (make_pass_early_remat): Declare.
12260 * early-remat.c: New file.
12261 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12263 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12266 Alan Hayward <alan.hayward@arm.com>
12267 David Sherwood <david.sherwood@arm.com>
12269 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12270 vfm1 with a bound_epilog parameter.
12271 (vect_do_peeling): Update calls accordingly, and move the prologue
12272 call earlier in the function. Treat the base bound_epilog as 0 for
12273 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12274 this base when peeling for gaps.
12275 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12276 with fully-masked loops.
12277 (vect_estimate_min_profitable_iters): Handle the single peeled
12278 iteration in that case.
12280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12285 single-element interleaving even if the size is not a power of 2.
12286 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12287 accesses for single-element interleaving if the group size is
12290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12291 Alan Hayward <alan.hayward@arm.com>
12292 David Sherwood <david.sherwood@arm.com>
12294 * doc/md.texi (fold_extract_last_@var{m}): Document.
12295 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12296 * optabs.def (fold_extract_last_optab): New optab.
12297 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12298 * internal-fn.c (fold_extract_direct): New macro.
12299 (expand_fold_extract_optab_fn): Likewise.
12300 (direct_fold_extract_optab_supported_p): Likewise.
12301 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12302 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12303 EXTRACT_LAST_REDUCTION.
12304 (get_initial_def_for_reduction): Do not create an initial vector
12305 for EXTRACT_LAST_REDUCTION reductions.
12306 (vectorizable_reduction): Leave the scalar phi in place for
12307 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12308 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12309 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12310 transform phase to vectorizable_condition.
12311 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12313 (vect_finish_stmt_generation): ...here.
12314 (vect_finish_replace_stmt): New function.
12315 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12316 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12318 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12321 Alan Hayward <alan.hayward@arm.com>
12322 David Sherwood <david.sherwood@arm.com>
12324 * doc/md.texi (extract_last_@var{m}): Document.
12325 * optabs.def (extract_last_optab): New optab.
12326 * internal-fn.def (EXTRACT_LAST): New internal function.
12327 * internal-fn.c (cond_unary_direct): New macro.
12328 (expand_cond_unary_optab_fn): Likewise.
12329 (direct_cond_unary_optab_supported_p): Likewise.
12330 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12331 loops using EXTRACT_LAST.
12332 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12333 (extract_last_<mode>): ...this optab.
12334 (vec_extract<mode><Vel>): Update accordingly.
12336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12337 Alan Hayward <alan.hayward@arm.com>
12338 David Sherwood <david.sherwood@arm.com>
12340 * target.def (empty_mask_is_expensive): New hook.
12341 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12342 * doc/tm.texi: Regenerate.
12343 * targhooks.h (default_empty_mask_is_expensive): Declare.
12344 * targhooks.c (default_empty_mask_is_expensive): New function.
12345 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12346 if the target says that empty masks are expensive.
12347 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12349 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12356 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12357 (vect_use_loop_mask_for_alignment_p): New function.
12358 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12359 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12360 niters_skip argument. Make sure that the first niters_skip elements
12361 of the first iteration are inactive.
12362 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12363 Update call to vect_set_loop_masks_directly.
12364 (get_misalign_in_elems): New function, split out from...
12365 (vect_gen_prolog_loop_niters): ...here.
12366 (vect_update_init_of_dr): Take a code argument that specifies whether
12367 the adjustment should be added or subtracted.
12368 (vect_update_init_of_drs): Likewise.
12369 (vect_prepare_for_masked_peels): New function.
12370 (vect_do_peeling): Skip prologue peeling if we're using a mask
12371 instead. Update call to vect_update_inits_of_drs.
12372 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12374 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12375 alignment. Do not include the number of peeled iterations in
12376 the minimum threshold in that case.
12377 (vectorizable_induction): Adjust the start value down by
12378 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12379 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12380 Take the number of skipped iterations into account when calculating
12382 * tree-vect-stmts.c (vect_gen_while_not): New function.
12384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12385 Alan Hayward <alan.hayward@arm.com>
12386 David Sherwood <david.sherwood@arm.com>
12388 * doc/sourcebuild.texi (vect_fully_masked): Document.
12389 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12390 default value to 0.
12391 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12393 (vect_analyze_loop_2): ...here. Don't check the vectorization
12394 factor against the number of loop iterations if the loop is
12397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12398 Alan Hayward <alan.hayward@arm.com>
12399 David Sherwood <david.sherwood@arm.com>
12401 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12402 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12403 (dump_groups): Update accordingly.
12404 (iv_use::mem_type): New member variable.
12405 (address_p): New function.
12406 (record_use): Add a mem_type argument and initialize the new
12408 (record_group_use): Add a mem_type argument. Use address_p.
12409 Remove obsolete null checks of base_object. Update call to record_use.
12410 (find_interesting_uses_op): Update call to record_group_use.
12411 (find_interesting_uses_cond): Likewise.
12412 (find_interesting_uses_address): Likewise.
12413 (get_mem_type_for_internal_fn): New function.
12414 (find_address_like_use): Likewise.
12415 (find_interesting_uses_stmt): Try find_address_like_use before
12416 calling find_interesting_uses_op.
12417 (addr_offset_valid_p): Use the iv mem_type field as the type
12418 of the addressed memory.
12419 (add_autoinc_candidates): Likewise.
12420 (get_address_cost): Likewise.
12421 (split_small_address_groups_p): Use address_p.
12422 (split_address_groups): Likewise.
12423 (add_iv_candidate_for_use): Likewise.
12424 (autoinc_possible_for_pair): Likewise.
12425 (rewrite_groups): Likewise.
12426 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12427 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12428 (get_alias_ptr_type_for_ptr_address): New function.
12429 (rewrite_use_address): Rewrite address uses in calls that were
12430 identified by find_address_like_use.
12432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12433 Alan Hayward <alan.hayward@arm.com>
12434 David Sherwood <david.sherwood@arm.com>
12436 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12438 * gimple-expr.h (is_gimple_addressable: Likewise.
12439 * gimple-expr.c (is_gimple_address): Likewise.
12440 * internal-fn.c (expand_call_mem_ref): New function.
12441 (expand_mask_load_optab_fn): Use it.
12442 (expand_mask_store_optab_fn): Likewise.
12444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12445 Alan Hayward <alan.hayward@arm.com>
12446 David Sherwood <david.sherwood@arm.com>
12448 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12449 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12450 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12451 (cond_umax@var{mode}): Document.
12452 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12453 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12454 (cond_umin_optab, cond_umax_optab): New optabs.
12455 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12456 (COND_IOR, COND_XOR): New internal functions.
12457 * internal-fn.h (get_conditional_internal_fn): Declare.
12458 * internal-fn.c (cond_binary_direct): New macro.
12459 (expand_cond_binary_optab_fn): Likewise.
12460 (direct_cond_binary_optab_supported_p): Likewise.
12461 (get_conditional_internal_fn): New function.
12462 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12463 Cope with reduction statements that are vectorized as calls rather
12465 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12466 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12467 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12468 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12469 (UNSPEC_COND_EOR): New unspecs.
12470 (optab): Add mappings for them.
12471 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12472 (sve_int_op, sve_fp_op): New int attributes.
12474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12475 Alan Hayward <alan.hayward@arm.com>
12476 David Sherwood <david.sherwood@arm.com>
12478 * optabs.def (while_ult_optab): New optab.
12479 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12480 * internal-fn.def (WHILE_ULT): New internal function.
12481 * internal-fn.h (direct_internal_fn_supported_p): New override
12482 that takes two types as argument.
12483 * internal-fn.c (while_direct): New macro.
12484 (expand_while_optab_fn): New function.
12485 (convert_optab_supported_p): Likewise.
12486 (direct_while_optab_supported_p): New macro.
12487 * wide-int.h (wi::udiv_ceil): New function.
12488 * tree-vectorizer.h (rgroup_masks): New structure.
12489 (vec_loop_masks): New typedef.
12490 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12491 and fully_masked_p.
12492 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12493 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12494 (vect_max_vf): New function.
12495 (slpeel_make_loop_iterate_ntimes): Delete.
12496 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12497 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12498 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12499 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12500 internal-fn.h, stor-layout.h and optabs-query.h.
12501 (vect_set_loop_mask): New function.
12502 (add_preheader_seq): Likewise.
12503 (add_header_seq): Likewise.
12504 (interleave_supported_p): Likewise.
12505 (vect_maybe_permute_loop_masks): Likewise.
12506 (vect_set_loop_masks_directly): Likewise.
12507 (vect_set_loop_condition_masked): Likewise.
12508 (vect_set_loop_condition_unmasked): New function, split out from
12509 slpeel_make_loop_iterate_ntimes.
12510 (slpeel_make_loop_iterate_ntimes): Rename to..
12511 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12512 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12513 (vect_do_peeling): Update call accordingly.
12514 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12516 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12517 mask_compare_type, can_fully_mask_p and fully_masked_p.
12518 (release_vec_loop_masks): New function.
12519 (_loop_vec_info): Use it to free the loop masks.
12520 (can_produce_all_loop_masks_p): New function.
12521 (vect_get_max_nscalars_per_iter): Likewise.
12522 (vect_verify_full_masking): Likewise.
12523 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12524 retries, and free the mask rgroups before retrying. Check loop-wide
12525 reasons for disallowing fully-masked loops. Make the final decision
12526 about whether use a fully-masked loop or not.
12527 (vect_estimate_min_profitable_iters): Do not assume that peeling
12528 for the number of iterations will be needed for fully-masked loops.
12529 (vectorizable_reduction): Disable fully-masked loops.
12530 (vectorizable_live_operation): Likewise.
12531 (vect_halve_mask_nunits): New function.
12532 (vect_double_mask_nunits): Likewise.
12533 (vect_record_loop_mask): Likewise.
12534 (vect_get_loop_mask): Likewise.
12535 (vect_transform_loop): Handle the case in which the final loop
12536 iteration might handle a partial vector. Call vect_set_loop_condition
12537 instead of slpeel_make_loop_iterate_ntimes.
12538 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12539 (check_load_store_masking): New function.
12540 (prepare_load_store_mask): Likewise.
12541 (vectorizable_store): Handle fully-masked loops.
12542 (vectorizable_load): Likewise.
12543 (supportable_widening_operation): Use vect_halve_mask_nunits for
12545 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12546 (vect_gen_while): New function.
12547 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12548 (aarch64_uqdec<mode>): New insn.
12550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12555 (reduc_xor_scal_optab): New optabs.
12556 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12557 (reduc_xor_scal_@var{m}): Document.
12558 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12559 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12560 internal functions.
12561 * fold-const-call.c (fold_const_call): Handle them.
12562 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12563 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12564 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12565 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12566 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12567 (UNSPEC_XORV): New unspecs.
12568 (optab): Add entries for them.
12569 (BITWISEV): New int iterator.
12570 (bit_reduc_op): New int attributes.
12572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12573 Alan Hayward <alan.hayward@arm.com>
12574 David Sherwood <david.sherwood@arm.com>
12576 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12577 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12578 * optabs.def (vec_shl_insert_optab): New optab.
12579 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12580 (duplicate_and_interleave): Likewise.
12581 * tree-vect-loop.c: Include internal-fn.h.
12582 (neutral_op_for_slp_reduction): New function, split out from
12583 get_initial_defs_for_reduction.
12584 (get_initial_def_for_reduction): Handle option 2 for variable-length
12585 vectors by loading the neutral value into a vector and then shifting
12586 the initial value into element 0.
12587 (get_initial_defs_for_reduction): Replace the code argument with
12588 the neutral value calculated by neutral_op_for_slp_reduction.
12589 Use gimple_build_vector for constant-length vectors.
12590 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12591 but the first group_size elements have a neutral value.
12592 Use duplicate_and_interleave otherwise.
12593 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12594 Update call to get_initial_defs_for_reduction. Handle SLP
12595 reductions for variable-length vectors by creating one vector
12596 result for each scalar result, with the elements associated
12597 with other scalar results stubbed out with the neutral value.
12598 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12599 Require IFN_VEC_SHL_INSERT for double reductions on
12600 variable-length vectors, or SLP reductions that have
12601 a neutral value. Require can_duplicate_and_interleave_p
12602 support for variable-length unchained SLP reductions if there
12603 is no neutral value, such as for MIN/MAX reductions. Also require
12604 the number of vector elements to be a multiple of the number of
12605 SLP statements when doing variable-length unchained SLP reductions.
12606 Update call to vect_create_epilog_for_reduction.
12607 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12608 and remove initial values.
12609 (duplicate_and_interleave): Make public.
12610 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12611 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12614 Alan Hayward <alan.hayward@arm.com>
12615 David Sherwood <david.sherwood@arm.com>
12617 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12618 (can_duplicate_and_interleave_p): New function.
12619 (vect_get_and_check_slp_defs): Take the vector of statements
12620 rather than just the current one. Remove excess parentheses.
12621 Restriction rejectinon of vect_constant_def and vect_external_def
12622 for variable-length vectors to boolean types, or types for which
12623 can_duplicate_and_interleave_p is false.
12624 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12625 (duplicate_and_interleave): New function.
12626 (vect_get_constant_vectors): Use gimple_build_vector for
12627 constant-length vectors and suitable variable-length constant
12628 vectors. Use duplicate_and_interleave for other variable-length
12629 vectors. Don't defer the update when inserting new statements.
12631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12632 Alan Hayward <alan.hayward@arm.com>
12633 David Sherwood <david.sherwood@arm.com>
12635 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12636 min_profitable_iters doesn't go negative.
12638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12639 Alan Hayward <alan.hayward@arm.com>
12640 David Sherwood <david.sherwood@arm.com>
12642 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12643 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12644 * optabs.def (vec_mask_load_lanes_optab): New optab.
12645 (vec_mask_store_lanes_optab): Likewise.
12646 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12647 (MASK_STORE_LANES): Likewise.
12648 * internal-fn.c (mask_load_lanes_direct): New macro.
12649 (mask_store_lanes_direct): Likewise.
12650 (expand_mask_load_optab_fn): Handle masked operations.
12651 (expand_mask_load_lanes_optab_fn): New macro.
12652 (expand_mask_store_optab_fn): Handle masked operations.
12653 (expand_mask_store_lanes_optab_fn): New macro.
12654 (direct_mask_load_lanes_optab_supported_p): Likewise.
12655 (direct_mask_store_lanes_optab_supported_p): Likewise.
12656 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12658 (vect_load_lanes_supported): Likewise.
12659 * tree-vect-data-refs.c (strip_conversion): New function.
12660 (can_group_stmts_p): Likewise.
12661 (vect_analyze_data_ref_accesses): Use it instead of checking
12662 for a pair of assignments.
12663 (vect_store_lanes_supported): Take a masked_p parameter.
12664 (vect_load_lanes_supported): Likewise.
12665 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12666 vect_store_lanes_supported and vect_load_lanes_supported.
12667 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12668 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12669 parameter. Don't allow gaps for masked accesses.
12670 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12671 and vect_load_lanes_supported.
12672 (get_load_store_type): Take a masked_p parameter and update
12673 call to get_group_load_store_type.
12674 (vectorizable_store): Update call to get_load_store_type.
12675 Handle IFN_MASK_STORE_LANES.
12676 (vectorizable_load): Update call to get_load_store_type.
12677 Handle IFN_MASK_LOAD_LANES.
12679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12680 Alan Hayward <alan.hayward@arm.com>
12681 David Sherwood <david.sherwood@arm.com>
12683 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12685 * config/aarch64/aarch64-protos.h
12686 (aarch64_sve_struct_memory_operand_p): Declare.
12687 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12688 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12689 (VPRED, vpred): Handle SVE structure modes.
12690 * config/aarch64/constraints.md (Utx): New constraint.
12691 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12692 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12693 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12694 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12695 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12696 structure modes. Split into pieces after RA.
12697 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12698 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12700 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12701 SVE structure modes.
12702 (aarch64_classify_address): Likewise.
12703 (sizetochar): Move earlier in file.
12704 (aarch64_print_operand): Handle SVE register lists.
12705 (aarch64_array_mode): New function.
12706 (aarch64_sve_struct_memory_operand_p): Likewise.
12707 (TARGET_ARRAY_MODE): Redefine.
12709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12710 Alan Hayward <alan.hayward@arm.com>
12711 David Sherwood <david.sherwood@arm.com>
12713 * target.def (array_mode): New target hook.
12714 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12715 * doc/tm.texi: Regenerate.
12716 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12717 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12718 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12719 targetm.array_mode.
12720 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12724 Alan Hayward <alan.hayward@arm.com>
12725 David Sherwood <david.sherwood@arm.com>
12727 * fold-const.c (fold_binary_loc): Check the argument types
12728 rather than the result type when testing for a vector operation.
12730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12732 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12733 * doc/tm.texi: Regenerate.
12735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12740 (sve): Document new AArch64 extension.
12741 * doc/md.texi (w): Extend the description of the AArch64
12742 constraint to include SVE vectors.
12743 (Upl, Upa): Document new AArch64 predicate constraints.
12744 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12746 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12747 (msve-vector-bits=): New option.
12748 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12749 SVE when these are disabled.
12750 (sve): New extension.
12751 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12752 modes. Adjust their number of units based on aarch64_sve_vg.
12753 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12754 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12755 aarch64_addr_query_type.
12756 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12757 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12758 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12759 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12760 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12761 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12762 (aarch64_simd_imm_zero_p): Delete.
12763 (aarch64_check_zero_based_sve_index_immediate): Declare.
12764 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12765 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12766 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12767 (aarch64_sve_float_mul_immediate_p): Likewise.
12768 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12769 rather than an rtx.
12770 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12771 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12772 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12773 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12774 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12775 (aarch64_regmode_natural_size): Likewise.
12776 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12777 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12779 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12780 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12781 for VG and the SVE predicate registers.
12782 (V_ALIASES): Add a "z"-prefixed alias.
12783 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12784 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12785 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12786 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12787 (REG_CLASS_NAMES): Add entries for them.
12788 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12789 and the predicate registers.
12790 (aarch64_sve_vg): Declare.
12791 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12792 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12793 (REGMODE_NATURAL_SIZE): Define.
12794 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12796 * config/aarch64/aarch64.c: Include cfgrtl.h.
12797 (simd_immediate_info): Add a constructor for series vectors,
12798 and an associated step field.
12799 (aarch64_sve_vg): New variable.
12800 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12801 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12802 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12803 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12804 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12805 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12806 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12807 (aarch64_get_mask_mode): New functions.
12808 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12809 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12810 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12811 predicate modes and predicate registers. Explicitly restrict
12812 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12813 to store a vector mode if it is recognized by
12814 aarch64_classify_vector_mode.
12815 (aarch64_regmode_natural_size): New function.
12816 (aarch64_hard_regno_caller_save_mode): Return the original mode
12818 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12819 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12820 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12821 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12823 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12824 does not overlap dest if the function is frame-related. Handle
12826 (aarch64_split_add_offset): New function.
12827 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12828 them aarch64_add_offset.
12829 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12830 and update call to aarch64_sub_sp.
12831 (aarch64_add_cfa_expression): New function.
12832 (aarch64_expand_prologue): Pass extra temporary registers to the
12833 functions above. Handle the case in which we need to emit new
12834 DW_CFA_expressions for registers that were originally saved
12835 relative to the stack pointer, but now have to be expressed
12836 relative to the frame pointer.
12837 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12839 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12840 IP0 and IP1 values for SVE frames.
12841 (aarch64_expand_vec_series): New function.
12842 (aarch64_expand_sve_widened_duplicate): Likewise.
12843 (aarch64_expand_sve_const_vector): Likewise.
12844 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12845 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12846 into the register, rather than emitting a SET directly.
12847 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12848 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12849 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12850 (offset_9bit_signed_scaled_p): New functions.
12851 (aarch64_replicate_bitmask_imm): New function.
12852 (aarch64_bitmask_imm): Use it.
12853 (aarch64_cannot_force_const_mem): Reject expressions involving
12854 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12855 (aarch64_classify_index): Handle SVE indices, by requiring
12856 a plain register index with a scale that matches the element size.
12857 (aarch64_classify_address): Handle SVE addresses. Assert that
12858 the mode of the address is VOIDmode or an integer mode.
12859 Update call to aarch64_classify_symbol.
12860 (aarch64_classify_symbolic_expression): Update call to
12861 aarch64_classify_symbol.
12862 (aarch64_const_vec_all_in_range_p): New function.
12863 (aarch64_print_vector_float_operand): Likewise.
12864 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12865 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12866 and the FP immediates 1.0 and 0.5.
12867 (aarch64_print_address_internal): Handle SVE addresses.
12868 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12869 (aarch64_regno_regclass): Handle predicate registers.
12870 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12872 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12873 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12874 (aarch64_convert_sve_vector_bits): New function.
12875 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12876 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12877 rather than an rtx.
12878 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12879 Handle SVE vector and predicate modes. Accept VL-based constants
12880 that need only one temporary register, and VL offsets that require
12881 no temporary registers.
12882 (aarch64_conditional_register_usage): Mark the predicate registers
12883 as fixed if SVE isn't available.
12884 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12885 Return true for SVE vector and predicate modes.
12886 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12887 rather than an unsigned int. Handle SVE modes.
12888 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12890 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12892 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12893 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12894 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12895 (aarch64_sve_float_mul_immediate_p): New functions.
12896 (aarch64_sve_valid_immediate): New function.
12897 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12898 Explicitly reject structure modes. Check for INDEX constants.
12899 Handle PTRUE and PFALSE constants.
12900 (aarch64_check_zero_based_sve_index_immediate): New function.
12901 (aarch64_simd_imm_zero_p): Delete.
12902 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12903 vector modes. Accept constants in the range of CNT[BHWD].
12904 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12905 ask for an Advanced SIMD mode.
12906 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12907 (aarch64_simd_vector_alignment): Handle SVE predicates.
12908 (aarch64_vectorize_preferred_vector_alignment): New function.
12909 (aarch64_simd_vector_alignment_reachable): Use it instead of
12911 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12912 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12914 (MAX_VECT_LEN): Delete.
12915 (expand_vec_perm_d): Add a vec_flags field.
12916 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12917 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12918 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12920 (aarch64_evpc_rev): Rename to...
12921 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12922 (aarch64_evpc_rev_global): New function.
12923 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12924 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12926 (aarch64_evpc_sve_tbl): New function.
12927 (aarch64_expand_vec_perm_const_1): Update after rename of
12928 aarch64_evpc_rev. Handle SVE permutes too, trying
12929 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12930 than aarch64_evpc_tbl.
12931 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12932 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12933 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12934 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12935 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12936 (aarch64_expand_sve_vcond): New functions.
12937 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12938 of aarch64_vector_mode_p.
12939 (aarch64_dwarf_poly_indeterminate_value): New function.
12940 (aarch64_compute_pressure_classes): Likewise.
12941 (aarch64_can_change_mode_class): Likewise.
12942 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12943 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12944 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12945 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12946 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12947 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12948 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12949 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12951 (Dn, Dl, Dr): Accept const as well as const_vector.
12952 (Dz): Likewise. Compare against CONST0_RTX.
12953 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12954 of "vector" where appropriate.
12955 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12956 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12957 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12958 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12959 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12960 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12961 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12962 (v_int_equiv): Extend to SVE modes.
12963 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12965 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12966 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12967 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12968 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12969 (SVE_COND_FP_CMP): New int iterators.
12970 (perm_hilo): Handle the new unpack unspecs.
12971 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12973 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12974 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12975 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12976 (aarch64_equality_operator, aarch64_constant_vector_operand)
12977 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12978 (aarch64_sve_nonimmediate_operand): Likewise.
12979 (aarch64_sve_general_operand): Likewise.
12980 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12981 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12982 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12983 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12984 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12985 (aarch64_sve_float_arith_immediate): Likewise.
12986 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12987 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12988 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12989 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12990 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12991 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12992 (aarch64_sve_float_arith_operand): Likewise.
12993 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12994 (aarch64_sve_float_mul_operand): Likewise.
12995 (aarch64_sve_vec_perm_operand): Likewise.
12996 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12997 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12998 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12999 as well as const_vector.
13000 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13001 in file. Use CONST0_RTX and CONSTM1_RTX.
13002 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13003 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13004 Use aarch64_simd_imm_zero.
13005 * config/aarch64/aarch64-sve.md: New file.
13006 * config/aarch64/aarch64.md: Include it.
13007 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13008 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13009 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13010 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13011 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13012 (sve): New attribute.
13013 (enabled): Disable instructions with the sve attribute unless
13015 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13016 aarch64_expand_mov_immediate.
13017 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13018 CNT[BHSD] immediates.
13019 (movti): Split CONST_POLY_INT moves into two halves.
13020 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13021 Split additions that need a temporary here if the destination
13022 is the stack pointer.
13023 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13024 (*add<mode>3_poly_1): New instruction.
13025 (set_clobber_cc): New expander.
13027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13029 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13030 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13031 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13032 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13033 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13034 Change innermode from fixed_mode_size to machine_mode.
13035 (simplify_subreg): Update call accordingly. Handle a constant-sized
13036 subreg of a variable-length CONST_VECTOR.
13038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13039 Alan Hayward <alan.hayward@arm.com>
13040 David Sherwood <david.sherwood@arm.com>
13042 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13043 (add_offset_to_base): New function, split out from...
13044 (create_mem_ref): ...here. When handling a scale other than 1,
13045 check first whether the address is valid without the offset.
13046 Add it into the base if so, leaving the index and scale as-is.
13048 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13051 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13052 fold_for_warn before checking if arg2 is INTEGER_CST.
13054 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13056 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13057 (store_multiple_operation): Delete.
13058 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13059 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13060 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13061 guarded by TARGET_STRING.
13062 (rs6000_output_load_multiple): Delete.
13063 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13064 OPTION_MASK_STRING / TARGET_STRING handling.
13065 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13066 (const rs6000_opt_masks) <"string">: Change mask to 0.
13067 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13068 (MASK_STRING): Delete.
13069 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13071 (load_multiple): Delete.
13078 (store_multiple): Delete.
13085 (movmemsi_8reg): Delete.
13086 (corresponding unnamed define_insn): Delete.
13087 (movmemsi_6reg): Delete.
13088 (corresponding unnamed define_insn): Delete.
13089 (movmemsi_4reg): Delete.
13090 (corresponding unnamed define_insn): Delete.
13091 (movmemsi_2reg): Delete.
13092 (corresponding unnamed define_insn): Delete.
13093 (movmemsi_1reg): Delete.
13094 (corresponding unnamed define_insn): Delete.
13095 * config/rs6000/rs6000.opt (mno-string): New.
13096 (mstring): Replace by deprecation warning stub.
13097 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13099 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13101 * regrename.c (regrename_do_replace): If replacing the same
13102 reg multiple times, try to reuse last created gen_raw_REG.
13105 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13106 main to workaround a bug in GDB.
13108 2018-01-12 Tom de Vries <tom@codesourcery.com>
13111 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13113 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13115 PR rtl-optimization/80481
13116 * ira-color.c (get_cap_member): New function.
13117 (allocnos_conflict_by_live_ranges_p): Use it.
13118 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13119 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13121 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13124 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13125 (*saddl_se_1): Ditto.
13126 (*ssubsi_1): Ditto.
13127 (*ssubl_se_1): Ditto.
13129 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13131 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13132 rather than wi::to_widest for DR_INITs.
13133 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13134 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13135 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13137 (vect_analyze_group_access_1): Note that here.
13139 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13141 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13142 polynomial type sizes.
13144 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13146 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13147 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13148 (gimple_add_tmp_var): Likewise.
13150 2018-01-12 Martin Liska <mliska@suse.cz>
13152 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13153 (gimple_alloc_sizes): Likewise.
13154 (dump_gimple_statistics): Use PRIu64 in printf format.
13155 * gimple.h: Change uint64_t to int.
13157 2018-01-12 Martin Liska <mliska@suse.cz>
13159 * tree-core.h: Use uint64_t instead of int.
13160 * tree.c (tree_node_counts): Likewise.
13161 (tree_node_sizes): Likewise.
13162 (dump_tree_statistics): Use PRIu64 in printf format.
13164 2018-01-12 Martin Liska <mliska@suse.cz>
13166 * Makefile.in: As qsort_chk is implemented in vec.c, add
13167 vec.o to linkage of gencfn-macros.
13168 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13169 passing the info to record_node_allocation_statistics.
13170 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13172 * ggc-common.c (struct ggc_usage): Add operator== and use
13173 it in operator< and compare function.
13174 * mem-stats.h (struct mem_usage): Likewise.
13175 * vec.c (struct vec_usage): Remove operator< and compare
13176 function. Can be simply inherited.
13178 2018-01-12 Martin Jambor <mjambor@suse.cz>
13181 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13182 * tree-ssa-math-opts.c: Include domwalk.h.
13183 (convert_mult_to_fma_1): New function.
13184 (fma_transformation_info): New type.
13185 (fma_deferring_state): Likewise.
13186 (cancel_fma_deferring): New function.
13187 (result_of_phi): Likewise.
13188 (last_fma_candidate_feeds_initial_phi): Likewise.
13189 (convert_mult_to_fma): Added deferring logic, split actual
13190 transformation to convert_mult_to_fma_1.
13191 (math_opts_dom_walker): New type.
13192 (math_opts_dom_walker::after_dom_children): New method, body moved
13193 here from pass_optimize_widening_mul::execute, added deferring logic
13195 (pass_optimize_widening_mul::execute): Moved most of code to
13196 math_opts_dom_walker::after_dom_children.
13197 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13198 * config/i386/i386.c (ix86_option_override_internal): Added
13199 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13201 2018-01-12 Richard Biener <rguenther@suse.de>
13204 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13205 inline instance vars.
13207 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13210 * config/rx/rx.c (rx_is_restricted_memory_address):
13211 Handle SUBREG case.
13213 2018-01-12 Richard Biener <rguenther@suse.de>
13215 PR tree-optimization/80846
13216 * target.def (split_reduction): New target hook.
13217 * targhooks.c (default_split_reduction): New function.
13218 * targhooks.h (default_split_reduction): Declare.
13219 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13220 target requests first reduce vectors by combining low and high
13222 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13223 (get_vectype_for_scalar_type_and_size): Export.
13224 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13225 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13226 * doc/tm.texi: Regenerate.
13227 * config/i386/i386.c (ix86_split_reduction): Implement
13228 TARGET_VECTORIZE_SPLIT_REDUCTION.
13230 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13233 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13234 in PIC mode except for TARGET_VXWORKS_RTP.
13235 * config/sparc/sparc.c: Include cfgrtl.h.
13236 (TARGET_INIT_PIC_REG): Define.
13237 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13238 (sparc_pic_register_p): New predicate.
13239 (sparc_legitimate_address_p): Use it.
13240 (sparc_legitimize_pic_address): Likewise.
13241 (sparc_delegitimize_address): Likewise.
13242 (sparc_mode_dependent_address_p): Likewise.
13243 (gen_load_pcrel_sym): Remove 4th parameter.
13244 (load_got_register): Adjust call to above. Remove obsolete stuff.
13245 (sparc_expand_prologue): Do not call load_got_register here.
13246 (sparc_flat_expand_prologue): Likewise.
13247 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13248 (sparc_use_pseudo_pic_reg): New function.
13249 (sparc_init_pic_reg): Likewise.
13250 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13251 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13253 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13255 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13256 Add item for branch_cost.
13258 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13260 PR rtl-optimization/83565
13261 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13262 not extend the result to a larger mode for rotate operations.
13263 (num_sign_bit_copies1): Likewise.
13265 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13268 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13270 Use values-Xc.o for -pedantic.
13271 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13273 2018-01-12 Martin Liska <mliska@suse.cz>
13276 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13278 (possible_polymorphic_call_targets): Use it.
13279 (ipa_devirt): Likewise.
13281 2018-01-12 Martin Liska <mliska@suse.cz>
13283 * profile-count.h (enum profile_quality): Use 0 as invalid
13284 enum value of profile_quality.
13286 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13288 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13289 -mext-string options.
13291 2018-01-12 Richard Biener <rguenther@suse.de>
13293 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13294 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13295 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13297 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13299 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13301 * configure.ac (--with-long-double-format): Add support for the
13302 configuration option to change the default long double format on
13304 * config.gcc (powerpc*-linux*-*): Likewise.
13305 * configure: Regenerate.
13306 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13307 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13308 used without modification.
13310 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13312 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13313 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13314 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13315 MISC_BUILTIN_SPEC_BARRIER.
13316 (rs6000_init_builtins): Likewise.
13317 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13319 (speculation_barrier): New define_insn.
13320 * doc/extend.texi: Document __builtin_speculation_barrier.
13322 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13325 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13326 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13327 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13329 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13330 integral modes instead of "ss" and "sd".
13331 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13332 vectors with 32-bit and 64-bit elements.
13333 (vecdupssescalarmodesuffix): New mode attribute.
13334 (vec_dup<mode>): Use it.
13336 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13339 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13340 frame if argument is passed on stack.
13342 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13345 * ree.c (combine_reaching_defs): Optimize also
13346 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13347 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13349 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13351 PR middle-end/83189
13352 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13354 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13356 PR middle-end/83718
13357 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13358 after they are computed.
13360 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13362 PR tree-optimization/83695
13363 * gimple-loop-linterchange.cc
13364 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13365 reset cached scev information after interchange.
13366 (pass_linterchange::execute): Remove call to scev_reset_htab.
13368 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13370 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13371 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13372 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13373 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13374 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13375 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13376 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13377 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13378 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13379 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13380 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13381 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13382 (V_lane_reg): Likewise.
13383 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13385 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13386 (vfmal_lane_low<mode>_intrinsic,
13387 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13388 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13389 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13390 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13391 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13392 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13394 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13396 * config/arm/arm-cpus.in (fp16fml): New feature.
13397 (ALL_SIMD): Add fp16fml.
13398 (armv8.2-a): Add fp16fml as an option.
13399 (armv8.3-a): Likewise.
13400 (armv8.4-a): Add fp16fml as part of fp16.
13401 * config/arm/arm.h (TARGET_FP16FML): Define.
13402 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13404 * config/arm/arm-modes.def (V2HF): Define.
13405 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13406 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13407 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13408 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13409 vfmsl_low, vfmsl_high): New set of builtins.
13410 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13411 (vfml_op): New code attribute.
13412 (VFMLHALVES): New int iterator.
13413 (VFML, VFMLSEL): New mode attributes.
13414 (V_reg): Define mapping for V2HF.
13415 (V_hi, V_lo): New mode attributes.
13416 (VF_constraint): Likewise.
13417 (vfml_half, vfml_half_selector): New int attributes.
13418 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13420 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13421 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13423 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13424 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13425 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13426 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13428 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13429 Document new effective target and option set.
13431 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13433 * config/arm/arm-cpus.in (armv8_4): New feature.
13434 (ARMv8_4a): New fgroup.
13435 (armv8.4-a): New arch.
13436 * config/arm/arm-tables.opt: Regenerate.
13437 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13438 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13439 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13440 Add matching rules for -march=armv8.4-a and extensions.
13441 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13443 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13446 * config/rx/rx.md (BW): New mode attribute.
13447 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13449 2018-01-11 Richard Biener <rguenther@suse.de>
13451 PR tree-optimization/83435
13452 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13453 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13454 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13456 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13457 Alan Hayward <alan.hayward@arm.com>
13458 David Sherwood <david.sherwood@arm.com>
13460 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13462 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13463 (aarch64_print_address_internal): Use it to check for a zero offset.
13465 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13466 Alan Hayward <alan.hayward@arm.com>
13467 David Sherwood <david.sherwood@arm.com>
13469 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13470 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13471 Return a poly_int64 rather than a HOST_WIDE_INT.
13472 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13473 rather than a HOST_WIDE_INT.
13474 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13475 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13476 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13477 final_offset from HOST_WIDE_INT to poly_int64.
13478 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13479 to_constant when getting the number of units in an Advanced SIMD
13481 (aarch64_builtin_vectorized_function): Check for a constant number
13483 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13485 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13486 attribute instead of GET_MODE_NUNITS.
13487 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13488 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13489 GET_MODE_SIZE for fixed-size registers.
13490 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13491 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13492 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13493 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13494 (aarch64_print_operand, aarch64_print_address_internal)
13495 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13496 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13497 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13498 Handle polynomial GET_MODE_SIZE.
13499 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13500 wider than SImode without modification.
13501 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13502 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13503 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13504 passing and returning SVE modes.
13505 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13506 rather than GEN_INT.
13507 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13508 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13509 (aarch64_allocate_and_probe_stack_space): Likewise.
13510 (aarch64_layout_frame): Cope with polynomial offsets.
13511 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13512 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13513 polynomial offsets.
13514 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13515 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13516 poly_int64 rather than a HOST_WIDE_INT.
13517 (aarch64_get_separate_components, aarch64_process_components)
13518 (aarch64_expand_prologue, aarch64_expand_epilogue)
13519 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13520 (aarch64_anchor_offset): New function, split out from...
13521 (aarch64_legitimize_address): ...here.
13522 (aarch64_builtin_vectorization_cost): Handle polynomial
13523 TYPE_VECTOR_SUBPARTS.
13524 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13526 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13527 number of elements from the PARALLEL rather than the mode.
13528 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13529 rather than GET_MODE_BITSIZE.
13530 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13531 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13532 (aarch64_expand_vec_perm_const_1): Handle polynomial
13533 d->perm.length () and d->perm elements.
13534 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13535 Apply to_constant to d->perm elements.
13536 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13537 polynomial CONST_VECTOR_NUNITS.
13538 (aarch64_move_pointer): Take amount as a poly_int64 rather
13540 (aarch64_progress_pointer): Avoid temporary variable.
13541 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13542 the mode attribute instead of GET_MODE.
13544 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13545 Alan Hayward <alan.hayward@arm.com>
13546 David Sherwood <david.sherwood@arm.com>
13548 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13549 x exists before using it.
13550 (aarch64_add_constant_internal): Rename to...
13551 (aarch64_add_offset_1): ...this. Replace regnum with separate
13552 src and dest rtxes. Handle the case in which they're different,
13553 including when the offset is zero. Replace scratchreg with an rtx.
13554 Use 2 additions if there is no spare register into which we can
13555 move a 16-bit constant.
13556 (aarch64_add_constant): Delete.
13557 (aarch64_add_offset): Replace reg with separate src and dest
13558 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13559 Use aarch64_add_offset_1.
13560 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13561 an rtx rather than an int. Take the delta as a poly_int64
13562 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13563 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13564 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13565 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13566 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13567 and aarch64_add_sp.
13568 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13569 aarch64_add_constant.
13571 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13573 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13574 Use scalar_float_mode.
13576 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13578 * config/aarch64/aarch64-simd.md
13579 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13580 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13581 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13582 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13583 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13584 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13585 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13586 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13587 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13588 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13590 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13593 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13594 targ_options->x_arm_arch_string is non NULL.
13596 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13598 * config/aarch64/aarch64.h
13599 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13601 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13604 * expmed.c (emit_store_flag_force): Swap if const op0
13605 and change VOIDmode to mode of op0.
13607 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13609 PR rtl-optimization/83761
13610 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13611 than bytes to mode_for_size.
13613 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13615 PR middle-end/83189
13616 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13617 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13620 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13622 PR middle-end/83575
13623 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13624 when in layout mode.
13625 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13626 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13629 2018-01-10 Michael Collison <michael.collison@arm.com>
13631 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13632 * config/aarch64/aarch64-option-extension.def: Add
13633 AARCH64_OPT_EXTENSION of 'fp16fml'.
13634 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13635 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13636 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13637 * config/aarch64/constraints.md (Ui7): New constraint.
13638 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13639 (VFMLA_SEL_W): Ditto.
13642 (VFMLA16_LOW): New int iterator.
13643 (VFMLA16_HIGH): Ditto.
13644 (UNSPEC_FMLAL): New unspec.
13645 (UNSPEC_FMLSL): Ditto.
13646 (UNSPEC_FMLAL2): Ditto.
13647 (UNSPEC_FMLSL2): Ditto.
13648 (f16mac): New code attribute.
13649 * config/aarch64/aarch64-simd-builtins.def
13650 (aarch64_fmlal_lowv2sf): Ditto.
13651 (aarch64_fmlsl_lowv2sf): Ditto.
13652 (aarch64_fmlalq_lowv4sf): Ditto.
13653 (aarch64_fmlslq_lowv4sf): Ditto.
13654 (aarch64_fmlal_highv2sf): Ditto.
13655 (aarch64_fmlsl_highv2sf): Ditto.
13656 (aarch64_fmlalq_highv4sf): Ditto.
13657 (aarch64_fmlslq_highv4sf): Ditto.
13658 (aarch64_fmlal_lane_lowv2sf): Ditto.
13659 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13660 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13661 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13662 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13663 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13664 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13665 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13666 (aarch64_fmlal_lane_highv2sf): Ditto.
13667 (aarch64_fmlsl_lane_highv2sf): Ditto.
13668 (aarch64_fmlal_laneq_highv2sf): Ditto.
13669 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13670 (aarch64_fmlalq_lane_highv4sf): Ditto.
13671 (aarch64_fmlsl_lane_highv4sf): Ditto.
13672 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13673 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13674 * config/aarch64/aarch64-simd.md:
13675 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13676 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13677 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13678 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13679 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13680 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13681 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13682 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13683 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13684 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13685 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13686 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13687 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13688 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13689 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13690 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13691 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13692 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13693 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13694 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13695 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13696 (vfmlsl_low_u32): Ditto.
13697 (vfmlalq_low_u32): Ditto.
13698 (vfmlslq_low_u32): Ditto.
13699 (vfmlal_high_u32): Ditto.
13700 (vfmlsl_high_u32): Ditto.
13701 (vfmlalq_high_u32): Ditto.
13702 (vfmlslq_high_u32): Ditto.
13703 (vfmlal_lane_low_u32): Ditto.
13704 (vfmlsl_lane_low_u32): Ditto.
13705 (vfmlal_laneq_low_u32): Ditto.
13706 (vfmlsl_laneq_low_u32): Ditto.
13707 (vfmlalq_lane_low_u32): Ditto.
13708 (vfmlslq_lane_low_u32): Ditto.
13709 (vfmlalq_laneq_low_u32): Ditto.
13710 (vfmlslq_laneq_low_u32): Ditto.
13711 (vfmlal_lane_high_u32): Ditto.
13712 (vfmlsl_lane_high_u32): Ditto.
13713 (vfmlal_laneq_high_u32): Ditto.
13714 (vfmlsl_laneq_high_u32): Ditto.
13715 (vfmlalq_lane_high_u32): Ditto.
13716 (vfmlslq_lane_high_u32): Ditto.
13717 (vfmlalq_laneq_high_u32): Ditto.
13718 (vfmlslq_laneq_high_u32): Ditto.
13719 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13720 (AARCH64_FL_FOR_ARCH8_4): New.
13721 (AARCH64_ISA_F16FML): New ISA flag.
13722 (TARGET_F16FML): New feature flag for fp16fml.
13723 (doc/invoke.texi): Document new fp16fml option.
13725 2018-01-10 Michael Collison <michael.collison@arm.com>
13727 * config/aarch64/aarch64-builtins.c:
13728 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13729 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13730 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13731 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13732 (AARCH64_ISA_SHA3): New ISA flag.
13733 (TARGET_SHA3): New feature flag for sha3.
13734 * config/aarch64/iterators.md (sha512_op): New int attribute.
13735 (CRYPTO_SHA512): New int iterator.
13736 (UNSPEC_SHA512H): New unspec.
13737 (UNSPEC_SHA512H2): Ditto.
13738 (UNSPEC_SHA512SU0): Ditto.
13739 (UNSPEC_SHA512SU1): Ditto.
13740 * config/aarch64/aarch64-simd-builtins.def
13741 (aarch64_crypto_sha512hqv2di): New builtin.
13742 (aarch64_crypto_sha512h2qv2di): Ditto.
13743 (aarch64_crypto_sha512su0qv2di): Ditto.
13744 (aarch64_crypto_sha512su1qv2di): Ditto.
13745 (aarch64_eor3qv8hi): Ditto.
13746 (aarch64_rax1qv2di): Ditto.
13747 (aarch64_xarqv2di): Ditto.
13748 (aarch64_bcaxqv8hi): Ditto.
13749 * config/aarch64/aarch64-simd.md:
13750 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13751 (aarch64_crypto_sha512su0qv2di): Ditto.
13752 (aarch64_crypto_sha512su1qv2di): Ditto.
13753 (aarch64_eor3qv8hi): Ditto.
13754 (aarch64_rax1qv2di): Ditto.
13755 (aarch64_xarqv2di): Ditto.
13756 (aarch64_bcaxqv8hi): Ditto.
13757 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13758 (vsha512h2q_u64): Ditto.
13759 (vsha512su0q_u64): Ditto.
13760 (vsha512su1q_u64): Ditto.
13761 (veor3q_u16): Ditto.
13762 (vrax1q_u64): Ditto.
13763 (vxarq_u64): Ditto.
13764 (vbcaxq_u16): Ditto.
13765 * config/arm/types.md (crypto_sha512): New type attribute.
13766 (crypto_sha3): Ditto.
13767 (doc/invoke.texi): Document new sha3 option.
13769 2018-01-10 Michael Collison <michael.collison@arm.com>
13771 * config/aarch64/aarch64-builtins.c:
13772 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13773 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13774 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13775 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13776 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13777 (AARCH64_ISA_SM4): New ISA flag.
13778 (TARGET_SM4): New feature flag for sm4.
13779 * config/aarch64/aarch64-simd-builtins.def
13780 (aarch64_sm3ss1qv4si): Ditto.
13781 (aarch64_sm3tt1aq4si): Ditto.
13782 (aarch64_sm3tt1bq4si): Ditto.
13783 (aarch64_sm3tt2aq4si): Ditto.
13784 (aarch64_sm3tt2bq4si): Ditto.
13785 (aarch64_sm3partw1qv4si): Ditto.
13786 (aarch64_sm3partw2qv4si): Ditto.
13787 (aarch64_sm4eqv4si): Ditto.
13788 (aarch64_sm4ekeyqv4si): Ditto.
13789 * config/aarch64/aarch64-simd.md:
13790 (aarch64_sm3ss1qv4si): Ditto.
13791 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13792 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13793 (aarch64_sm4eqv4si): Ditto.
13794 (aarch64_sm4ekeyqv4si): Ditto.
13795 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13796 (sm3part_op): Ditto.
13797 (CRYPTO_SM3TT): Ditto.
13798 (CRYPTO_SM3PART): Ditto.
13799 (UNSPEC_SM3SS1): New unspec.
13800 (UNSPEC_SM3TT1A): Ditto.
13801 (UNSPEC_SM3TT1B): Ditto.
13802 (UNSPEC_SM3TT2A): Ditto.
13803 (UNSPEC_SM3TT2B): Ditto.
13804 (UNSPEC_SM3PARTW1): Ditto.
13805 (UNSPEC_SM3PARTW2): Ditto.
13806 (UNSPEC_SM4E): Ditto.
13807 (UNSPEC_SM4EKEY): Ditto.
13808 * config/aarch64/constraints.md (Ui2): New constraint.
13809 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13810 * config/arm/types.md (crypto_sm3): New type attribute.
13811 (crypto_sm4): Ditto.
13812 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13813 (vsm3tt1aq_u32): Ditto.
13814 (vsm3tt1bq_u32): Ditto.
13815 (vsm3tt2aq_u32): Ditto.
13816 (vsm3tt2bq_u32): Ditto.
13817 (vsm3partw1q_u32): Ditto.
13818 (vsm3partw2q_u32): Ditto.
13819 (vsm4eq_u32): Ditto.
13820 (vsm4ekeyq_u32): Ditto.
13821 (doc/invoke.texi): Document new sm4 option.
13823 2018-01-10 Michael Collison <michael.collison@arm.com>
13825 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13826 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13827 (AARCH64_FL_FOR_ARCH8_4): New.
13828 (AARCH64_FL_V8_4): New flag.
13829 (doc/invoke.texi): Document new armv8.4-a option.
13831 2018-01-10 Michael Collison <michael.collison@arm.com>
13833 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13834 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13835 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13836 * config/aarch64/aarch64-option-extension.def: Add
13837 AARCH64_OPT_EXTENSION of 'sha2'.
13838 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13839 (crypto): Disable sha2 and aes if crypto disabled.
13840 (crypto): Enable aes and sha2 if enabled.
13841 (simd): Disable sha2 and aes if simd disabled.
13842 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13844 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13845 (TARGET_SHA2): New feature flag for sha2.
13846 (TARGET_AES): New feature flag for aes.
13847 * config/aarch64/aarch64-simd.md:
13848 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13849 conditional on TARGET_AES.
13850 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13851 (aarch64_crypto_sha1hsi): Make pattern conditional
13853 (aarch64_crypto_sha1hv4si): Ditto.
13854 (aarch64_be_crypto_sha1hv4si): Ditto.
13855 (aarch64_crypto_sha1su1v4si): Ditto.
13856 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13857 (aarch64_crypto_sha1su0v4si): Ditto.
13858 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13859 (aarch64_crypto_sha256su0v4si): Ditto.
13860 (aarch64_crypto_sha256su1v4si): Ditto.
13861 (doc/invoke.texi): Document new aes and sha2 options.
13863 2018-01-10 Martin Sebor <msebor@redhat.com>
13865 PR tree-optimization/83781
13866 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13869 2018-01-11 Martin Sebor <msebor@gmail.com>
13870 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13872 PR tree-optimization/83501
13873 PR tree-optimization/81703
13875 * tree-ssa-strlen.c (get_string_cst): Rename...
13876 (get_string_len): ...to this. Handle global constants.
13877 (handle_char_store): Adjust.
13879 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13880 Jim Wilson <jimw@sifive.com>
13882 * config/riscv/riscv-protos.h (riscv_output_return): New.
13883 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13884 (riscv_attribute_table, riscv_output_return),
13885 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13886 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13887 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13889 (riscv_expand_prologue): Add early return for naked function.
13890 (riscv_expand_epilogue): Likewise.
13891 (riscv_function_ok_for_sibcall): Return false for naked function.
13892 (riscv_set_current_function): New.
13893 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13894 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13895 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13896 * doc/extend.texi (RISC-V Function Attributes): New.
13898 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13900 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13901 check for 128-bit long double before checking TCmode.
13902 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13903 128-bit long doubles before checking TFmode or TCmode.
13904 (FLOAT128_IBM_P): Likewise.
13906 2018-01-10 Martin Sebor <msebor@redhat.com>
13908 PR tree-optimization/83671
13909 * builtins.c (c_strlen): Unconditionally return zero for the empty
13911 Use -Warray-bounds for warnings.
13912 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13913 for non-constant array indices with COMPONENT_REF, arrays of
13914 arrays, and pointers to arrays.
13915 (gimple_fold_builtin_strlen): Determine and set length range for
13916 non-constant character arrays.
13918 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13920 PR middle-end/81897
13921 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13924 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13926 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13928 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13931 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13932 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13933 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13934 indexed_or_indirect_operand predicate.
13935 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13936 (*vsx_le_perm_load_v8hi): Likewise.
13937 (*vsx_le_perm_load_v16qi): Likewise.
13938 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13939 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13940 (*vsx_le_perm_store_v8hi): Likewise.
13941 (*vsx_le_perm_store_v16qi): Likewise.
13942 (eight unnamed splitters): Likewise.
13944 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13946 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13947 * config/rs6000/emmintrin.h: Likewise.
13948 * config/rs6000/mmintrin.h: Likewise.
13949 * config/rs6000/xmmintrin.h: Likewise.
13951 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13954 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13956 * tree.c (tree_nop_conversion): Return true for location wrapper
13958 (maybe_wrap_with_location): New function.
13959 (selftest::check_strip_nops): New function.
13960 (selftest::test_location_wrappers): New function.
13961 (selftest::tree_c_tests): Call it.
13962 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13963 (maybe_wrap_with_location): New decl.
13964 (EXPR_LOCATION_WRAPPER_P): New macro.
13965 (location_wrapper_p): New inline function.
13966 (tree_strip_any_location_wrapper): New inline function.
13968 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13971 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13972 stack_realign_offset for the largest alignment of stack slot
13974 (ix86_find_max_used_stack_alignment): New function.
13975 (ix86_finalize_stack_frame_flags): Use it. Set
13976 max_used_stack_alignment if we don't realign stack.
13977 * config/i386/i386.h (machine_function): Add
13978 max_used_stack_alignment.
13980 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13982 * config/arm/arm.opt (-mbranch-cost): New option.
13983 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13986 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13989 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13990 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13992 2018-01-10 Richard Biener <rguenther@suse.de>
13995 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13996 early out so it also covers the case where we have a non-NULL
13999 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14001 PR tree-optimization/83753
14002 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14003 for non-strided grouped accesses if the number of elements is 1.
14005 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14008 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14009 * i386.h (TARGET_USE_GATHER): Define.
14010 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14012 2018-01-10 Martin Liska <mliska@suse.cz>
14015 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14016 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14018 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14019 CLEANUP_NO_PARTITIONING is not set.
14021 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14023 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14024 for vectors, as a partial revert of r254296.
14025 * rtl.h (const_vec_p): Delete.
14026 (const_vec_duplicate_p): Don't test for vector CONSTs.
14027 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14028 * expmed.c (make_tree): Likewise.
14031 * common.md (E, F): Use CONSTANT_P instead of checking for
14033 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14034 checking for CONST_VECTOR.
14036 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14038 PR middle-end/83575
14039 * predict.c (force_edge_cold): Handle in more sane way edges
14040 with no prediction.
14042 2018-01-09 Carl Love <cel@us.ibm.com>
14044 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14046 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14047 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14048 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14049 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14050 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14051 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14052 * config/rs6000/rs6000-protos.h: Add extern defition for
14053 rs6000_generate_float2_double_code.
14054 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14056 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14057 (float2_v2df): Add define_expand.
14059 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14062 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14063 op_mode in the force_to_mode call.
14065 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14067 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14068 instead of checking each element individually.
14069 (aarch64_evpc_uzp): Likewise.
14070 (aarch64_evpc_zip): Likewise.
14071 (aarch64_evpc_ext): Likewise.
14072 (aarch64_evpc_rev): Likewise.
14073 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14074 instead of checking each element individually. Return true without
14076 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14077 whether all selected elements come from the same input, instead of
14078 checking each element individually. Remove calls to gen_rtx_REG,
14079 start_sequence and end_sequence and instead assert that no rtl is
14082 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14084 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14085 order of HIGH and CONST checks.
14087 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14089 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14090 if the destination isn't an SSA_NAME.
14092 2018-01-09 Richard Biener <rguenther@suse.de>
14094 PR tree-optimization/83668
14095 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14097 (canonicalize_loop_form): ... here, renamed from ...
14098 (canonicalize_loop_closed_ssa_form): ... this and amended to
14099 swap successor edges for loop exit blocks to make us use
14100 the RPO order we need for initial schedule generation.
14102 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14104 PR tree-optimization/64811
14105 * match.pd: When optimizing comparisons with Inf, avoid
14106 introducing or losing exceptions from comparisons with NaN.
14108 2018-01-09 Martin Liska <mliska@suse.cz>
14111 * asan.c (shadow_mem_size): Add gcc_assert.
14113 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14115 Don't save registers in main().
14118 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14119 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14120 * config/avr/avr.c (avr_set_current_function): Don't error if
14121 naked, OS_task or OS_main are specified at the same time.
14122 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14124 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14126 * common/config/avr/avr-common.c (avr_option_optimization_table):
14127 Switch on -mmain-is-OS_task for optimizing compilations.
14129 2018-01-09 Richard Biener <rguenther@suse.de>
14131 PR tree-optimization/83572
14132 * graphite.c: Include cfganal.h.
14133 (graphite_transform_loops): Connect infinite loops to exit
14134 and remove fake edges at the end.
14136 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14138 * ipa-inline.c (edge_badness): Revert accidental checkin.
14140 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14143 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14144 symbols; not inline clones.
14146 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14149 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14150 hard registers. Formatting fixes.
14152 PR preprocessor/83722
14153 * gcc.c (try_generate_repro): Pass
14154 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14155 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14158 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14159 Kito Cheng <kito.cheng@gmail.com>
14161 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14162 (riscv_leaf_function_p): Delete.
14163 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14165 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14167 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14169 (do_ifelse): New function.
14170 (do_isel): New function.
14171 (do_sub3): New function.
14172 (do_add3): New function.
14173 (do_load_mask_compare): New function.
14174 (do_overlap_load_compare): New function.
14175 (expand_compare_loop): New function.
14176 (expand_block_compare): Call expand_compare_loop() when appropriate.
14177 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14178 option description.
14179 (-mblock-compare-inline-loop-limit): New option.
14181 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14184 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14185 Reverse order of second and third operands in first alternative.
14186 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14187 of first and second elements in UNSPEC_VPERMR vector.
14188 (altivec_expand_vec_perm_le): Likewise.
14190 2018-01-08 Jeff Law <law@redhat.com>
14192 PR rtl-optimizatin/81308
14193 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14194 (process_switch): If group_case_labels makes a change, then set
14196 (pass_convert_switch::execute): If a switch is converted, then
14197 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14199 PR rtl-optimization/81308
14200 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14203 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14205 PR target/83663 - Revert r255946
14206 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14207 generation for cases where splatting a value is not useful.
14208 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14209 across a vec_duplicate and a paradoxical subreg forming a vector
14210 mode to a vec_concat.
14212 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14214 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14215 -march=armv8.3-a variants.
14216 * config/arm/t-multilib: Likewise.
14217 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14219 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14221 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14223 (cceq_ior_compare_complement): Give it a name so I can use it, and
14224 change boolean_or_operator predicate to boolean_operator so it can
14225 be used to generate a crand.
14226 (eqne): New code iterator.
14227 (bd/bd_neg): New code_attrs.
14228 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14229 a single define_insn.
14230 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14231 decrement (bdnzt/bdnzf/bdzt/bdzf).
14232 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14233 with the new names of the branch decrement patterns, and added the
14234 names of the branch decrement conditional patterns.
14236 2018-01-08 Richard Biener <rguenther@suse.de>
14238 PR tree-optimization/83563
14239 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14242 2018-01-08 Richard Biener <rguenther@suse.de>
14244 PR middle-end/83713
14245 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14247 2018-01-08 Richard Biener <rguenther@suse.de>
14249 PR tree-optimization/83685
14250 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14251 references to abnormals.
14253 2018-01-08 Richard Biener <rguenther@suse.de>
14256 * dwarf2out.c (output_indirect_strings): Handle empty
14257 skeleton_debug_str_hash.
14258 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14260 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14262 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14263 (emit_store_direct): Likewise.
14264 (arc_trampoline_adjust_address): Likewise.
14265 (arc_asm_trampoline_template): New function.
14266 (arc_initialize_trampoline): Use asm_trampoline_template.
14267 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14268 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14269 * config/arc/arc.md (flush_icache): Delete pattern.
14271 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14273 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14274 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14277 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14280 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14281 by not USED_FOR_TARGET.
14282 (make_pass_resolve_sw_modes): Likewise.
14284 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14286 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14289 2018-01-08 Richard Biener <rguenther@suse.de>
14291 PR middle-end/83580
14292 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14294 2018-01-08 Richard Biener <rguenther@suse.de>
14296 PR middle-end/83517
14297 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14299 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14301 PR middle-end/81897
14302 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14303 basic blocks with a small number of successors.
14304 (convert_control_dep_chain_into_preds): Improve handling of
14306 (dump_predicates): Split apart into...
14307 (dump_pred_chain): ...here...
14308 (dump_pred_info): ...and here.
14309 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14310 (can_chain_union_be_invalidated_p): Improve check for invalidation
14312 (uninit_uses_cannot_happen): Avoid unnecessary if
14313 convert_control_dep_chain_into_preds yielded nothing.
14315 2018-01-06 Martin Sebor <msebor@redhat.com>
14317 PR tree-optimization/83640
14318 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14319 subtracting negative offset from size.
14320 (builtin_access::overlap): Adjust offset bounds of the access to fall
14321 within the size of the object if possible.
14323 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14325 PR rtl-optimization/83699
14326 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14327 extract_bit_field_as_subreg to cases in which the extracted
14328 value is also a vector.
14330 * lra-constraints.c (process_alt_operands): Test for the equivalence
14331 substitutions when detecting a possible reload cycle.
14333 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14336 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14337 by default if flag_selective_schedling{,2}. Formatting fixes.
14339 PR rtl-optimization/83682
14340 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14341 if it has non-VECTOR_MODE element mode.
14342 (vec_duplicate_p): Likewise.
14344 PR middle-end/83694
14345 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14346 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14348 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14351 * config/i386/i386-builtin.def
14352 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14353 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14354 Require also OPTION_MASK_ISA_AVX512F in addition to
14355 OPTION_MASK_ISA_GFNI.
14356 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14357 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14358 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14359 to OPTION_MASK_ISA_GFNI.
14360 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14361 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14362 OPTION_MASK_ISA_AVX512BW.
14363 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14364 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14365 addition to OPTION_MASK_ISA_GFNI.
14366 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14367 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14368 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14369 to OPTION_MASK_ISA_GFNI.
14370 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14371 a requirement for all ISAs rather than any of them with a few
14373 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14375 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14376 bitmasks to be enabled with 3 exceptions, instead of requiring any
14377 enabled ISA with lots of exceptions.
14378 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14379 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14380 Change avx512bw in isa attribute to avx512f.
14381 * config/i386/sgxintrin.h: Add license boilerplate.
14382 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14383 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14384 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14385 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14387 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14388 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14389 temporarily sse2 rather than sse if not enabled already.
14392 * config/i386/sse.md (VI248_VLBW): Rename to ...
14393 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14394 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14395 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14396 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14397 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14398 mode iterator instead of VI248_VLBW.
14400 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14402 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14403 (record_modified): Skip clobbers; add debug output.
14404 (param_change_prob): Use sreal frequencies.
14406 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14408 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14409 punt for user-aligned variables.
14411 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14413 * tree-chrec.c (chrec_contains_symbols): Return true for
14416 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14419 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14420 of (x|y) == x for BICS pattern.
14422 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14424 PR tree-optimization/83605
14425 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14426 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14429 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14431 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14432 * config/epiphany/rtems.h: New file.
14434 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14435 Uros Bizjak <ubizjak@gmail.com>
14438 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14439 QIreg_operand instead of register_operand predicate.
14440 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14441 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14442 comments instead of -fmitigate[-_]rop.
14444 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14447 * cgraphunit.c (symbol_table::compile): Switch to text_section
14448 before calling assembly_start debug hook.
14449 * run-rtl-passes.c (run_rtl_passes): Likewise.
14452 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14454 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14455 range_int_cst_p rather than !symbolic_range_p before calling
14456 extract_range_from_multiplicative_op_1.
14458 2018-01-04 Jeff Law <law@redhat.com>
14460 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14461 redundant test in assertion.
14463 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14465 * doc/rtl.texi: Document machine_mode wrapper classes.
14467 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14469 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14470 using tree_to_uhwi.
14472 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14474 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14475 the VEC_PERM_EXPR fold to fail.
14477 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14480 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14481 to switched_sections.
14483 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14486 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14487 test for d.testing.
14489 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14492 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14493 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14495 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14498 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14499 is BLKmode and bitpos not zero or mode change is needed.
14501 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14504 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14507 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14510 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14511 instead of MULT rtx. Update all corresponding splitters.
14512 (*saddl_se): Ditto.
14513 (*ssub<modesuffix>): Ditto.
14514 (*ssubl_se): Ditto.
14515 (*cmp_sadd_di): Update split patterns.
14516 (*cmp_sadd_si): Ditto.
14517 (*cmp_sadd_sidi): Ditto.
14518 (*cmp_ssub_di): Ditto.
14519 (*cmp_ssub_si): Ditto.
14520 (*cmp_ssub_sidi): Ditto.
14521 * config/alpha/predicates.md (const23_operand): New predicate.
14522 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14523 Look for ASHIFT, not MULT inner operand.
14524 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14526 2018-01-04 Martin Liska <mliska@suse.cz>
14528 PR gcov-profile/83669
14529 * gcov.c (output_intermediate_file): Add version to intermediate
14531 * doc/gcov.texi: Document new field 'version' in intermediate
14532 file format. Fix location of '-k' option of gcov command.
14534 2018-01-04 Martin Liska <mliska@suse.cz>
14537 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14539 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14541 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14543 2018-01-03 Martin Sebor <msebor@redhat.com>
14545 PR tree-optimization/83655
14546 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14547 checking calls with invalid arguments.
14549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14551 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14552 (vectorizable_mask_load_store): Delete.
14553 (vectorizable_call): Return false for masked loads and stores.
14554 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14555 instead of gimple_assign_rhs1.
14556 (vectorizable_load): Handle IFN_MASK_LOAD.
14557 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14563 (vectorizable_mask_load_store): ...here.
14564 (vectorizable_load): ...and here.
14566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14568 * tree-vect-stmts.c (vect_build_all_ones_mask)
14569 (vect_build_zero_merge_argument): New functions, split out from...
14570 (vectorizable_load): ...here.
14572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14574 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14576 (vectorizable_mask_load_store): ...here.
14577 (vectorizable_store): ...and here.
14579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14581 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14583 (vectorizable_mask_load_store): ...here.
14585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14587 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14588 (vect_model_store_cost): Take a vec_load_store_type instead of a
14590 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14591 (vect_model_store_cost): Take a vec_load_store_type instead of a
14593 (vectorizable_mask_load_store): Update accordingly.
14594 (vectorizable_store): Likewise.
14595 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14599 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14600 IFN_MASK_LOAD calls here rather than...
14601 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14604 Alan Hayward <alan.hayward@arm.com>
14605 David Sherwood <david.sherwood@arm.com>
14607 * expmed.c (extract_bit_field_1): For vector extracts,
14608 fall back to extract_bit_field_as_subreg if vec_extract
14611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14612 Alan Hayward <alan.hayward@arm.com>
14613 David Sherwood <david.sherwood@arm.com>
14615 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14616 they are variable or constant sized.
14617 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14618 slots for constant-sized data.
14620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14621 Alan Hayward <alan.hayward@arm.com>
14622 David Sherwood <david.sherwood@arm.com>
14624 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14625 handling COND_EXPRs with boolean comparisons, try to find a better
14626 basis for the mask type than the boolean itself.
14628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14630 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14631 is calculated and how it can be overridden.
14632 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14633 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14635 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14639 Alan Hayward <alan.hayward@arm.com>
14640 David Sherwood <david.sherwood@arm.com>
14642 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14643 Remove the mode argument.
14644 (aarch64_simd_valid_immediate): Remove the mode and inverse
14646 * config/aarch64/iterators.md (bitsize): New iterator.
14647 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14648 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14649 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14650 aarch64_simd_valid_immediate.
14651 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14652 (aarch64_reg_or_bic_imm): Likewise.
14653 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14654 with an insn_type enum and msl with a modifier_type enum.
14655 Replace element_width with a scalar_mode. Change the shift
14656 to unsigned int. Add constructors for scalar_float_mode and
14657 scalar_int_mode elements.
14658 (aarch64_vect_float_const_representable_p): Delete.
14659 (aarch64_can_const_movi_rtx_p)
14660 (aarch64_simd_scalar_immediate_valid_for_move)
14661 (aarch64_simd_make_constant): Update call to
14662 aarch64_simd_valid_immediate.
14663 (aarch64_advsimd_valid_immediate_hs): New function.
14664 (aarch64_advsimd_valid_immediate): Likewise.
14665 (aarch64_simd_valid_immediate): Remove mode and inverse
14666 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14667 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14668 and aarch64_float_const_representable_p on the result.
14669 (aarch64_output_simd_mov_immediate): Remove mode argument.
14670 Update call to aarch64_simd_valid_immediate and use of
14671 simd_immediate_info.
14672 (aarch64_output_scalar_simd_mov_immediate): Update call
14675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14676 Alan Hayward <alan.hayward@arm.com>
14677 David Sherwood <david.sherwood@arm.com>
14679 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14680 (mode_nunits): Likewise CONST_MODE_NUNITS.
14681 * machmode.def (ADJUST_NUNITS): Document.
14682 * genmodes.c (mode_data::need_nunits_adj): New field.
14683 (blank_mode): Update accordingly.
14684 (adj_nunits): New variable.
14685 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14687 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14688 listed in adj_nunits.
14689 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14690 listed in adj_nunits. Don't emit case statements for such modes.
14691 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14692 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14693 nothing if adj_nunits is nonnull.
14694 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14695 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14696 (emit_mode_fbit): Update use of print_maybe_const_decl.
14697 (emit_move_size): Likewise. Treat the array as non-const
14699 (emit_mode_adjustments): Handle adj_nunits.
14701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14703 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14704 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14705 (VECTOR_MODES): Use it.
14706 (make_vector_modes): Take the prefix as an argument.
14708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14709 Alan Hayward <alan.hayward@arm.com>
14710 David Sherwood <david.sherwood@arm.com>
14712 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14713 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14714 for MODE_VECTOR_BOOL.
14715 * machmode.def (VECTOR_BOOL_MODE): Document.
14716 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14717 (make_vector_bool_mode): New function.
14718 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14720 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14721 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14723 * stor-layout.c (int_mode_for_mode): Likewise.
14724 * tree.c (build_vector_type_for_mode): Likewise.
14725 * varasm.c (output_constant_pool_2): Likewise.
14726 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14727 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14728 for MODE_VECTOR_BOOL.
14729 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14730 of mode class checks.
14731 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14732 instead of a list of mode class checks.
14733 (expand_vector_scalar_condition): Likewise.
14734 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14737 Alan Hayward <alan.hayward@arm.com>
14738 David Sherwood <david.sherwood@arm.com>
14740 * machmode.h (mode_size): Change from unsigned short to
14742 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14743 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14744 or if measurement_type is not polynomial.
14745 (fixed_size_mode::includes_p): Check for constant-sized modes.
14746 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14747 return a poly_uint16 rather than an unsigned short.
14748 (emit_mode_size): Change the type of mode_size from unsigned short
14749 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14750 (emit_mode_adjustments): Cope with polynomial vector sizes.
14751 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14753 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14755 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14756 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14757 * caller-save.c (setup_save_areas): Likewise.
14758 (replace_reg_with_saved_mem): Likewise.
14759 * calls.c (emit_library_call_value_1): Likewise.
14760 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14761 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14762 (gen_lowpart_for_combine): Likewise.
14763 * convert.c (convert_to_integer_1): Likewise.
14764 * cse.c (equiv_constant, cse_insn): Likewise.
14765 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14766 (cselib_subst_to_values): Likewise.
14767 * dce.c (word_dce_process_block): Likewise.
14768 * df-problems.c (df_word_lr_mark_ref): Likewise.
14769 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14770 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14771 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14772 (rtl_for_decl_location): Likewise.
14773 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14774 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14775 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14776 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14777 (expand_expr_real_1): Likewise.
14778 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14779 (pad_below): Likewise.
14780 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14781 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14782 * ira.c (get_subreg_tracking_sizes): Likewise.
14783 * ira-build.c (ira_create_allocno_objects): Likewise.
14784 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14785 (ira_sort_regnos_for_alter_reg): Likewise.
14786 * ira-costs.c (record_operand_costs): Likewise.
14787 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14788 (resolve_simple_move): Likewise.
14789 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14790 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14791 (lra_constraints): Likewise.
14792 (CONST_POOL_OK_P): Reject variable-sized modes.
14793 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14794 (add_pseudo_to_slot, lra_spill): Likewise.
14795 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14796 * optabs-query.c (get_best_extraction_insn): Likewise.
14797 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14798 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14799 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14800 * recog.c (offsettable_address_addr_space_p): Likewise.
14801 * regcprop.c (maybe_mode_change): Likewise.
14802 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14803 * regrename.c (build_def_use): Likewise.
14804 * regstat.c (dump_reg_info): Likewise.
14805 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14806 (find_reloads, find_reloads_subreg_address): Likewise.
14807 * reload1.c (eliminate_regs_1): Likewise.
14808 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14809 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14810 (simplify_binary_operation_1, simplify_subreg): Likewise.
14811 * targhooks.c (default_function_arg_padding): Likewise.
14812 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14813 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14814 (verify_gimple_assign_ternary): Likewise.
14815 * tree-inline.c (estimate_move_cost): Likewise.
14816 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14817 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14818 (get_address_cost_ainc): Likewise.
14819 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14820 (vect_supportable_dr_alignment): Likewise.
14821 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14822 (vectorizable_reduction): Likewise.
14823 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14824 (vectorizable_operation, vectorizable_load): Likewise.
14825 * tree.c (build_same_sized_truth_vector_type): Likewise.
14826 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14827 * var-tracking.c (emit_note_insn_var_location): Likewise.
14828 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14829 (ADDR_VEC_ALIGN): Likewise.
14831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14832 Alan Hayward <alan.hayward@arm.com>
14833 David Sherwood <david.sherwood@arm.com>
14835 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14837 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14838 or if measurement_type is polynomial.
14839 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14840 * combine.c (make_extraction): Likewise.
14841 * dse.c (find_shift_sequence): Likewise.
14842 * dwarf2out.c (mem_loc_descriptor): Likewise.
14843 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14844 (extract_bit_field, extract_low_bits): Likewise.
14845 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14846 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14847 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14848 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14849 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14850 * reload.c (find_reloads): Likewise.
14851 * reload1.c (alter_reg): Likewise.
14852 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14853 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14854 * tree-if-conv.c (predicate_mem_writes): Likewise.
14855 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14856 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14857 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14858 * valtrack.c (dead_debug_insert_temp): Likewise.
14859 * varasm.c (mergeable_constant_section): Likewise.
14860 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14863 Alan Hayward <alan.hayward@arm.com>
14864 David Sherwood <david.sherwood@arm.com>
14866 * expr.c (expand_assignment): Cope with polynomial mode sizes
14867 when assigning to a CONCAT.
14869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14870 Alan Hayward <alan.hayward@arm.com>
14871 David Sherwood <david.sherwood@arm.com>
14873 * machmode.h (mode_precision): Change from unsigned short to
14875 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14877 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14878 or if measurement_type is not polynomial.
14879 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14880 in which the mode is already known to be a scalar_int_mode.
14881 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14882 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14884 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14885 for GET_MODE_PRECISION.
14886 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14887 for GET_MODE_PRECISION.
14888 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14890 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14891 (expand_field_assignment, make_extraction): Likewise.
14892 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14893 (get_last_value): Likewise.
14894 * convert.c (convert_to_integer_1): Likewise.
14895 * cse.c (cse_insn): Likewise.
14896 * expr.c (expand_expr_real_1): Likewise.
14897 * lra-constraints.c (simplify_operand_subreg): Likewise.
14898 * optabs-query.c (can_atomic_load_p): Likewise.
14899 * optabs.c (expand_atomic_load): Likewise.
14900 (expand_atomic_store): Likewise.
14901 * ree.c (combine_reaching_defs): Likewise.
14902 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14903 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14904 * tree.h (type_has_mode_precision_p): Likewise.
14905 * ubsan.c (instrument_si_overflow): Likewise.
14907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14908 Alan Hayward <alan.hayward@arm.com>
14909 David Sherwood <david.sherwood@arm.com>
14911 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14912 polynomial numbers of units.
14913 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14914 (valid_vector_subparts_p): New function.
14915 (build_vector_type): Remove temporary shim and take the number
14916 of units as a poly_uint64 rather than an int.
14917 (build_opaque_vector_type): Take the number of units as a
14918 poly_uint64 rather than an int.
14919 * tree.c (build_vector_from_ctor): Handle polynomial
14920 TYPE_VECTOR_SUBPARTS.
14921 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14922 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14923 (build_vector_from_val): If the number of units is variable,
14924 use build_vec_duplicate_cst for constant operands and
14925 VEC_DUPLICATE_EXPR otherwise.
14926 (make_vector_type): Remove temporary is_constant ().
14927 (build_vector_type, build_opaque_vector_type): Take the number of
14928 units as a poly_uint64 rather than an int.
14929 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14931 * cfgexpand.c (expand_debug_expr): Likewise.
14932 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14933 (store_constructor, expand_expr_real_1): Likewise.
14934 (const_scalar_mask_from_tree): Likewise.
14935 * fold-const-call.c (fold_const_reduction): Likewise.
14936 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14937 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14938 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14939 (fold_relational_const): Likewise.
14940 (native_interpret_vector): Likewise. Change the size from an
14941 int to an unsigned int.
14942 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14943 TYPE_VECTOR_SUBPARTS.
14944 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14945 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14946 duplicating a non-constant operand into a variable-length vector.
14947 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14948 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14949 * ipa-icf.c (sem_variable::equals): Likewise.
14950 * match.pd: Likewise.
14951 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14952 * print-tree.c (print_node): Likewise.
14953 * stor-layout.c (layout_type): Likewise.
14954 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14955 * tree-cfg.c (verify_gimple_comparison): Likewise.
14956 (verify_gimple_assign_binary): Likewise.
14957 (verify_gimple_assign_ternary): Likewise.
14958 (verify_gimple_assign_single): Likewise.
14959 * tree-pretty-print.c (dump_generic_node): Likewise.
14960 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14961 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14962 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14963 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14964 (vect_shift_permute_load_chain): Likewise.
14965 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14966 (expand_vector_condition, optimize_vector_constructor): Likewise.
14967 (lower_vec_perm, get_compute_type): Likewise.
14968 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14969 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14970 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14971 (vect_recog_mask_conversion_pattern): Likewise.
14972 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14973 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14974 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14975 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14976 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14977 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14978 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14979 (supportable_widening_operation): Likewise.
14980 (supportable_narrowing_operation): Likewise.
14981 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14983 * varasm.c (output_constant): Likewise.
14985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14986 Alan Hayward <alan.hayward@arm.com>
14987 David Sherwood <david.sherwood@arm.com>
14989 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14990 so that both the length == 3 and length != 3 cases set up their
14991 own permute vectors. Add comments explaining why we know the
14992 number of elements is constant.
14993 (vect_permute_load_chain): Likewise.
14995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14996 Alan Hayward <alan.hayward@arm.com>
14997 David Sherwood <david.sherwood@arm.com>
14999 * machmode.h (mode_nunits): Change from unsigned char to
15001 (ONLY_FIXED_SIZE_MODES): New macro.
15002 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15003 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15004 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15006 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15007 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15008 or if measurement_type is not polynomial.
15009 * genmodes.c (ZERO_COEFFS): New macro.
15010 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15012 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15013 Use ZERO_COEFFS when emitting initializers.
15014 * data-streamer.h (bp_pack_poly_value): New function.
15015 (bp_unpack_poly_value): Likewise.
15016 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15017 for GET_MODE_NUNITS.
15018 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15019 for GET_MODE_NUNITS.
15020 * tree.c (make_vector_type): Remove temporary shim and make
15021 the real function take the number of units as a poly_uint64
15022 rather than an int.
15023 (build_vector_type_for_mode): Handle polynomial nunits.
15024 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15025 * emit-rtl.c (const_vec_series_p_1): Likewise.
15026 (gen_rtx_CONST_VECTOR): Likewise.
15027 * fold-const.c (test_vec_duplicate_folding): Likewise.
15028 * genrecog.c (validate_pattern): Likewise.
15029 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15030 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15031 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15032 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15033 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15034 * rtlanal.c (subreg_get_info): Likewise.
15035 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15036 (vect_grouped_load_supported): Likewise.
15037 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15038 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15039 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15040 (simplify_const_unary_operation, simplify_binary_operation_1)
15041 (simplify_const_binary_operation, simplify_ternary_operation)
15042 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15043 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15044 instead of CONST_VECTOR_NUNITS.
15045 * varasm.c (output_constant_pool_2): Likewise.
15046 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15047 explicit-encoded elements in the XVEC for variable-length vectors.
15049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15051 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15054 Alan Hayward <alan.hayward@arm.com>
15055 David Sherwood <david.sherwood@arm.com>
15057 * coretypes.h (fixed_size_mode): Declare.
15058 (fixed_size_mode_pod): New typedef.
15059 * builtins.h (target_builtins::x_apply_args_mode)
15060 (target_builtins::x_apply_result_mode): Change type to
15061 fixed_size_mode_pod.
15062 * builtins.c (apply_args_size, apply_result_size, result_vector)
15063 (expand_builtin_apply_args_1, expand_builtin_apply)
15064 (expand_builtin_return): Update accordingly.
15066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15068 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15069 * cselib.c (cselib_hash_rtx): Likewise.
15070 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15071 CONST_VECTOR encoding.
15073 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15074 Jeff Law <law@redhat.com>
15077 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15078 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15079 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15080 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15083 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15084 explicitly probe *sp in a noreturn function if there were any callee
15085 register saves or frame pointer is needed.
15087 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15090 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15091 BLKmode for ternary, binary or unary expressions.
15094 * var-tracking.c (delete_vta_debug_insn): New inline function.
15095 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15096 insns from get_insns () to NULL instead of each bb separately.
15097 Use delete_vta_debug_insn. No longer static.
15098 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15099 delete_vta_debug_insns callers.
15100 * rtl.h (delete_vta_debug_insns): Declare.
15101 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15102 instead of variable_tracking_main.
15104 2018-01-03 Martin Sebor <msebor@redhat.com>
15106 PR tree-optimization/83603
15107 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15108 arguments past the endof the argument list in functions declared
15109 without a prototype.
15110 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15111 Avoid checking when arguments are null.
15113 2018-01-03 Martin Sebor <msebor@redhat.com>
15116 * doc/extend.texi (attribute const): Fix a typo.
15117 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15118 issuing -Wsuggest-attribute for void functions.
15120 2018-01-03 Martin Sebor <msebor@redhat.com>
15122 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15123 offset_int::from instead of wide_int::to_shwi.
15124 (maybe_diag_overlap): Remove assertion.
15125 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15126 * gimple-ssa-sprintf.c (format_directive): Same.
15127 (parse_directive): Same.
15128 (sprintf_dom_walker::compute_format_length): Same.
15129 (try_substitute_return_value): Same.
15131 2018-01-03 Jeff Law <law@redhat.com>
15133 PR middle-end/83654
15134 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15135 non-constant residual for zero at runtime and avoid probing in
15136 that case. Reorganize code for trailing problem to mirror handling
15139 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15141 PR tree-optimization/83501
15142 * tree-ssa-strlen.c (get_string_cst): New.
15143 (handle_char_store): Call get_string_cst.
15145 2018-01-03 Martin Liska <mliska@suse.cz>
15147 PR tree-optimization/83593
15148 * tree-ssa-strlen.c: Include tree-cfg.h.
15149 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15150 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15151 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15153 (strlen_dom_walker::before_dom_children): Call
15154 gimple_purge_dead_eh_edges. Dump tranformation with details
15156 (strlen_dom_walker::before_dom_children): Update call by adding
15157 new argument cleanup_eh.
15158 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15160 2018-01-03 Martin Liska <mliska@suse.cz>
15163 * cif-code.def (VARIADIC_THUNK): New enum value.
15164 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15167 2018-01-03 Jan Beulich <jbeulich@suse.com>
15169 * sse.md (mov<mode>_internal): Tighten condition for when to use
15170 vmovdqu<ssescalarsize> for TI and OI modes.
15172 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15174 Update copyright years.
15176 2018-01-03 Martin Liska <mliska@suse.cz>
15179 * ipa-visibility.c (function_and_variable_visibility): Skip
15180 functions with noipa attribure.
15182 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15184 * gcc.c (process_command): Update copyright notice dates.
15185 * gcov-dump.c (print_version): Ditto.
15186 * gcov.c (print_version): Ditto.
15187 * gcov-tool.c (print_version): Ditto.
15188 * gengtype.c (create_file): Ditto.
15189 * doc/cpp.texi: Bump @copying's copyright year.
15190 * doc/cppinternals.texi: Ditto.
15191 * doc/gcc.texi: Ditto.
15192 * doc/gccint.texi: Ditto.
15193 * doc/gcov.texi: Ditto.
15194 * doc/install.texi: Ditto.
15195 * doc/invoke.texi: Ditto.
15197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15199 * vector-builder.h (vector_builder::m_full_nelts): Change from
15200 unsigned int to poly_uint64.
15201 (vector_builder::full_nelts): Update prototype accordingly.
15202 (vector_builder::new_vector): Likewise.
15203 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15204 (vector_builder::operator ==): Likewise.
15205 (vector_builder::finalize): Likewise.
15206 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15207 Take the number of elements as a poly_uint64 rather than an
15209 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15210 from unsigned int to poly_uint64.
15211 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15212 (vec_perm_indices::new_vector): Likewise.
15213 (vec_perm_indices::length): Likewise.
15214 (vec_perm_indices::nelts_per_input): Likewise.
15215 (vec_perm_indices::input_nelts): Likewise.
15216 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15217 number of elements per input as a poly_uint64 rather than an
15218 unsigned int. Use the original encoding for variable-length
15219 vectors, rather than clamping each individual element.
15220 For the second and subsequent elements in each pattern,
15221 clamp the step and base before clamping their sum.
15222 (vec_perm_indices::series_p): Handle polynomial element counts.
15223 (vec_perm_indices::all_in_range_p): Likewise.
15224 (vec_perm_indices_to_tree): Likewise.
15225 (vec_perm_indices_to_rtx): Likewise.
15226 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15227 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15228 (tree_vector_builder::new_binary_operation): Handle polynomial
15229 element counts. Return false if we need to know the number
15230 of elements at compile time.
15231 * fold-const.c (fold_vec_perm): Punt if the number of elements
15232 isn't known at compile time.
15234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15236 * vec-perm-indices.h (vec_perm_builder): Change element type
15237 from HOST_WIDE_INT to poly_int64.
15238 (vec_perm_indices::element_type): Update accordingly.
15239 (vec_perm_indices::clamp): Handle polynomial element_types.
15240 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15241 (vec_perm_indices::all_in_range_p): Likewise.
15242 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15244 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15245 polynomial vec_perm_indices element types.
15246 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15247 * fold-const.c (fold_vec_perm): Likewise.
15248 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15249 * tree-vect-generic.c (lower_vec_perm): Likewise.
15250 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15251 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15252 element type to HOST_WIDE_INT.
15254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15255 Alan Hayward <alan.hayward@arm.com>
15256 David Sherwood <david.sherwood@arm.com>
15258 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15259 rather than an int. Use plus_constant.
15260 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15261 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15264 Alan Hayward <alan.hayward@arm.com>
15265 David Sherwood <david.sherwood@arm.com>
15267 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15268 a HOST_WIDE_INT to a poly_int64.
15270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15271 Alan Hayward <alan.hayward@arm.com>
15272 David Sherwood <david.sherwood@arm.com>
15274 * calls.c (load_register_parameters): Cope with polynomial
15275 mode sizes. Require a constant size for BLKmode parameters
15276 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15277 forces a parameter to be padded at the lsb end in order to
15278 fill a complete number of words, require the parameter size
15279 to be ordered wrt UNITS_PER_WORD.
15281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15282 Alan Hayward <alan.hayward@arm.com>
15283 David Sherwood <david.sherwood@arm.com>
15285 * reload1.c (spill_stack_slot_width): Change element type
15286 from unsigned int to poly_uint64_pod.
15287 (alter_reg): Treat mode sizes as polynomial.
15289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15290 Alan Hayward <alan.hayward@arm.com>
15291 David Sherwood <david.sherwood@arm.com>
15293 * reload.c (complex_word_subreg_p): New function.
15294 (reload_inner_reg_of_subreg, push_reload): Use it.
15296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15297 Alan Hayward <alan.hayward@arm.com>
15298 David Sherwood <david.sherwood@arm.com>
15300 * lra-constraints.c (process_alt_operands): Reject matched
15301 operands whose sizes aren't ordered.
15302 (match_reload): Refer to this check here.
15304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15305 Alan Hayward <alan.hayward@arm.com>
15306 David Sherwood <david.sherwood@arm.com>
15308 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15309 that the mode size is in the set {1, 2, 4, 8, 16}.
15311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15312 Alan Hayward <alan.hayward@arm.com>
15313 David Sherwood <david.sherwood@arm.com>
15315 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15316 Use plus_constant instead of gen_rtx_PLUS.
15318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15319 Alan Hayward <alan.hayward@arm.com>
15320 David Sherwood <david.sherwood@arm.com>
15322 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15323 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15324 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15325 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15326 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15327 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15328 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15329 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15330 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15331 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15333 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15334 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15335 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15336 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15337 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15338 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15339 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15340 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15341 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15342 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15344 * expr.c (emit_move_resolve_push): Treat the input and result
15345 of PUSH_ROUNDING as a poly_int64.
15346 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15347 (emit_push_insn): Likewise.
15348 * lra-eliminations.c (mark_not_eliminable): Likewise.
15349 * recog.c (push_operand): Likewise.
15350 * reload1.c (elimination_effects): Likewise.
15351 * rtlanal.c (nonzero_bits1): Likewise.
15352 * calls.c (store_one_arg): Likewise. Require the padding to be
15353 known at compile time.
15355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15356 Alan Hayward <alan.hayward@arm.com>
15357 David Sherwood <david.sherwood@arm.com>
15359 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15360 Use plus_constant instead of gen_rtx_PLUS.
15362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15363 Alan Hayward <alan.hayward@arm.com>
15364 David Sherwood <david.sherwood@arm.com>
15366 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15367 rather than an int.
15369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15370 Alan Hayward <alan.hayward@arm.com>
15371 David Sherwood <david.sherwood@arm.com>
15373 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15374 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15375 via stack temporaries. Treat the mode size as polynomial too.
15377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15378 Alan Hayward <alan.hayward@arm.com>
15379 David Sherwood <david.sherwood@arm.com>
15381 * expr.c (expand_expr_real_2): When handling conversions involving
15382 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15383 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15384 as a poly_uint64 too.
15386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15387 Alan Hayward <alan.hayward@arm.com>
15388 David Sherwood <david.sherwood@arm.com>
15390 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15393 Alan Hayward <alan.hayward@arm.com>
15394 David Sherwood <david.sherwood@arm.com>
15396 * combine.c (can_change_dest_mode): Handle polynomial
15397 REGMODE_NATURAL_SIZE.
15398 * expmed.c (store_bit_field_1): Likewise.
15399 * expr.c (store_constructor): Likewise.
15400 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15401 and polynomial REGMODE_NATURAL_SIZE.
15402 (gen_lowpart_common): Likewise.
15403 * reginfo.c (record_subregs_of_mode): Likewise.
15404 * rtlanal.c (read_modify_subreg_p): Likewise.
15406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15407 Alan Hayward <alan.hayward@arm.com>
15408 David Sherwood <david.sherwood@arm.com>
15410 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15411 numbers of elements.
15413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15414 Alan Hayward <alan.hayward@arm.com>
15415 David Sherwood <david.sherwood@arm.com>
15417 * match.pd: Cope with polynomial numbers of vector elements.
15419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15420 Alan Hayward <alan.hayward@arm.com>
15421 David Sherwood <david.sherwood@arm.com>
15423 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15424 in a POINTER_PLUS_EXPR.
15426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15427 Alan Hayward <alan.hayward@arm.com>
15428 David Sherwood <david.sherwood@arm.com>
15430 * omp-simd-clone.c (simd_clone_subparts): New function.
15431 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15432 (ipa_simd_modify_function_body): Likewise.
15434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15435 Alan Hayward <alan.hayward@arm.com>
15436 David Sherwood <david.sherwood@arm.com>
15438 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15439 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15440 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15441 (expand_vector_condition, vector_element): Likewise.
15442 (subparts_gt): New function.
15443 (get_compute_type): Use subparts_gt.
15444 (count_type_subparts): Delete.
15445 (expand_vector_operations_1): Use subparts_gt instead of
15446 count_type_subparts.
15448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15449 Alan Hayward <alan.hayward@arm.com>
15450 David Sherwood <david.sherwood@arm.com>
15452 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15453 (vect_compile_time_alias): ...this new function. Do the calculation
15454 on poly_ints rather than trees.
15455 (vect_prune_runtime_alias_test_list): Update call accordingly.
15457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15458 Alan Hayward <alan.hayward@arm.com>
15459 David Sherwood <david.sherwood@arm.com>
15461 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15463 (vect_schedule_slp_instance): Likewise.
15465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15466 Alan Hayward <alan.hayward@arm.com>
15467 David Sherwood <david.sherwood@arm.com>
15469 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15470 constant and extern definitions for variable-length vectors.
15471 (vect_get_constant_vectors): Note that the number of units
15472 is known to be constant.
15474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15475 Alan Hayward <alan.hayward@arm.com>
15476 David Sherwood <david.sherwood@arm.com>
15478 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15479 of units as polynomial. Choose between WIDE and NARROW based
15482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15483 Alan Hayward <alan.hayward@arm.com>
15484 David Sherwood <david.sherwood@arm.com>
15486 * tree-vect-stmts.c (simd_clone_subparts): New function.
15487 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15490 Alan Hayward <alan.hayward@arm.com>
15491 David Sherwood <david.sherwood@arm.com>
15493 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15494 vectors as polynomial. Use build_index_vector for
15495 IFN_GOMP_SIMD_LANE.
15497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15498 Alan Hayward <alan.hayward@arm.com>
15499 David Sherwood <david.sherwood@arm.com>
15501 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15502 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15503 for variable-length vectors.
15504 (vectorizable_mask_load_store): Treat the number of units as
15505 polynomial, asserting that it is constant if the condition has
15506 already been enforced.
15507 (vectorizable_store, vectorizable_load): Likewise.
15509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15510 Alan Hayward <alan.hayward@arm.com>
15511 David Sherwood <david.sherwood@arm.com>
15513 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15514 of units as polynomial. Punt if we can't tell at compile time
15515 which vector contains the final result.
15517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15518 Alan Hayward <alan.hayward@arm.com>
15519 David Sherwood <david.sherwood@arm.com>
15521 * tree-vect-loop.c (vectorizable_induction): Treat the number
15522 of units as polynomial. Punt on SLP inductions. Use an integer
15523 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15524 cast of such a series for variable-length floating-point
15527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15528 Alan Hayward <alan.hayward@arm.com>
15529 David Sherwood <david.sherwood@arm.com>
15531 * tree.h (build_index_vector): Declare.
15532 * tree.c (build_index_vector): New function.
15533 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15534 of units as polynomial, forcibly converting it to a constant if
15535 vectorizable_reduction has already enforced the condition.
15536 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15537 to create a {1,2,3,...} vector.
15538 (vectorizable_reduction): Treat the number of units as polynomial.
15539 Choose vectype_in based on the largest scalar element size rather
15540 than the smallest number of units. Enforce the restrictions
15543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15544 Alan Hayward <alan.hayward@arm.com>
15545 David Sherwood <david.sherwood@arm.com>
15547 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15548 number of units as polynomial.
15550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15551 Alan Hayward <alan.hayward@arm.com>
15552 David Sherwood <david.sherwood@arm.com>
15554 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15555 * target.def (autovectorize_vector_sizes): Return the vector sizes
15556 by pointer, using vector_sizes rather than a bitmask.
15557 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15558 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15559 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15561 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15562 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15563 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15564 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15565 * omp-general.c (omp_max_vf): Likewise.
15566 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15567 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15568 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15569 * tree-vect-slp.c (vect_slp_bb): Likewise.
15570 * doc/tm.texi: Regenerate.
15571 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15573 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15574 the vector size as a poly_uint64 rather than an unsigned int.
15575 (current_vector_size): Change from an unsigned int to a poly_uint64.
15576 (get_vectype_for_scalar_type): Update accordingly.
15577 * tree.h (build_truth_vector_type): Take the size and number of
15578 units as a poly_uint64 rather than an unsigned int.
15579 (build_vector_type): Add a temporary overload that takes
15580 the number of units as a poly_uint64 rather than an unsigned int.
15581 * tree.c (make_vector_type): Likewise.
15582 (build_truth_vector_type): Take the number of units as a poly_uint64
15583 rather than an unsigned int.
15585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15586 Alan Hayward <alan.hayward@arm.com>
15587 David Sherwood <david.sherwood@arm.com>
15589 * target.def (get_mask_mode): Take the number of units and length
15590 as poly_uint64s rather than unsigned ints.
15591 * targhooks.h (default_get_mask_mode): Update accordingly.
15592 * targhooks.c (default_get_mask_mode): Likewise.
15593 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15594 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15601 * omp-general.c (omp_max_vf): Likewise.
15602 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15603 (expand_omp_simd): Handle polynomial safelen.
15604 * omp-low.c (omplow_simd_context): Add a default constructor.
15605 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15606 (lower_rec_simd_input_clauses): Update accordingly.
15607 (lower_rec_input_clauses): Likewise.
15609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15610 Alan Hayward <alan.hayward@arm.com>
15611 David Sherwood <david.sherwood@arm.com>
15613 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15614 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15615 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15616 (vect_analyze_slp_cost): Likewise.
15617 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15618 (vect_model_load_cost): Likewise.
15620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15621 Alan Hayward <alan.hayward@arm.com>
15622 David Sherwood <david.sherwood@arm.com>
15624 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15625 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15626 from an unsigned int * to a poly_uint64_pod *.
15627 (calculate_unrolling_factor): New function.
15628 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15631 Alan Hayward <alan.hayward@arm.com>
15632 David Sherwood <david.sherwood@arm.com>
15634 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15635 from an unsigned int to a poly_uint64.
15636 (_loop_vec_info::slp_unrolling_factor): Likewise.
15637 (_loop_vec_info::vectorization_factor): Change from an int
15639 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15640 (vect_get_num_vectors): New function.
15641 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15642 (vect_get_num_copies): Use vect_get_num_vectors.
15643 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15644 to an unsigned int *.
15645 (vect_analyze_data_refs): Change min_vf from an int * to a
15647 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15648 than an unsigned HOST_WIDE_INT.
15649 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15650 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15651 to an unsigned int *.
15652 (vect_analyze_data_ref_dependences): Likewise.
15653 (vect_compute_data_ref_alignment): Handle polynomial vf.
15654 (vect_enhance_data_refs_alignment): Likewise.
15655 (vect_prune_runtime_alias_test_list): Likewise.
15656 (vect_shift_permute_load_chain): Likewise.
15657 (vect_supportable_dr_alignment): Likewise.
15658 (dependence_distance_ge_vf): Take the vectorization factor as a
15659 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15660 (vect_analyze_data_refs): Change min_vf from an int * to a
15662 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15663 vfm1 as a poly_uint64 rather than an int. Make the same change
15664 for the returned bound_scalar.
15665 (vect_gen_vector_loop_niters): Handle polynomial vf.
15666 (vect_do_peeling): Likewise. Update call to
15667 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15668 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15670 * tree-vect-loop.c (vect_determine_vectorization_factor)
15671 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15672 (vect_get_known_peeling_cost): Likewise.
15673 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15674 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15675 (vect_transform_loop): Likewise. Use the lowest possible VF when
15676 updating the upper bounds of the loop.
15677 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15678 rather than an int.
15679 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15680 polynomial unroll factors.
15681 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15682 (vect_make_slp_decision): Likewise.
15683 (vect_supported_load_permutation_p): Likewise, and polynomial
15685 (vect_analyze_slp_cost): Handle polynomial vf.
15686 (vect_slp_analyze_node_operations): Likewise.
15687 (vect_slp_analyze_bb_1): Likewise.
15688 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15689 than an unsigned HOST_WIDE_INT.
15690 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15691 (vectorizable_load): Handle polynomial vf.
15692 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15694 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15697 Alan Hayward <alan.hayward@arm.com>
15698 David Sherwood <david.sherwood@arm.com>
15700 * match.pd: Handle bit operations involving three constants
15701 and try to fold one pair.
15703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15705 * tree-vect-loop-manip.c: Include gimple-fold.h.
15706 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15707 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15708 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15709 Add a path that uses a step of VF instead of 1, but disable it
15711 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15712 and niters_no_overflow parameters. Update calls to
15713 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15714 Create a new SSA name if the latter choses to use a ste other
15715 than zero, and return it via niters_vector_mult_vf_var.
15716 * tree-vect-loop.c (vect_transform_loop): Update calls to
15717 vect_do_peeling, vect_gen_vector_loop_niters and
15718 slpeel_make_loop_iterate_ntimes.
15719 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15720 (vect_gen_vector_loop_niters): Update declarations after above changes.
15722 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15724 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15725 128-bit round to integer instructions.
15726 (ceil<mode>2): Likewise.
15727 (btrunc<mode>2): Likewise.
15728 (round<mode>2): Likewise.
15730 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15732 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15733 unaligned VSX load/store on P8/P9.
15734 (expand_block_clear): Allow the use of unaligned VSX
15735 load/store on P8/P9.
15737 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15739 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15741 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15742 swap associated with both a load and a store.
15744 2018-01-02 Andrew Waterman <andrew@sifive.com>
15746 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15747 * config/riscv/riscv.md (clear_cache): Use it.
15749 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15751 * web.c: Remove out-of-date comment.
15753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15755 * expr.c (fixup_args_size_notes): Check that any existing
15756 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15757 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15758 (emit_single_push_insn): ...here.
15760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15762 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15763 (const_vector_encoded_nelts): New function.
15764 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15765 (const_vector_int_elt, const_vector_elt): Declare.
15766 * emit-rtl.c (const_vector_int_elt_1): New function.
15767 (const_vector_elt): Likewise.
15768 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15769 of CONST_VECTOR_ELT.
15771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15773 * expr.c: Include rtx-vector-builder.h.
15774 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15775 directly on the tree encoding.
15776 (const_vector_from_tree): Likewise.
15777 * optabs.c: Include rtx-vector-builder.h.
15778 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15779 sequence of "u" values.
15780 * vec-perm-indices.c: Include rtx-vector-builder.h.
15781 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15782 directly on the vec_perm_indices encoding.
15784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15786 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15787 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15788 * rtx-vector-builder.h: New file.
15789 * rtx-vector-builder.c: Likewise.
15790 * rtl.h (rtx_def::u2): Add a const_vector field.
15791 (CONST_VECTOR_NPATTERNS): New macro.
15792 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15793 (CONST_VECTOR_DUPLICATE_P): Likewise.
15794 (CONST_VECTOR_STEPPED_P): Likewise.
15795 (CONST_VECTOR_ENCODED_ELT): Likewise.
15796 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15797 (unwrap_const_vec_duplicate): Likewise.
15798 (const_vec_series_p): Check for a non-duplicated vector encoding.
15799 Say that the function only returns true for integer vectors.
15800 * emit-rtl.c: Include rtx-vector-builder.h.
15801 (gen_const_vec_duplicate_1): Delete.
15802 (gen_const_vector): Call gen_const_vec_duplicate instead of
15803 gen_const_vec_duplicate_1.
15804 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15805 (gen_const_vec_duplicate): Use rtx_vector_builder.
15806 (gen_const_vec_series): Likewise.
15807 (gen_rtx_CONST_VECTOR): Likewise.
15808 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15809 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15810 Build a new vector rather than modifying a CONST_VECTOR in-place.
15811 (handle_special_swappables): Update call accordingly.
15812 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15813 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15814 Build a new vector rather than modifying a CONST_VECTOR in-place.
15815 (handle_special_swappables): Update call accordingly.
15817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15819 * simplify-rtx.c (simplify_const_binary_operation): Use
15820 CONST_VECTOR_ELT instead of XVECEXP.
15822 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15824 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15825 the selector elements to be different from the data elements
15826 if the selector is a VECTOR_CST.
15827 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15828 ssizetype for the selector.
15830 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15832 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15833 before testing each element individually.
15834 * tree-vect-generic.c (lower_vec_perm): Likewise.
15836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15838 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15839 * selftest-run-tests.c (selftest::run_tests): Call it.
15840 * vector-builder.h (vector_builder::operator ==): New function.
15841 (vector_builder::operator !=): Likewise.
15842 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15843 (vec_perm_indices::all_from_input_p): New function.
15844 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15845 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15846 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15847 instead of reading the VECTOR_CST directly. Detect whether both
15848 vector inputs are the same before constructing the vec_perm_indices,
15849 and update the number of inputs argument accordingly. Use the
15850 utility functions added above. Only construct sel2 if we need to.
15852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15854 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15855 the broadcast of the low byte.
15856 (expand_mult_highpart): Use an explicit encoding for the permutes.
15857 * optabs-query.c (can_mult_highpart_p): Likewise.
15858 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15859 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15860 (vectorizable_bswap): Likewise.
15861 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15862 explicit encoding for the power-of-2 permutes.
15863 (vect_permute_store_chain): Likewise.
15864 (vect_grouped_load_supported): Likewise.
15865 (vect_permute_load_chain): Likewise.
15867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15869 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15870 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15871 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15872 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15873 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15874 (vect_gen_perm_mask_any): Likewise.
15876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15878 * int-vector-builder.h: New file.
15879 * vec-perm-indices.h: Include int-vector-builder.h.
15880 (vec_perm_indices): Redefine as an int_vector_builder.
15881 (auto_vec_perm_indices): Delete.
15882 (vec_perm_builder): Redefine as a stand-alone class.
15883 (vec_perm_indices::vec_perm_indices): New function.
15884 (vec_perm_indices::clamp): Likewise.
15885 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15886 (vec_perm_indices::new_vector): New function.
15887 (vec_perm_indices::new_expanded_vector): Update for new
15888 vec_perm_indices class.
15889 (vec_perm_indices::rotate_inputs): New function.
15890 (vec_perm_indices::all_in_range_p): Operate directly on the
15891 encoded form, without computing elided elements.
15892 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15893 encoding. Update for new vec_perm_indices class.
15894 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15895 the given vec_perm_builder.
15896 (expand_vec_perm_var): Update vec_perm_builder constructor.
15897 (expand_mult_highpart): Use vec_perm_builder instead of
15898 auto_vec_perm_indices.
15899 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15900 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15901 or double series encoding as appropriate.
15902 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15903 vec_perm_indices instead of auto_vec_perm_indices.
15904 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15905 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15906 (vect_permute_store_chain): Likewise.
15907 (vect_grouped_load_supported): Likewise.
15908 (vect_permute_load_chain): Likewise.
15909 (vect_shift_permute_load_chain): Likewise.
15910 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15911 (vect_transform_slp_perm_load): Likewise.
15912 (vect_schedule_slp_instance): Likewise.
15913 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15914 (vectorizable_mask_load_store): Likewise.
15915 (vectorizable_bswap): Likewise.
15916 (vectorizable_store): Likewise.
15917 (vectorizable_load): Likewise.
15918 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15919 vec_perm_indices instead of auto_vec_perm_indices. Use
15920 tree_to_vec_perm_builder to read the vector from a tree.
15921 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15922 vec_perm_builder instead of a vec_perm_indices.
15923 (have_whole_vector_shift): Use vec_perm_builder and
15924 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15925 truncation to calc_vec_perm_mask_for_shift.
15926 (vect_create_epilog_for_reduction): Likewise.
15927 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15928 from auto_vec_perm_indices to vec_perm_indices.
15929 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15930 instead of changing individual elements.
15931 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15932 the vector in d.perm.
15933 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15934 from auto_vec_perm_indices to vec_perm_indices.
15935 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15936 instead of changing individual elements.
15937 (arm_vectorize_vec_perm_const): Use new_vector to install
15938 the vector in d.perm.
15939 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15940 Update vec_perm_builder constructor.
15941 (rs6000_expand_interleave): Likewise.
15942 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15943 (rs6000_expand_interleave): Likewise.
15945 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15947 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15948 to qimode could truncate the indices.
15949 * optabs.c (expand_vec_perm_var): Likewise.
15951 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15953 * Makefile.in (OBJS): Add vec-perm-indices.o.
15954 * vec-perm-indices.h: New file.
15955 * vec-perm-indices.c: Likewise.
15956 * target.h (vec_perm_indices): Replace with a forward class
15958 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15959 * optabs.h: Include vec-perm-indices.h.
15960 (expand_vec_perm): Delete.
15961 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15962 (expand_vec_perm_const): Declare.
15963 * target.def (vec_perm_const_ok): Replace with...
15964 (vec_perm_const): ...this new hook.
15965 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15966 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15967 * doc/tm.texi: Regenerate.
15968 * optabs.def (vec_perm_const): Delete.
15969 * doc/md.texi (vec_perm_const): Likewise.
15970 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15971 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15972 expand_vec_perm for constant permutation vectors. Assert that
15973 the mode of variable permutation vectors is the integer equivalent
15974 of the mode that is being permuted.
15975 * optabs-query.h (selector_fits_mode_p): Declare.
15976 * optabs-query.c: Include vec-perm-indices.h.
15977 (selector_fits_mode_p): New function.
15978 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15979 is defined, instead of checking whether the vec_perm_const_optab
15980 exists. Use targetm.vectorize.vec_perm_const instead of
15981 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15982 fit in the vector mode before using a variable permute.
15983 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15984 vec_perm_indices instead of an rtx.
15985 (expand_vec_perm): Replace with...
15986 (expand_vec_perm_const): ...this new function. Take the selector
15987 as a vec_perm_indices rather than an rtx. Also take the mode of
15988 the selector. Update call to shift_amt_for_vec_perm_mask.
15989 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15990 Use vec_perm_indices::new_expanded_vector to expand the original
15991 selector into bytes. Check whether the indices fit in the vector
15992 mode before using a variable permute.
15993 (expand_vec_perm_var): Make global.
15994 (expand_mult_highpart): Use expand_vec_perm_const.
15995 * fold-const.c: Includes vec-perm-indices.h.
15996 * tree-ssa-forwprop.c: Likewise.
15997 * tree-vect-data-refs.c: Likewise.
15998 * tree-vect-generic.c: Likewise.
15999 * tree-vect-loop.c: Likewise.
16000 * tree-vect-slp.c: Likewise.
16001 * tree-vect-stmts.c: Likewise.
16002 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16004 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16005 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16006 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16007 (aarch64_vectorize_vec_perm_const): ...this new function.
16008 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16009 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16010 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16011 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16012 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16013 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16014 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16016 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16017 check for NEON modes.
16018 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16019 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16020 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16021 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16023 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16024 the old VEC_PERM_CONST conditions.
16025 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16026 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16027 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16028 (ia64_vectorize_vec_perm_const_ok): Merge into...
16029 (ia64_vectorize_vec_perm_const): ...this new function.
16030 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16031 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16032 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16033 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16034 * config/mips/mips.c (mips_expand_vec_perm_const)
16035 (mips_vectorize_vec_perm_const_ok): Merge into...
16036 (mips_vectorize_vec_perm_const): ...this new function.
16037 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16038 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16039 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16040 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16041 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16042 (rs6000_expand_vec_perm_const): Delete.
16043 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16045 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16046 (altivec_expand_vec_perm_const_le): Take each operand individually.
16047 Operate on constant selectors rather than rtxes.
16048 (altivec_expand_vec_perm_const): Likewise. Update call to
16049 altivec_expand_vec_perm_const_le.
16050 (rs6000_expand_vec_perm_const): Delete.
16051 (rs6000_vectorize_vec_perm_const_ok): Delete.
16052 (rs6000_vectorize_vec_perm_const): New function.
16053 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16054 an element count and rtx array.
16055 (rs6000_expand_extract_even): Update call accordingly.
16056 (rs6000_expand_interleave): Likewise.
16057 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16058 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16059 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16060 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16061 (rs6000_expand_vec_perm_const): Delete.
16062 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16063 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16064 (altivec_expand_vec_perm_const_le): Take each operand individually.
16065 Operate on constant selectors rather than rtxes.
16066 (altivec_expand_vec_perm_const): Likewise. Update call to
16067 altivec_expand_vec_perm_const_le.
16068 (rs6000_expand_vec_perm_const): Delete.
16069 (rs6000_vectorize_vec_perm_const_ok): Delete.
16070 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16071 reference to the SPE evmerge intructions.
16072 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16073 an element count and rtx array.
16074 (rs6000_expand_extract_even): Update call accordingly.
16075 (rs6000_expand_interleave): Likewise.
16076 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16077 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16079 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16083 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16084 vector mode and that that mode matches the mode of the data
16086 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16087 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16088 directly using expand_vec_perm_1 when forcing selectors into
16090 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16092 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16094 * optabs-query.h (can_vec_perm_p): Delete.
16095 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16096 * optabs-query.c (can_vec_perm_p): Split into...
16097 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16098 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16099 particular selector is valid.
16100 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16101 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16102 (vect_grouped_load_supported): Likewise.
16103 (vect_shift_permute_load_chain): Likewise.
16104 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16105 (vect_transform_slp_perm_load): Likewise.
16106 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16107 (vectorizable_bswap): Likewise.
16108 (vect_gen_perm_mask_checked): Likewise.
16109 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16110 implementations of variable permutation vectors into account
16111 when deciding which selector to use.
16112 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16113 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16114 with a false third argument.
16115 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16116 to test whether the constant selector is valid and can_vec_perm_var_p
16117 to test whether a variable selector is valid.
16119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16121 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16122 * optabs-query.c (can_vec_perm_p): Likewise.
16123 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16124 instead of vec_perm_indices.
16125 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16126 (vect_gen_perm_mask_checked): Likewise,
16127 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16128 (vect_gen_perm_mask_checked): Likewise,
16130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16132 * optabs-query.h (qimode_for_vec_perm): Declare.
16133 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16134 (qimode_for_vec_perm): ...this new function.
16135 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16137 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16139 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16140 does not have a conditional at the top.
16142 2018-01-02 Richard Biener <rguenther@suse.de>
16144 * ipa-inline.c (big_speedup_p): Fix expression.
16146 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16149 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16152 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16156 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16157 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16158 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16159 cond_taken_branch_cost 3->4.
16161 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16163 PR tree-optimization/83581
16164 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16165 TODO_cleanup_cfg if any changes have been made.
16167 PR middle-end/83608
16168 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16169 convert_modes if target mode has the right side, but different mode
16172 PR middle-end/83609
16173 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16174 last argument when extracting from CONCAT. If either from_real or
16175 from_imag is NULL, use expansion through memory. If result is not
16176 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16177 the parts directly to inner mode, if even that fails, use expansion
16180 PR middle-end/83623
16181 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16182 check for bswap in mode rather than HImode and use that in expand_unop
16185 Copyright (C) 2018 Free Software Foundation, Inc.
16187 Copying and distribution of this file, with or without modification,
16188 are permitted in any medium without royalty provided the copyright
16189 notice and this notice are preserved.