1 2018-06-08 Martin Liska <mliska@suse.cz>
3 * ipa-pure-const.c (struct funct_state_d): Do it class instead
5 (class funct_state_summary_t): New function_summary class.
6 (has_function_state): Remove.
7 (get_function_state): Likewise.
8 (set_function_state): Likewise.
9 (add_new_function): Likewise.
10 (funct_state_summary_t::insert): New function.
11 (duplicate_node_data): Remove.
12 (remove_node_data): Remove.
13 (funct_state_summary_t::duplicate): New function.
14 (register_hooks): Create new funct_state_summaries.
15 (pure_const_generate_summary): Use it.
16 (pure_const_write_summary): Likewise.
17 (pure_const_read_summary): Likewise.
18 (propagate_pure_const): Likewise.
19 (propagate_nothrow): Likewise.
20 (dump_malloc_lattice): Likewise.
21 (propagate_malloc): Likewise.
22 (execute): Do not register hooks, just remove summary
24 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
27 2018-06-08 Martin Liska <mliska@suse.cz>
29 * ipa-reference.c (remove_node_data): Remove.
30 (duplicate_node_data): Likewise.
31 (class ipa_ref_var_info_summary_t): New class.
32 (class ipa_ref_opt_summary_t): Likewise.
33 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
34 (get_reference_optimization_summary): Use
35 ipa_ref_opt_sum_summaries.
36 (set_reference_vars_info): Remove.
37 (set_reference_optimization_summary): Likewise.
38 (ipa_init): Create summaries.
39 (init_function_info): Use function summary.
40 (ipa_ref_opt_summary_t::duplicate): New function.
41 (ipa_ref_opt_summary_t::remove): New function.
42 (get_read_write_all_from_node): Fix GNU coding style.
43 (propagate): Use function summary.
44 (write_node_summary_p): Fix GNU coding style.
45 (stream_out_bitmap): Likewise.
46 (ipa_reference_read_optimization_summary): Use function summary.
47 (ipa_reference_c_finalize): Do not release hooks.
49 2018-06-08 Martin Liska <mliska@suse.cz>
51 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
52 (analyze_function_body): Extract multiple calls of get_create.
53 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
54 * ipa-inline.c (recursive_inlining): Use ::get method.
55 * ipa-inline.h (estimate_edge_growth): Likewise.
57 2018-06-08 Martin Liska <mliska@suse.cz>
59 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
61 (hsa_function_summary::hsa_function_summary): Use the new enum
63 (hsa_gpu_implementation_p): Use hsa_summaries::get.
64 * hsa-gen.c (hsa_get_host_function): Likewise.
65 (get_brig_function_name): Likewise.
66 * ipa-hsa.c (process_hsa_functions): Likewise.
67 (ipa_hsa_write_summary): Likewise.
68 * symbol-summary.h (symtab_duplication): Use ::get function/
71 2018-06-08 Martin Liska <mliska@suse.cz>
73 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
75 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
76 (hsa_register_kernel): Likewise.
77 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
78 * hsa-gen.c (hsa_get_host_function): Likewise.
79 (get_brig_function_name): Likewise.
80 (generate_hsa): Likewise.
81 (pass_gen_hsail::execute): Likewise.
82 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
83 (devirtualization_time_bonus): Likewise.
84 (ipcp_propagate_stage): Likewise.
85 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
86 (edge_set_predicate): Likewise.
87 (evaluate_conditions_for_known_args): Likewise.
88 (evaluate_properties_for_edge): Likewise.
89 (ipa_fn_summary::reset): Likewise.
90 (ipa_fn_summary_t::duplicate): Likewise.
91 (dump_ipa_call_summary): Likewise.
92 (ipa_dump_fn_summary): Likewise.
93 (analyze_function_body): Likewise.
94 (compute_fn_summary): Likewise.
95 (estimate_edge_devirt_benefit): Likewise.
96 (estimate_edge_size_and_time): Likewise.
97 (estimate_calls_size_and_time): Likewise.
98 (estimate_node_size_and_time): Likewise.
99 (inline_update_callee_summaries): Likewise.
100 (remap_edge_change_prob): Likewise.
101 (remap_edge_summaries): Likewise.
102 (ipa_merge_fn_summary_after_inlining): Likewise.
103 (ipa_update_overall_fn_summary): Likewise.
104 (read_ipa_call_summary): Likewise.
105 (inline_read_section): Likewise.
106 (write_ipa_call_summary): Likewise.
107 (ipa_fn_summary_write): Likewise.
108 (ipa_free_fn_summary): Likewise.
109 * ipa-hsa.c (process_hsa_functions): Likewise.
110 (ipa_hsa_write_summary): Likewise.
111 (ipa_hsa_read_section): Likewise.
112 * ipa-icf.c (sem_function::merge): Likewise.
113 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
114 (do_estimate_edge_time): Likewise.
115 (estimate_size_after_inlining): Likewise.
116 (estimate_growth): Likewise.
117 (growth_likely_positive): Likewise.
118 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
119 (inline_call): Likewise.
120 * ipa-inline.c (caller_growth_limits): Likewise.
121 (can_inline_edge_p): Likewise.
122 (can_inline_edge_by_limits_p): Likewise.
123 (compute_uninlined_call_time): Likewise.
124 (compute_inlined_call_time): Likewise.
125 (want_inline_small_function_p): Likewise.
126 (edge_badness): Likewise.
127 (update_caller_keys): Likewise.
128 (update_callee_keys): Likewise.
129 (recursive_inlining): Likewise.
130 (inline_small_functions): Likewise.
131 (inline_to_all_callers_1): Likewise.
132 (dump_overall_stats): Likewise.
133 (early_inline_small_functions): Likewise.
134 (early_inliner): Likewise.
135 * ipa-inline.h (estimate_edge_growth): Likewise.
136 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
137 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
138 * ipa-prop.h (IPA_NODE_REF): Likewise.
139 (IPA_EDGE_REF): Likewise.
140 * ipa-pure-const.c (malloc_candidate_p): Likewise.
141 (propagate_malloc): Likewise.
142 * ipa-split.c (execute_split_functions): Likewise.
143 * symbol-summary.h: Rename get to get_create.
145 (get_create): Likewise.
146 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
148 2018-06-08 Martin Liska <mliska@suse.cz>
150 * symbol-summary.h (release): Move definition out of class
152 (symtab_removal): Likewise.
153 (symtab_duplication): Likewise.
155 2018-06-08 Martin Liska <mliska@suse.cz>
157 * symbol-summary.h (function_summary): Move constructor
158 implementation out of class declaration.
160 (symtab_insertion): Likewise.
161 (symtab_removal): Likewise.
162 (symtab_duplication): Likewise.
165 2018-06-08 Martin Liska <mliska@suse.cz>
167 * Makefile.in: Remove support for MPX (macros, related functions,
168 fields in cgraph_node, ...).
169 * builtin-types.def (BT_BND): Likewise.
170 (BT_FN_BND_CONST_PTR): Likewise.
171 (BT_FN_CONST_PTR_BND): Likewise.
172 (BT_FN_VOID_PTR_BND): Likewise.
173 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
174 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
175 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
176 (expand_builtin_mempcpy_with_bounds): Likewise.
177 (expand_builtin_memset_with_bounds): Likewise.
178 (expand_builtin_memset_args): Likewise.
179 (std_expand_builtin_va_start): Likewise.
180 (expand_builtin): Likewise.
181 (expand_builtin_with_bounds): Likewise.
182 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
183 (DEF_LIB_BUILTIN_CHKP): Likewise.
184 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
185 (DEF_CHKP_BUILTIN): Likewise.
186 (BUILT_IN_MEMCPY): Likewise.
187 (BUILT_IN_MEMMOVE): Likewise.
188 (BUILT_IN_MEMPCPY): Likewise.
189 (BUILT_IN_MEMSET): Likewise.
190 (BUILT_IN_STPCPY): Likewise.
191 (BUILT_IN_STRCAT): Likewise.
192 (BUILT_IN_STRCHR): Likewise.
193 (BUILT_IN_STRCPY): Likewise.
194 (BUILT_IN_STRLEN): Likewise.
195 (BUILT_IN_MEMCPY_CHK): Likewise.
196 (BUILT_IN_MEMMOVE_CHK): Likewise.
197 (BUILT_IN_MEMPCPY_CHK): Likewise.
198 (BUILT_IN_MEMSET_CHK): Likewise.
199 (BUILT_IN_STPCPY_CHK): Likewise.
200 (BUILT_IN_STRCAT_CHK): Likewise.
201 (BUILT_IN_STRCPY_CHK): Likewise.
202 * calls.c (store_bounds): Likewise.
203 (emit_call_1): Likewise.
204 (special_function_p): Likewise.
205 (maybe_warn_nonstring_arg): Likewise.
206 (initialize_argument_information): Likewise.
207 (finalize_must_preallocate): Likewise.
208 (compute_argument_addresses): Likewise.
209 (expand_call): Likewise.
210 * cfgexpand.c (expand_call_stmt): Likewise.
211 (expand_return): Likewise.
212 (expand_gimple_stmt_1): Likewise.
213 (pass_expand::execute): Likewise.
214 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
215 (cgraph_node::remove): Likewise.
216 (cgraph_node::dump): Likewise.
217 (cgraph_node::verify_node): Likewise.
218 * cgraph.h (chkp_function_instrumented_p): Likewise.
219 (symtab_node::get_alias_target): Likewise.
220 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
221 (cgraph_local_p): Likewise.
222 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
223 (cgraph_edge::rebuild_references): Likewise.
224 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
225 (walk_polymorphic_call_targets): Likewise.
226 (cgraph_node::expand_thunk): Likewise.
227 (symbol_table::output_weakrefs): Likewise.
228 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
229 (ix86_handle_option): Likewise.
230 * config/i386/constraints.md: Likewise.
231 * config/i386/i386-builtin-types.def (BND): Likewise.
235 * config/i386/i386-builtin.def (BDESC_END): Likewise.
236 (BDESC_FIRST): Likewise.
238 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
239 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
240 * config/i386/i386.c (enum reg_class): Likewise.
241 (ix86_target_string): Likewise.
242 (ix86_option_override_internal): Likewise.
243 (ix86_conditional_register_usage): Likewise.
244 (ix86_valid_target_attribute_inner_p): Likewise.
245 (ix86_set_indirect_branch_type): Likewise.
246 (ix86_set_current_function): Likewise.
247 (ix86_function_arg_regno_p): Likewise.
248 (init_cumulative_args): Likewise.
249 (ix86_function_arg_advance): Likewise.
250 (ix86_function_arg): Likewise.
251 (ix86_pass_by_reference): Likewise.
252 (ix86_function_value_regno_p): Likewise.
253 (ix86_function_value_1): Likewise.
254 (ix86_function_value_bounds): Likewise.
255 (ix86_return_in_memory): Likewise.
256 (ix86_setup_incoming_vararg_bounds): Likewise.
257 (ix86_va_start): Likewise.
258 (indirect_thunk_need_prefix): Likewise.
259 (print_reg): Likewise.
260 (ix86_print_operand): Likewise.
261 (ix86_expand_call): Likewise.
262 (ix86_output_function_return): Likewise.
263 (reg_encoded_number): Likewise.
264 (BDESC_VERIFYS): Likewise.
265 (ix86_init_mpx_builtins): Likewise.
266 (ix86_init_builtins): Likewise.
267 (ix86_emit_cmove): Likewise.
268 (ix86_emit_move_max): Likewise.
269 (ix86_expand_builtin): Likewise.
270 (ix86_builtin_mpx_function): Likewise.
271 (ix86_get_arg_address_for_bt): Likewise.
272 (ix86_load_bounds): Likewise.
273 (ix86_store_bounds): Likewise.
274 (ix86_load_returned_bounds): Likewise.
275 (ix86_store_returned_bounds): Likewise.
276 (ix86_class_likely_spilled_p): Likewise.
277 (ix86_hard_regno_mode_ok): Likewise.
278 (x86_order_regs_for_local_alloc): Likewise.
279 (ix86_mitigate_rop): Likewise.
280 (ix86_bnd_prefixed_insn_p): Likewise.
281 (ix86_mpx_bound_mode): Likewise.
282 (ix86_make_bounds_constant): Likewise.
283 (ix86_initialize_bounds): Likewise.
284 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
285 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
286 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
287 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
288 (TARGET_CHKP_BOUND_MODE): Likewise.
289 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
290 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
291 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
292 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
293 * config/i386/i386.h (TARGET_MPX): Likewise.
294 (TARGET_MPX_P): Likewise.
295 (VALID_BND_REG_MODE): Likewise.
296 (FIRST_BND_REG): Likewise.
297 (LAST_BND_REG): Likewise.
298 (enum reg_class): Likewise.
299 (BND_REG_P): Likewise.
300 (BND_REGNO_P): Likewise.
302 (ADJUST_INSN_LENGTH): Likewise.
303 * config/i386/i386.md: Likewise.
304 * config/i386/i386.opt: Likewise.
305 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
307 (LINK_MPX): Likewise.
308 (MPX_SPEC): Likewise.
309 (LIBMPX_SPEC): Likewise.
310 (LIBMPXWRAPPERS_SPEC): Likewise.
311 (CHKP_SPEC): Likewise.
312 * config/i386/predicates.md: Likewise.
313 * dbxout.c (dbxout_type): Likewise.
314 * doc/extend.texi: Likewise.
315 * doc/invoke.texi: Likewise.
316 * doc/md.texi: Likewise.
317 * doc/tm.texi: Likewise.
318 * doc/tm.texi.in: Likewise.
319 * dwarf2out.c (is_base_type): Likewise.
320 (gen_formal_types_die): Likewise.
321 (gen_subprogram_die): Likewise.
322 (gen_type_die_with_usage): Likewise.
323 (gen_decl_die): Likewise.
324 (dwarf2out_late_global_decl): Likewise.
325 * expr.c (expand_assignment): Likewise.
326 (emit_storent_insn): Likewise.
327 (store_expr_with_bounds): Likewise.
328 (store_expr): Likewise.
329 (expand_expr_real_1): Likewise.
330 * expr.h (store_expr_with_bounds): Likewise.
331 * function.c (use_register_for_decl): Likewise.
332 (struct bounds_parm_data): Likewise.
333 (assign_parms_augmented_arg_list): Likewise.
334 (assign_parm_find_entry_rtl): Likewise.
335 (assign_parm_is_stack_parm): Likewise.
336 (assign_parm_load_bounds): Likewise.
337 (assign_bounds): Likewise.
338 (assign_parms): Likewise.
339 (expand_function_start): Likewise.
340 * gcc.c (CHKP_SPEC): Likewise.
341 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
342 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
343 (wrestrict_dom_walker::check_call): Likewise.
344 * gimple.c (gimple_build_call_from_tree): Likewise.
345 * gimple.h (enum gf_mask): Likewise.
346 (gimple_call_with_bounds_p): Likewise.
347 (gimple_call_set_with_bounds): Likewise.
348 * gimplify.c (gimplify_init_constructor): Likewise.
349 * ipa-cp.c (initialize_node_lattices): Likewise.
350 (propagate_constants_across_call): Likewise.
351 (find_more_scalar_values_for_callers_subset): Likewise.
352 * ipa-hsa.c (process_hsa_functions): Likewise.
353 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
354 * ipa-icf.c (sem_function::merge): Likewise.
355 * ipa-inline.c (early_inliner): Likewise.
356 * ipa-pure-const.c (warn_function_noreturn): Likewise.
357 (warn_function_cold): Likewise.
358 (propagate_pure_const): Likewise.
359 * ipa-ref.h (enum GTY): Likewise.
360 * ipa-split.c (find_retbnd): Likewise.
361 (consider_split): Likewise.
362 (split_function): Likewise.
363 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
364 * ipa.c (walk_polymorphic_call_targets): Likewise.
365 (symbol_table::remove_unreachable_nodes): Likewise.
366 (process_references): Likewise.
367 (cgraph_build_static_cdtor_1): Likewise.
368 * lto-cgraph.c (lto_output_node): Likewise.
369 (output_refs): Likewise.
370 (compute_ltrans_boundary): Likewise.
371 (input_overwrite_node): Likewise.
372 (input_node): Likewise.
373 (input_cgraph_1): Likewise.
374 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
375 * passes.c (pass_manager::execute_early_local_passes): Likewise.
376 (class pass_chkp_instrumentation_passes): Likewise.
377 (make_pass_chkp_instrumentation_passes): Likewise.
378 * passes.def: Likewise.
379 * rtl.h (struct GTY): Likewise.
380 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
381 * stor-layout.c (layout_type): Likewise.
382 * symtab.c: Likewise.
383 * target.def: Likewise.
384 * targhooks.c (default_chkp_bound_type): Likewise.
385 (default_chkp_bound_mode): Likewise.
386 (default_builtin_chkp_function): Likewise.
387 (default_chkp_function_value_bounds): Likewise.
388 (default_chkp_make_bounds_constant): Likewise.
389 (default_chkp_initialize_bounds): Likewise.
390 * targhooks.h (default_chkp_bound_type): Likewise.
391 (default_chkp_bound_mode): Likewise.
392 (default_builtin_chkp_function): Likewise.
393 (default_chkp_function_value_bounds): Likewise.
394 (default_chkp_make_bounds_constant): Likewise.
395 (default_chkp_initialize_bounds): Likewise.
396 * toplev.c (compile_file): Likewise.
397 (process_options): Likewise.
398 * tree-core.h (DEF_BUILTIN): Likewise.
399 (DEF_BUILTIN_CHKP): Likewise.
400 * tree-inline.c (declare_return_variable): Likewise.
401 (remap_gimple_stmt): Likewise.
403 (initialize_inlined_parameters): Likewise.
404 (expand_call_inline): Likewise.
405 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
406 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
407 (make_pass_ipa_chkp_produce_thunks): Likewise.
408 (make_pass_chkp): Likewise.
409 (make_pass_chkp_opt): Likewise.
410 (make_pass_chkp_instrumentation_passes): Likewise.
411 * tree-pretty-print.c (dump_generic_node): Likewise.
412 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
413 * tree-ssa-dce.c (propagate_necessity): Likewise.
414 (eliminate_unnecessary_stmts): Likewise.
415 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
416 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
417 * tree-ssa-sccvn.h: Likewise.
418 * tree-ssa-strlen.c (get_string_length): Likewise.
419 (valid_builtin_call): Likewise.
420 (adjust_last_stmt): Likewise.
421 (handle_builtin_strchr): Likewise.
422 (handle_builtin_strcpy): Likewise.
423 (handle_builtin_stxncpy): Likewise.
424 (handle_builtin_memcpy): Likewise.
425 (handle_builtin_strcat): Likewise.
426 (strlen_check_and_optimize_stmt): Likewise.
427 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
428 * tree-streamer-in.c: Likewise.
429 * tree-streamer.c (record_common_node): Likewise.
430 * tree.c (tree_code_size): Likewise.
431 (wide_int_to_tree_1): Likewise.
432 (type_contains_placeholder_1): Likewise.
433 (build_common_tree_nodes): Likewise.
434 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
435 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
436 (POINTER_BOUNDS_P): Likewise.
437 (BOUNDED_TYPE_P): Likewise.
438 (BOUNDED_P): Likewise.
439 (CALL_WITH_BOUNDS_P): Likewise.
440 (pointer_bounds_type_node): Likewise.
441 * value-prof.c (gimple_ic): Likewise.
442 * var-tracking.c (vt_add_function_parameters): Likewise.
443 * varasm.c (make_decl_rtl): Likewise.
444 (assemble_start_function): Likewise.
445 (output_constant): Likewise.
446 (maybe_assemble_visibility): Likewise.
447 * varpool.c (ctor_for_folding): Likewise.
448 * chkp-builtins.def: Remove.
449 * ipa-chkp.c: Remove.
450 * ipa-chkp.h: Remove.
451 * rtl-chkp.c: Remove.
452 * rtl-chkp.h: Remove.
453 * tree-chkp-opt.c: Remove.
454 * tree-chkp.c: Remove.
455 * tree-chkp.h: Remove.
457 2018-06-07 Carl Love <cel@us.ibm.com>
459 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
460 vextract_fp_from_shortl): Add BE support.
462 2018-06-07 Paul Koning <ni1d@arrl.net>
464 * compare-elim.c (try_merge_compare): Don't merge compare if
465 address contains a side effect.
466 (try_eliminate_compare): Likewise.
468 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
470 * config.gcc: Support "tremont".
471 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
472 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
474 * config/i386/i386.c (m_TREMONT): Define.
475 (processor_target_table): Add "tremont".
476 (PTA_TREMONT): Define.
477 (ix86_lea_outperforms): Add TARGET_TREMONT.
478 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
479 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
480 and M_INTEL_GOLDMONT_PLUS.
481 (fold_builtin_cpu): Add "tremont".
482 (ix86_add_stmt_cost): Add TARGET_TREMONT.
483 (ix86_option_override_internal): Add "tremont".
484 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
485 (processor_type): Add PROCESSOR_TREMONT.
486 * config/i386/x86-tune.def: Add m_TREMONT.
487 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
489 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
491 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
492 symbol defined for msp430i* devices to be lower case.
494 gcc/testsuite/gcc.target/msp430/
495 * msp430i-device-symbol.c: New test.
496 * msp430f-device-symbol.c: New test.
497 * msp430.h: New test header file.
499 2018-06-07 Richard Biener <rguenther@suse.de>
501 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
502 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
503 Properly wrap signed arithmetic if overflow wraps.
505 2018-06-07 Jakub Jelinek <jakub@redhat.com>
507 PR tree-optimization/69615
508 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
509 of a cast from a same precision integral SSA_NAME in a bb dominated
510 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
511 cast to utype if rhs2 has already a compatible type.
513 2018-06-07 Richard Biener <rguenther@suse.de>
515 PR tree-optimization/85935
516 * graphite-scop-detection.c (find_params_in_bb): Analyze
517 condition operands with respect to the correct loop. Assert
518 the analysis doesn't fail.
520 2018-06-04 Carl Love <cel@us.ibm.com>
522 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
523 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
524 as it is slightly cheaper.
525 (first_match_or_eos_index_<mode>):
526 Calculate index using natural element order.
527 (first_match_index_<mode>):
528 Calculate index using natural element order.
529 (first_match_or_eos_index_<mode>):
530 Calculate index using natural order.
531 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
533 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
534 P9V_BUILTIN_VCLZLSBB_V16QI.
535 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
538 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
540 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
541 indentation and line wrap for many prototypes. Add missing
542 @smallexample directives around block of prototypes for vec_xl and
545 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
547 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
548 track if we pass or return IEEE 128-bit floating point.
549 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
550 C++ mangling that is compatible with GCC 8.1.
551 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
552 (init_cumulative_args): Note if we pass or return IEEE 128-bit
553 floating point types.
554 (rs6000_function_arg_advance_1): Likewise.
555 (rs6000_mangle_type): Optionally generate mangled names that match
556 what GCC 8.1 generated for IEEE 128-bit floating point types.
557 (rs6000_globalize_decl_name): If we have an external function that
558 passes or returns IEEE 128-bit types, generate a weak reference
559 from the mangled name used in GCC 8.1 to the current mangled
561 (rs6000_init_builtins): Make __ibm128 use the long double type if
562 long double is IBM extended double. Make __float128 use the long
563 double type if long double is IEEE 128-bit.
566 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
567 macro for __ibm128 built-in functions.
568 (PACK_IF): Add __ibm128 pack/unpack functions.
569 (UNPACK_IF): Likewise.
570 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
571 enable long double built-in functions if long double is IEEE
572 128-bit floating point.
573 (rs6000_invalid_builtin): Update long double built-in function
575 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
576 functions, adjust the built-in function to use the long double
577 built-in function if __ibm128 and long double are the same type.
578 * doc/extend.texi (PowerPC builtins): Update documention for
579 __builtin_{,un}pack_longdouble. Add documentation for
580 __builtin_{,un}pack_ibm128.
582 2018-06-06 Jim Wilson <jimw@sifive.com>
584 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
585 (struct machine_function): New field interrupt_mode.
586 (riscv_handle_type_attribute): New function. Add forward declaration.
587 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
588 (riscv_expand_epilogue): Check interrupt_mode field.
589 (riscv_set_current_function): Check interrupt attribute args and
590 set interrupt_mode field.
591 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
592 (riscv_sret, riscv_uret): New.
593 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
594 new arguments to interrupt attribute.
596 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
599 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
600 Don't handle -mcpu=power8 if -mpower9-vector is also used.
602 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
604 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
605 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
606 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
607 several redundant entries.
609 2018-06-06 David Malcolm <dmalcolm@redhat.com>
611 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
612 type from "rtx" to "rtx_insn *".
613 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
614 for local "call_insn", removing cast.
615 (ix86_expand_call): Likewise, introducing a "call_insn" local.
617 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
619 PR tree-optimization/86066
620 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
621 for BIT_INSERT_EXPR stores.
623 2018-06-06 Richard Biener <rguenther@suse.de>
625 PR tree-optimization/86062
626 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
628 of to be offsetted base.
630 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
632 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
633 to be static and remove check on interrupt attribute name.
635 gcc/testsuite/gcc.target/msp430/
636 * function-attributes-4.c: New test.
637 * static-interrupts.c: New test.
639 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
641 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
642 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
644 2018-06-05 Steve Ellcey <sellcey@cavium.com>
647 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
649 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
650 Remove second argument, change how error is called.
651 (aarch64_layout_arg): Remove second argument from
652 aarch64_err_no_fpadvsimd call.
653 (aarch64_init_cumulative_args): Ditto.
654 (aarch64_gimplify_va_arg_expr): Ditto.
655 * config/aarch64/aarch64.md (mov<mode>): Ditto.
657 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
659 * config/i386/i386.md (simple_return_indirect_internal): New expander.
660 (*simple_return_indirect_internal<mode>): Rename from
661 simple_return_indirect_internal. Use W mode iterator.
662 (rstorssp): New expander.
663 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
664 (clrssbsy): New expander.
665 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
667 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
669 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
671 (cmse_check_pointed_object): Likewise.
673 2018-06-05 Martin Liska <mliska@suse.cz>
675 PR gcov-profile/47618
676 * doc/invoke.texi: Document how -fprofile-dir format
679 2018-06-05 Richard Biener <rguenther@suse.de>
681 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
682 removal pretend DOM info isn't available so we do not update
683 it and only remove edges, not dominated blocks. Actually free
684 DOM info in case we removed something. Remove unreachable blocks.
685 (mfb_keep_latches): Work with either DOM info or marked backedges.
686 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
687 first. Mark backedges if DOM info isn't available.
688 (Re-)compute DOM info after cleanup_control_flow_pre.
690 2018-06-05 Richard Biener <rguenther@suse.de>
692 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
693 (locus_discrim_hasher::hash): Adjust.
694 (locus_discrim_hasher::equal): Likewise.
695 (next_discriminator_for_locus): Work on line directly.
696 (same_line_p): Pass in expanded locus1 as well.
697 (assign_discriminators): Avoid redundant location expansions.
699 2018-06-05 Richard Biener <rguenther@suse.de>
701 PR tree-optimization/86046
702 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
703 if required after clearing TREE_ADDRESSABLE.
705 2018-06-05 Richard Biener <rguenther@suse.de>
707 PR tree-optimization/86047
708 * tree-ssa-loop.c (for_each_index): Glob handling of all
709 decls and constants and really handle all of them.
711 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
714 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
715 qualifier_void_pointer and qualifier_const_void_pointer.
716 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
717 (arm_init_builtins): Handle the above.
718 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
719 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
720 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
723 2018-06-05 Martin Liska <mliska@suse.cz>
725 * auto-profile.c (read_autofdo_file): Do not use
726 gcov_ctr_summary struct.
727 (afdo_callsite_hot_enough_for_early_inline): Likewise.
728 * coverage.c (struct counts_entry): Likewise.
729 (read_counts_file): Read just single summary entry.
730 (get_coverage_counts): Use gcov_summary struct.
731 * coverage.h (get_coverage_counts): Likewise.
732 * gcov-dump.c (dump_working_sets): Likewise.
733 (tag_summary): Dump just single summary.
734 * gcov-io.c (gcov_write_summary): Write just histogram
736 (gcov_read_summary): Read just single summary.
737 (compute_working_sets): Use gcov_summary struct.
738 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
739 of GCOV_COUNTERS_SUMMABLE.
740 (GCOV_COUNTERS_SUMMABLE): Remove.
741 (GCOV_FIRST_VALUE_COUNTER): Replace with
742 GCOV_COUNTER_V_INTERVAL.
743 (struct gcov_ctr_summary): Remove.
744 (struct gcov_summary): Directly use fields of former
746 (compute_working_sets): Use gcov_summary struct.
747 * gcov.c (read_count_file): Do not use ctrs fields.
748 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
750 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
752 * profile.c: Likewise.
753 * profile.h: Likewise.
755 2018-06-05 Martin Liska <mliska@suse.cz>
757 PR gcov-profile/84846
758 * gcov.c (output_lines): Print working directory only
759 in intermediate format.
761 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
763 * config/s390/s390-builtin-types.def: Add void function type.
764 * config/s390/s390-builtins.def: Use the function type for the
767 2018-06-04 Jim Wilson <jimw@sifive.com>
769 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
771 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
772 and maybe_eh_return. Change regno to unsigned int. Use new args to
773 handle EH_RETURN_DATA_REGNO registers properly.
774 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
775 (riscv_expand_epilogue): Update comment. Change argument name and
776 type. Update code to use new name and type. Pass new args to
777 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
779 * config/riscv/riscv.md (NORMAL_RETURN): New.
780 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
781 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
782 (eh_return): Call gen_eh_return_internal and emit barrier.
783 (eh_return_internal): Call riscv_expand_epilogue.
785 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
787 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
788 bit_insertion field and declare can_be_merged_into method.
789 (merged_store_group::can_be_merged_into): New method.
790 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
791 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
792 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
794 2018-06-04 Richard Biener <rguenther@suse.de>
796 PR tree-optimization/85955
797 * builtins.c (fold_builtin_sincos): Convert pointers to
798 destination to appropriate type before dereferencing.
800 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
802 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
804 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
806 * expr.c (expand_expr_real_1): Force the operand into memory if
807 its TYPE_MODE is BLKmode and if there is no integer mode for
808 the number of bits being extracted.
810 2018-06-04 Jakub Jelinek <jakub@redhat.com>
814 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
815 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
817 2018-06-04 Richard Biener <rguenther@suse.de>
819 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
820 (cleanup_tree_cfg_noloop): ... single caller. Do
821 start_recording_case_labels later.
823 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
825 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
826 to _IMMINTRIN_H_INCLUDED.
827 * config/i386/pconfigintrin.h: Ditto.
828 * config/i386/waitpkgintrin.h: Ditto.
829 * config/i386/immintrin.h: Add includes for sgxintrin.h,
830 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
831 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
832 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
833 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
834 waitpkgintrin.h and cldemoteintrin.h.
836 2018-06-04 Richard Biener <rguenther@suse.de>
838 PR tree-optimization/86038
839 * tracer.c (find_best_successor): Check probability for
840 being initialized, bail out if not.
842 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
845 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
846 of bits to ignore when comparing architectures.
848 2018-06-04 Jakub Jelinek <jakub@redhat.com>
850 PR tree-optimization/69615
851 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
852 maximum or minimum of the type, try to merge it also as if
853 range1 is + [-, x - 1] or + [x + 1, -].
856 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
858 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
860 PR tree-optimization/86034
861 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
862 the unsigned bitfield type in a bit insertion sequence if it does not
863 have a larger precision than the bitfield size.
864 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
866 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
868 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
870 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
872 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
873 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
874 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
875 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
877 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
879 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
880 Disable -fdelete-null-pointer-checks for ELF toolchain.
882 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
883 Kito Cheng <kito.cheng@gmail.com>
885 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
886 (nds32le-*-*, nds32be-*-*): Integrate checking process.
887 (nds32*-*-*): Add glibc and uclibc conditions.
888 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
889 (TARGET_EXCEPT_UNWIND_INFO): Define.
890 * config/nds32/elf.h: New file.
891 * config/nds32/linux.h: New file.
892 * config/nds32/nds32-elf.opt: New file.
893 * config/nds32/nds32-linux.opt: New file.
894 * config/nds32/nds32-fp-as-gp.c
895 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
896 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
898 (nds32_asm_file_end): Ditto.
899 (nds32_print_operand): Ditto.
900 (nds32_insert_attributes): Ditto.
901 (nds32_init_libfuncs): New function.
902 (TARGET_HAVE_TLS): Define.
903 (TARGET_INIT_LIBFUNCS): Define.
904 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
906 (TARGET_ELF): Apply different mcmodel setting.
907 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
908 been migrated into elf.h and linux.h files.
909 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
910 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
911 (mcmodel): The content has been migrated into nds32-elf.opt and
912 nds32-linux.opt files.
913 * config/nds32/t-elf: New file.
914 * config/nds32/t-linux: New file.
916 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
917 Shiva Chen <shiva0217@gmail.com>
919 * config/nds32/constants.md (unspec_volatile_element): Add
920 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
921 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
923 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
924 (make_pass_nds32_fp_as_gp): Declare.
925 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
927 (nds32_asm_function_end_prologue): Remove unused asm output.
928 (nds32_asm_function_begin_epilogue): Remove unused asm output.
929 (nds32_asm_file_start): Output necessary fp_as_gp information.
930 (nds32_option_override): Adjust register usage.
931 (nds32_expand_prologue): Consider fp_as_gp situation.
932 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
933 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
936 (simple_return): Ditto.
937 (omit_fp_begin): Output special directive for fp_as_gp.
938 (omit_fp_end): Output special directive for fp_as_gp.
939 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
940 mforbid-fp-as-gp): New options.
942 2018-06-01 Mark Wielaard <mark@klomp.org>
944 * dwarf2out.c (dwarf2out_finish): Remove generation of
947 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
949 * gimple-ssa-store-merging.c: Include gimple-fold.h.
950 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
951 (struct merged_store_group): Add bit_insertion field.
952 (dump_char_array): Use standard hexadecimal format.
953 (merged_store_group::merged_store_group): Set bit_insertion to false.
954 (merged_store_group::apply_stores): Use optimal buffer size. Deal
955 with BIT_INSERT_EXPR stores. Move up code updating the mask and
956 also print the mask in the dump file.
957 (pass_store_merging::gate): Minor tweak.
958 (imm_store_chain_info::coalesce_immediate): Fix wrong association
959 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
960 stores with INTEGER_CST stores.
961 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
962 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
963 and use it throughout. Generate bit insertion sequences if need be.
964 (pass_store_merging::process_store): Remove redundant condition.
965 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
967 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
969 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
970 the 128-bit floating point types. Fix function comment.
972 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
974 * config/aarch64/aarch64-simd.md
975 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
977 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
980 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
982 PR tree-optimization/85989
983 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
985 (backprop::intersect_uses): Check it when deciding whether this
986 is a backedge reference.
987 (backprop::process_block): Add each phi to m_visited_phis
988 after visiting it, then clear it at the end.
990 2018-06-01 Richard Biener <rguenther@suse.de>
992 * tree-vectorizer.h (vect_dr_stmt): New function.
993 (vect_get_load_cost): Adjust.
994 (vect_get_store_cost): Likewise.
995 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
996 Use vect_dr_stmt instead of DR_SMTT.
997 (vect_record_base_alignments): Likewise.
998 (vect_calculate_target_alignment): Likewise.
999 (vect_compute_data_ref_alignment): Likewise and make static.
1000 (vect_update_misalignment_for_peel): Likewise.
1001 (vect_verify_datarefs_alignment): Likewise.
1002 (vector_alignment_reachable_p): Likewise.
1003 (vect_get_data_access_cost): Likewise. Pass down
1004 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1005 (vect_get_peeling_costs_all_drs): Likewise.
1006 (vect_peeling_hash_get_lowest_cost): Likewise.
1007 (vect_enhance_data_refs_alignment): Likewise.
1008 (vect_find_same_alignment_drs): Likewise.
1009 (vect_analyze_data_refs_alignment): Likewise.
1010 (vect_analyze_group_access_1): Likewise.
1011 (vect_analyze_group_access): Likewise.
1012 (vect_analyze_data_ref_access): Likewise.
1013 (vect_analyze_data_ref_accesses): Likewise.
1014 (vect_vfa_segment_size): Likewise.
1015 (vect_small_gap_p): Likewise.
1016 (vectorizable_with_step_bound_p): Likewise.
1017 (vect_prune_runtime_alias_test_list): Likewise.
1018 (vect_analyze_data_refs): Likewise.
1019 (vect_supportable_dr_alignment): Likewise.
1020 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1021 (vect_gen_prolog_loop_niters): Likewise.
1022 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1023 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1025 (vect_recog_mask_conversion_pattern): Likewise.
1026 (vect_try_gather_scatter_pattern): Likewise.
1027 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1028 to vect_get_store_cost.
1029 (vect_get_store_cost): Get stmt_info instead of DR.
1030 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1031 (vect_get_load_cost): Get stmt_info instead of DR.
1033 2018-06-01 Richard Biener <rguenther@suse.de>
1036 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1037 vars wrapped in MEM_REFs.
1039 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1041 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1042 Fix subreg tests so that we only return a choice between
1043 GENERAL_REGS and FP_REGS if the original classes included both.
1045 2018-06-01 Richard Biener <rguenther@suse.de>
1048 * tree-ssa-structalias.c (get_function_part_constraint):
1049 Handle NULL fi->decl.
1050 (find_func_aliases_for_call): Properly handle indirect
1051 fi from direct call.
1052 (find_func_clobbers): Likewise.
1053 (ipa_pta_execute): Likewise.
1054 (create_variable_info_for): For functions that are ifunc_resolver
1055 resolve to a varinfo that contains the result of the resolver call.
1056 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1059 2018-05-31 Michael Collison <michael.collison@arm.com>
1061 * config/aarch64/aarch64.md:
1062 (*fix_to_zero_extenddfdi2): New pattern.
1063 * gcc.target/aarch64/fix_extend1.c: New testcase.
1065 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1069 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1070 and BUILT_IN_STRNCMP_EQ.
1071 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1072 BUILT_IN_STRNCMP_EQ.
1073 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1074 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1075 (gimple_fold_builtin): Likewise.
1076 * tree-ssa-strlen.c (compute_string_length): New function.
1077 (determine_min_obsize): New function.
1078 (handle_builtin_string_cmp): New function to handle calls to
1079 string compare functions.
1080 (strlen_optimize_stmt): Add handling to builtin string compare
1082 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1083 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1084 * tree.c (build_common_builtin_nodes): Add new defines of
1085 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1087 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1090 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1091 functions with naked attribute.
1093 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1095 * config/i386/sse.md (avx_vec_concat<mode>):
1096 Substitute concat_tg_mode mode attribute with xtg_mode.
1097 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1098 (concat_tg_mode): Remove mode attribute.
1100 2018-05-31 Martin Sebor <msebor@redhat.com>
1103 * calls.c (alloc_max_size): Correct a logic error/typo.
1104 Treat excessive arguments as infinite. Warn for invalid arguments.
1105 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1107 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1110 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1111 and movx for Haswell.
1113 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1114 Cesar Philippidis <cesar@codesourcery.com>
1117 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1118 when emitting error on private/firstprivate reductions.
1119 * omp-low.c (lower_omp_target): Avoid reference-type processing
1120 on pointers for firstprivate clause.
1122 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1124 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1127 * config/aarch64/aarch64-simd.md
1128 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1129 (aarch64_ld1_x3_<mode>): Likewise
1130 (aarch64_st1x2<VALLDIF:mode>): Likewise
1131 (aarch64_st1_x2_<mode>): Likewise
1132 (aarch64_st1x3<VALLDIF:mode>): Likewise
1133 (aarch64_st1_x3_<mode>): Likewise
1134 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1135 (vld1_s8_x3): Likewise.
1136 (vld1_u16_x3): Likewise.
1137 (vld1_s16_x3): Likewise.
1138 (vld1_u32_x3): Likewise.
1139 (vld1_s32_x3): Likewise.
1140 (vld1_u64_x3): Likewise.
1141 (vld1_s64_x3): Likewise.
1142 (vld1_f16_x3): Likewise.
1143 (vld1_f32_x3): Likewise.
1144 (vld1_f64_x3): Likewise.
1145 (vld1_p8_x3): Likewise.
1146 (vld1_p16_x3): Likewise.
1147 (vld1_p64_x3): Likewise.
1148 (vld1q_u8_x3): Likewise.
1149 (vld1q_s8_x3): Likewise.
1150 (vld1q_u16_x3): Likewise.
1151 (vld1q_s16_x3): Likewise.
1152 (vld1q_u32_x3): Likewise.
1153 (vld1q_s32_x3): Likewise.
1154 (vld1q_u64_x3): Likewise.
1155 (vld1q_s64_x3): Likewise.
1156 (vld1q_f16_x3): Likewise.
1157 (vld1q_f32_x3): Likewise.
1158 (vld1q_f64_x3): Likewise.
1159 (vld1q_p8_x3): Likewise.
1160 (vld1q_p16_x3): Likewise.
1161 (vld1q_p64_x3): Likewise.
1162 (vst1_s64_x2): Likewise.
1163 (vst1_u64_x2): Likewise.
1164 (vst1_f64_x2): Likewise.
1165 (vst1_s8_x2): Likewise.
1166 (vst1_p8_x2): Likewise.
1167 (vst1_s16_x2): Likewise.
1168 (vst1_p16_x2): Likewise.
1169 (vst1_s32_x2): Likewise.
1170 (vst1_u8_x2): Likewise.
1171 (vst1_u16_x2): Likewise.
1172 (vst1_u32_x2): Likewise.
1173 (vst1_f16_x2): Likewise.
1174 (vst1_f32_x2): Likewise.
1175 (vst1_p64_x2): Likewise.
1176 (vst1q_s8_x2): Likewise.
1177 (vst1q_p8_x2): Likewise.
1178 (vst1q_s16_x2): Likewise.
1179 (vst1q_p16_x2): Likewise.
1180 (vst1q_s32_x2): Likewise.
1181 (vst1q_s64_x2): Likewise.
1182 (vst1q_u8_x2): Likewise.
1183 (vst1q_u16_x2): Likewise.
1184 (vst1q_u32_x2): Likewise.
1185 (vst1q_u64_x2): Likewise.
1186 (vst1q_f16_x2): Likewise.
1187 (vst1q_f32_x2): Likewise.
1188 (vst1q_f64_x2): Likewise.
1189 (vst1q_p64_x2): Likewise.
1190 (vst1_s64_x3): Likewise.
1191 (vst1_u64_x3): Likewise.
1192 (vst1_f64_x3): Likewise.
1193 (vst1_s8_x3): Likewise.
1194 (vst1_p8_x3): Likewise.
1195 (vst1_s16_x3): Likewise.
1196 (vst1_p16_x3): Likewise.
1197 (vst1_s32_x3): Likewise.
1198 (vst1_u8_x3): Likewise.
1199 (vst1_u16_x3): Likewise.
1200 (vst1_u32_x3): Likewise.
1201 (vst1_f16_x3): Likewise.
1202 (vst1_f32_x3): Likewise.
1203 (vst1_p64_x3): Likewise.
1204 (vst1q_s8_x3): Likewise.
1205 (vst1q_p8_x3): Likewise.
1206 (vst1q_s16_x3): Likewise.
1207 (vst1q_p16_x3): Likewise.
1208 (vst1q_s32_x3): Likewise.
1209 (vst1q_s64_x3): Likewise.
1210 (vst1q_u8_x3): Likewise.
1211 (vst1q_u16_x3): Likewise.
1212 (vst1q_u32_x3): Likewise.
1213 (vst1q_u64_x3): Likewise.
1214 (vst1q_f16_x3): Likewise.
1215 (vst1q_f32_x3): Likewise.
1216 (vst1q_f64_x3): Likewise.
1217 (vst1q_p64_x3): Likewise.
1219 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1221 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1222 user_label_prefix to name.
1224 * tree-core.h: Update comment about the format of NAME string
1225 passed to handler in attribute_spec.
1227 * config/msp430/msp430.md: Remove erroneous subreg expression from
1228 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1229 zero_extend{q,h}isi2.
1231 2018-05-30 Borislav Petkov <bp@suse.de>
1233 * doc/extend.texi: Document some architecture specific
1234 constraints and sort entries.
1236 2018-05-30 Martin Sebor <msebor@redhat.com>
1239 * builtins.c (expand_builtin_stpcpy_1): New function.
1240 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1241 only if the former succeeds.
1243 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1245 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1248 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1250 * doc/invoke.texi (-flinker-output): Document
1252 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1254 * passes.c (ipa_write_summaries): Only modify statements if body
1256 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1257 incrementally linking.
1258 (ipa_passes): Likewise.
1259 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1260 pass down resolution info.
1261 * common.opt (flag_incremental_link): Update info.
1262 * gcc.c (plugin specs): Turn flinker-output=* to
1263 -plugin-opt=-linker-output-known
1264 * toplev.c (compile_file): Also cut compilation when doing incremental
1266 * flag-types. (enum lto_partition_model): Add
1267 LTO_LINKER_OUTPUT_NOLTOREL.
1268 (invoke.texi): Add -flinker-output docs.
1269 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1270 link same way as WPA; do not stream in dead initializers.
1272 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1273 darf2out_register_external_decl): Support incremental link.
1275 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1277 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1279 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1281 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1282 it down to simple_object_copy_lto_debug_sections.
1283 (run_gcc): Determine incremental LTO link time and configure
1284 lto1 into non-wpa mode, disable renaming of debug sections.
1286 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1288 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1289 descriptions of various incorrectly documented functions.
1291 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1294 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1295 address check not strict.
1297 2018-05-30 Richard Biener <rguenther@suse.de>
1299 PR tree-optimization/85964
1300 * tracer.c (better_p): Drop initialized count check, we only
1301 call the function with initialized counts now.
1302 (find_best_successor): Do find a best edge if one
1303 has uninitialized count.
1304 (find_best_predecessor): Likewise. Do BB frequency check only
1305 if count is initialized.
1307 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1309 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1310 (aarch64_ldrstr_offset_compare): New.
1311 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1312 load/store orderings.
1313 (aarch64_gen_adjusted_ldpstp): Likewise.
1315 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1317 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1318 Check for subset of GENERAL_REGS and FP_REGS.
1319 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1322 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1324 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1325 and wi::to_poly_offset. Add the current offset and then check
1326 whether the sum fits, rather than using an unchecked addition of
1327 a checked term. Check for a shwi rather than a uhwi.
1328 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1329 (store_constructor): Use poly_int_tree_p.
1330 (expand_expr_real_1): Likewise.
1331 * function.c (assign_temp): Likewise.
1332 * fold-const.c (const_binop): Use poly_int_tree_p and
1334 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1336 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1337 to_poly_offset for MEM offsets.
1338 * ipa-icf.c (sem_variable::equals): Likewise.
1339 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1340 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1341 wi::to_poly_offset for BIT_FIELD_REF offsets.
1342 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1344 * var-tracking.c (emit_note_insn_var_location): Use
1345 tree_to_poly_uint64.
1347 2018-05-29 Jim Wilson <jimw@sifive.com>
1349 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1351 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1354 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1355 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1357 (sse4_1_round<mode>2): Use nonimmediate_operand
1358 for operand 1 predicate.
1360 2018-05-29 Martin Sebor <msebor@redhat.com>
1361 Richard Biener <rguenther@suse.de>
1364 * calls.c (get_size_range): Call determine_value_range instead
1365 of get_value_range..
1366 * tree-vrp.h (determine_value_range): Declared new function.
1367 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1369 2018-05-29 Richard Biener <rguenther@suse.de>
1371 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1372 sure to use non-pattern stmts for get_earlier_stmt arguments.
1373 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1374 called on pattern stmts.
1375 (get_later_stmt): Likewise.
1377 2018-05-29 Martin Liska <mliska@suse.cz>
1379 PR gcov-profile/85759
1380 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1383 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1385 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1387 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1391 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1392 VEC_PACK_FLOAT_EXPR): New tree codes.
1393 * tree-pretty-print.c (op_code_prio): Handle
1394 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1395 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1396 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1397 * tree-inline.c (estimate_operator_cost): Likewise.
1398 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1399 * fold-const.c (const_binop): Likewise.
1400 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1401 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1402 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1403 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1404 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1405 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1406 * expr.c (expand_expr_real_2): Likewise.
1407 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1408 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1409 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1411 * optabs.c (expand_widen_pattern_expr): For
1412 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1413 sign from result type rather than operand's type.
1414 (expand_binop_directly): For vec_packu_float_optab and
1415 vec_packs_float_optab allow result type to be different from operand's
1417 * optabs-tree.c (optab_for_tree_code): Handle
1418 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1419 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1420 * tree-vect-generic.c (expand_vector_operations_1): Handle
1421 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1422 VEC_PACK_FLOAT_EXPR.
1423 * tree-vect-stmts.c (supportable_widening_operation): Handle
1425 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1426 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1427 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1428 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1429 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1431 (vec_pack<floatprefix>_float_<mode>): New expander.
1432 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1434 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1435 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1436 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1437 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1438 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1440 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1441 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1442 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1443 VEC_PACK_FLOAT_EXPR): Document.
1445 2018-05-29 Richard Biener <rguenther@suse.de>
1447 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1449 (stmt_vec_info_vec): Make pointer.
1450 (init_stmt_vec_info_vec): Remove.
1451 (free_stmt_vec_info_vec): Likewise.
1452 (set_stmt_vec_info_vec): New function.
1453 (free_stmt_vec_infos): Likewise.
1454 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1455 (set_vinfo_for_stmt): Likewise.
1456 (get_earlier_stmt): Likewise.
1457 (get_later_stmt): Likewise.
1458 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1459 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1460 (vec_info::~vec_info): Free stmt_vec_infos.
1461 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1462 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1463 (pass_slp_vectorize::execute): Likewise.
1464 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1465 (free_stmt_vec_info_vec): Likewise.
1466 (set_stmt_vec_info_vec): New function.
1467 (free_stmt_vec_infos): Likewise.
1468 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1469 the global stmt_vec_info_vec.
1470 * tree-parloops.c (gather_scalar_reductions): Use
1471 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1474 2018-05-29 Richard Biener <rguenther@suse.de>
1476 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1478 2018-05-29 Martin Liska <mliska@suse.cz>
1479 David Malcolm <dmalcolm@redhat.com>
1481 * vec.c (test_reverse): New.
1482 (vec_c_tests): Add new test.
1483 * vec.h (vl_ptr>::reverse): New function.
1485 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1487 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1489 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1492 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1494 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1496 2018-05-28 Richard Biener <rguenther@suse.de>
1498 PR tree-optimization/85933
1499 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1500 look at stmts marked as vectorizable.
1502 2018-05-28 Richard Biener <rguenther@suse.de>
1504 PR tree-optimization/85934
1505 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1506 vector boolean check before scalar optimization.
1508 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1510 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1513 2018-05-28 Mark Wielaard <mark@klomp.org>
1515 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1516 if it is an expression containing a minus sign.
1518 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1520 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1522 2018-05-27 Paul Koning <ni1d@arrl.net>
1524 * config/pdp11/pdp11.md (truncsihi2): Remove.
1526 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1527 Chung-Ju Wu <jasonwucj@gmail.com>
1529 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1531 (unaligned_store_dw): Ditto.
1532 * config/nds32/nds32-memory-manipulation.c
1533 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1534 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1535 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1536 (emit_setmem_word_loop): Rename to ...
1537 (emit_setmem_doubleword_loop): ... this.
1538 (nds32_gen_dup_4_byte_to_word_value): New function.
1539 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1540 (nds32_expand_setmem_loop): Refine implementation.
1541 (nds32_expand_setmem_loop_v3m): Ditto.
1542 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1545 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1547 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1549 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1551 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1552 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1553 machine->attr_no_prologue_p.
1554 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1555 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1556 (nds32_expand_epilogue): Consider attr_naked_p.
1557 (nds32_expand_epilogue_v3pop): Likewise.
1558 (nds32_can_use_return_insn): Likewise.
1559 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1560 attr_no_prologue_p fields.
1561 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1563 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1566 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1568 * config/i386/sse.md
1569 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1571 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1573 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1575 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1577 (*<floatsuffix>floatv2div2sf2): Rename to ...
1578 (*float<floatunssuffix>v2div2sf2): ... this.
1579 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1580 (float<floatunssuffix>v2div2sf2_mask): ... this.
1581 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1582 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1583 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1585 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1587 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1589 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1591 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1593 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1595 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1596 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1597 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1598 gen_ufix_truncv8dfv8si2.
1599 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1600 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1601 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1602 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1603 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1604 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1605 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1606 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1608 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1612 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1614 2018-05-25 Jim Wilson <jimw@sifive.com>
1616 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1617 * config/riscv/riscv.c (struct machine_function): Add
1618 interrupt_handler_p and attribute_checked_p fields.
1619 (riscv_attribute_table): Add interrupt.
1620 (riscv_interrupt_type_p): New.
1621 (riscv_save_reg_p): Save extra regs for interrupt handler.
1622 (riscv_use_save_libcall): Return false for interrupt handler.
1623 (riscv_first_stack_step): Add forward declaration.
1624 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1625 for interrupt handler with large frame. Use it for saved reg list.
1626 (riscv_expand_prologue): Move flag_stack_usage_info support to
1627 eliminate duplication.
1628 (riscv_expand_epilogue): Generate mret for interrupt handler.
1629 (riscv_epilogue_uses): New.
1630 (riscv_can_use_return_insn): Return false for interrupt handler.
1631 (riscv_function_ok_for_sibcall): Likewise.
1632 (riscv_set_current_function): Add interrupt handler support.
1633 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1634 * config/riscv/riscv.md (UNSPECV_MRET): New.
1636 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1638 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1640 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1642 PR tree-optimization/85712
1643 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1644 this candidate has already been replaced in-situ by a copy.
1646 2018-05-25 Jason Merrill <jason@redhat.com>
1648 PR c++/80485 - inline function non-zero address.
1649 * symtab.c (nonzero_address): Check DECL_COMDAT.
1651 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1654 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1655 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1656 extension of SImode operation. Use const123_operand predicate.
1657 (*saddsi_1): Remove.
1658 (*saddl_se_1): Ditto.
1660 (*ssubl_se_1): Ditto.
1661 * config/alpha/predicates.md (const123_operand): New predicate.
1662 * config/alpha/constraints.md (P): Use IN_RANGE.
1664 2018-05-25 Richard Biener <rguenther@suse.de>
1666 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1668 (ref_maybe_used_by_stmt_p): Likewise.
1669 (stmt_may_clobber_ref_p): Likewise.
1670 (stmt_may_clobber_ref_p_1): Likewise.
1671 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1673 (ref_maybe_used_by_stmt_p): Likewise.
1674 (stmt_may_clobber_ref_p): Likewise.
1675 (stmt_may_clobber_ref_p_1): Likewise.
1676 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1677 the alias oracle to disambiguate DRs with stmts DR analysis
1679 (vect_analyze_data_refs): Do not give up on not analyzable
1680 DRs for BB vectorization. Remove code truncating the dataref
1683 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1686 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1687 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1688 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1690 2018-05-25 Richard Biener <rguenther@suse.de>
1692 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1693 function, combining stmt data ref gathering and fatal analysis
1695 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1696 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1697 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1698 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1699 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1701 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1703 PR tree-optimization/85720
1704 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1705 SCC if all partitions are builtins.
1706 (version_loop_by_alias_check): New parameter. Generate cancelable
1707 runtime alias check if all partitions are builtins.
1708 (distribute_loop): Update call to above function.
1710 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1712 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1713 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1714 (parm_default_def_partition_arg): Ditto.
1715 (set_parm_default_def_partition): Ditto.
1716 (get_parm_default_def_partitions): Ditto and make it static.
1717 (get_undefined_value_partitions): Ditto and make it static.
1718 (remove_ssa_form): Refactor call to init_var_map here.
1719 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1720 computation for loop region.
1721 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1722 (register_default_def): Delete.
1723 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1724 (parm_default_def_partition_arg): Ditto.
1725 (set_parm_default_def_partition): Ditto.
1726 (get_parm_default_def_partitions): Ditto and make it static.
1727 (get_undefined_value_partitions): Ditto and make it static.
1728 (coalesce_with_default, coalesce_with_default): Update comment.
1729 (create_coalesce_list_for_region): New func factored out from
1730 create_outofssa_var_map.
1731 (populate_coalesce_list_for_outofssa): New func factored out from
1732 create_outofssa_var_map and coalesce_ssa_name.
1733 (create_outofssa_var_map): Delete.
1734 (coalesce_ssa_name): Refactor to support live range computation.
1735 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1736 (get_parm_default_def_partitions): Delete.
1737 (get_undefined_value_partitions): Ditto.
1738 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1739 computation for loop region.
1740 (new_tree_live_info, loe_visit_block): Ditto.
1741 (live_worklist, set_var_live_on_entry): Ditto.
1742 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1743 * tree-ssa-live.h (struct _var_map): New fields.
1744 (init_var_map): Change decl.
1745 (region_contains_p): New.
1747 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1749 * tree-ssa-live.h (live_merge_and_clear): Delete.
1751 2018-05-25 Richard Biener <rguenther@suse.de>
1754 * tree-dump.c (dequeue_and_dump): Remove access to removed
1755 operand 2 of a SWITCH_EXPR.
1757 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1759 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1760 multiplication and division.
1761 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1762 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1763 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1764 (cond_udiv_optab, cond_umod_optab): New optabs.
1765 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1766 (IFN_COND_RDIV): New internal functions.
1767 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1768 TRUNC_MOD_EXPR and RDIV_EXPR.
1769 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1770 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1772 (SVE_INT_BINARY): Include mult.
1773 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1774 (optab, sve_int_op): Handle mult.
1775 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1777 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1778 for SVE_INT_BINARY_SD.
1780 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1782 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1783 (optab, sve_int_op): Handle div and udiv.
1784 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1785 for SVE_INT_BINARY_SD.
1786 (*<optab><mode>3): New insn for the same.
1788 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1790 * tree-vect-patterns.c: Include predict.h.
1791 (vect_recog_divmod_pattern): Restrict check for division support
1792 to when optimizing for size.
1794 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1796 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1797 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1798 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1799 (gimple_match_op::set_op): Likewise.
1800 (gimple_resimplify4): Declare.
1801 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1802 (expr::gen_transform): Likewise.
1803 (decision_tree::gen): Generate a simplification routine for 4 operands.
1804 * gimple-match-head.c (gimple_simplify): Add an overload for
1805 4 operands. In the top-level function, handle up to 4 call
1806 arguments and call gimple_resimplify4.
1807 (gimple_resimplify4): New function.
1808 (build_call_internal): Pass a fourth operand.
1809 (maybe_push_to_seq): Likewise.
1810 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1811 Fold VEC_COND_EXPRs of an operation and a default value into
1812 an IFN_COND_* function if possible.
1813 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1815 (SVE_COND_FP_BINARY): Include them.
1816 (optab, sve_fp_op): Handle them.
1817 (SVE_INT_BINARY_REV): New code iterator.
1818 (SVE_COND_FP_BINARY_REV): New int iterator.
1819 (commutative): New int attribute.
1820 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1822 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1824 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1825 (*cond_<optab><mode>): New patterns for reversed operands.
1827 2018-05-25 Richard Biener <rguenther@suse.de>
1829 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1830 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1831 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1832 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1833 * tree-vect-data-refs.c (everywhere): Adjust users.
1834 * tree-vect-loop.c (everywhere): Likewise.
1835 * tree-vect-slp.c (everywhere): Likewise.
1836 * tree-vect-stmts.c (everywhere): Likewise.
1837 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1839 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1841 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1843 (gcc_cv_as_section_exclude): ... this.
1844 Try Solaris as #exclude syntax.
1845 * configure: Regenerate.
1846 * config.in: Regenerate.
1847 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1849 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1850 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1852 * varasm.c (default_elf_asm_named_section): Don't check if
1853 HAVE_GAS_SECTION_EXCLUDE is defined.
1855 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1857 * doc/md.texi: Update the documentation of the cond_* optabs
1858 to mention the new final operand. Fix GET_MODE_NUNITS call.
1859 Describe the scalar case too.
1860 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1861 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1863 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1864 (get_conditional_internal_fn): Update comment.
1865 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1866 accumulator value as a final argument to conditional functions.
1867 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1868 a define_expand and add an "else" operand. Assert for now that
1869 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1870 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1871 (*cond_<optab><mode>): New patterns.
1872 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1873 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1874 (UNSPEC_COND_EOR): Delete.
1875 (optab): Remove associated mappings.
1876 (SVE_INT_BINARY): New code iterator.
1877 (sve_int_op): Remove int attribute and add "minus" to the code
1879 (SVE_COND_INT_OP): Delete.
1880 (SVE_COND_FP_OP): Rename to...
1881 (SVE_COND_FP_BINARY): ...this.
1883 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1885 * optabs.c (can_reuse_operands_p): New function.
1886 (maybe_legitimize_operands): Try to reuse the results for
1889 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1891 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1892 Add {q} suffix to insn mnemonic.
1894 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1896 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1897 (msp430_warn_func_return): New.
1899 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1901 * fold-const.c (tree_nonzero_bits): New function.
1902 * fold-const.h (tree_nonzero_bits): Likewise.
1903 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1904 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1906 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1910 * varasm.c (assemble_alias): Check ifunc_resolver only on
1913 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1916 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1917 when memory input operand is handled.
1919 2018-05-24 Luis Machado <luis.machado@linaro.org>
1921 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1923 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1925 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1927 * match.pd: Delay FMA folds until after vectorization.
1929 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1932 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1933 address check not strict.
1935 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1937 * gimple-match.h (gimple_match_op): New class.
1938 (mprts_hook): Replace parameters with a gimple_match_op *.
1939 (maybe_build_generic_op): Likewise.
1940 (gimple_simplified_result_is_gimple_val): Replace parameters with
1941 a const gimple_match_op *.
1942 (gimple_simplify): Replace code_helper * and tree * parameters with
1943 a gimple_match_op * parameter.
1944 (gimple_resimplify1): Replace code_helper *, tree and tree *
1945 parameters with a gimple_match_op * parameter.
1946 (gimple_resimplify2): Likewise.
1947 (gimple_resimplify3): Likewise.
1948 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1949 parameters with a gimple_match_op * parameter.
1950 * gimple-match-head.c (gimple_simplify): Change prototypes of
1951 auto-generated functions to take a gimple_match_op * instead of
1952 separate code_helper * and tree * parameters. Make the same
1953 change in the top-level overload and update calls to the
1954 gimple_resimplify routines. Update calls to the auto-generated
1955 functions and to maybe_push_res_to_seq in the publicly-facing
1956 operation-specific gimple_simplify overloads.
1957 (gimple_match_op::MAX_NUM_OPS): Define.
1958 (gimple_resimplify1): Replace rcode and ops with a single res_op
1959 parameter. Update call to gimple_simplify.
1960 (gimple_resimplify2): Likewise.
1961 (gimple_resimplify3): Likewise.
1962 (mprts_hook): Replace parameters with a gimple_match_op *.
1963 (maybe_build_generic_op): Likewise.
1964 (build_call_internal): Replace type, nargs and ops with
1965 a gimple_match_op *.
1966 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1967 with a single gimple_match_op *. Update calls to mprts_hook,
1968 build_call_internal and gimple_simplified_result_is_gimple_val.
1969 Factor out code that is common to the tree_code and combined_fn cases.
1970 * genmatch.c (expr::gen_transform): Replace tem_code and
1971 tem_ops with a gimple_match_op called tem_op. Update calls
1972 to the gimple_resimplify functions and maybe_push_res_to_seq.
1973 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1974 res_ops. Update call to the gimple_resimplify functions.
1975 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1976 (decision_tree::gen): Make the functions take a gimple_match_op *
1977 called res_op instead of separate res_code and res_ops parameters.
1978 Update call accordingly.
1979 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1980 and ops with a single res_op parameter. Update calls to
1981 maybe_build_generic_op and maybe_push_res_to_seq.
1982 (fold_stmt_1): Update calls to gimple_simplify and
1983 replace_stmt_with_simplification.
1984 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1985 and gimple_simplified_result_is_gimple_val.
1986 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1988 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1989 with a gimple_match_op *.
1990 (vn_nary_build_or_lookup): Likewise. Update call to
1991 vn_nary_build_or_lookup_1.
1992 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1993 gimple_match_op *. Update calls to the gimple_resimplify routines
1994 and to gimple_simplified_result_is_gimple_val.
1995 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1996 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1997 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1998 (visit_nary_op): Likewise.
1999 (visit_reference_op_load): Likewise.
2001 2018-05-23 Luis Machado <luis.machado@linaro.org>
2003 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2004 modifier for printing the step amount.
2006 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2009 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2012 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2014 * doc/sourcebuild.texi (Endianness): New subsubsection.
2016 2018-05-23 Luis Machado <luis.machado@linaro.org>
2018 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2019 <prefetch_dynamic_strides>: New const bool field.
2020 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2021 prefetch_dynamic_strides.
2022 (exynosm1_prefetch_tune): Likewise.
2023 (thunderxt88_prefetch_tune): Likewise.
2024 (thunderx_prefetch_tune): Likewise.
2025 (thunderx2t99_prefetch_tune): Likewise.
2026 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2028 (aarch64_override_options_internal): Update to set
2029 PARAM_PREFETCH_DYNAMIC_STRIDES.
2030 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2031 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2032 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2033 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2034 prefetch-dynamic-strides setting.
2036 2018-05-23 Luis Machado <luis.machado@linaro.org>
2038 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2039 <minimum_stride>: New const int field.
2040 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2041 minimum_stride field defaulting to -1.
2042 (exynosm1_prefetch_tune): Likewise.
2043 (thunderxt88_prefetch_tune): Likewise.
2044 (thunderx_prefetch_tune): Likewise.
2045 (thunderx2t99_prefetch_tune): Likewise.
2046 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2047 <default_opt_level>: Set to 3.
2048 (aarch64_override_options_internal): Update to set
2049 PARAM_PREFETCH_MINIMUM_STRIDE.
2050 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2051 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2052 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2053 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2054 stride is constant and is below the minimum stride threshold.
2056 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2058 * config/arm/arm-cpus.in (mode26): Delete.
2059 (armv4): Delete mode26 reference.
2060 * config/arm/arm.c (arm_configure_build_target): Delete use of
2063 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2065 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2067 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2068 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2070 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2071 (floatunsdidf2): Ditto.
2073 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2075 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2076 (fixuns_trunc<mode>si2_avx512f): Ditto.
2077 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2078 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2079 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2081 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2083 PR rtl-optimization/79985
2084 * df-scan.c (df_insn_refs_collect): Remove special case for
2085 global registers and asm statements.
2087 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2089 * extend.texi (Global Register Variables): Rewrite the bullet list.
2090 Note that the register is available for allocation. Note that access
2091 via inline asm must use constraints. Add note about async-signal
2092 handlers. Remove paragraph about automagic register selection.
2094 2018-05-23 Richard Biener <rguenther@suse.de>
2096 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2097 of fixed offset from memset VN.
2099 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2101 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2103 (alloc_cand_and_find_basis): Initialize first_interp field.
2104 (slsr_process_mul): Modify first_interp field.
2105 (slsr_process_add): Likewise.
2106 (slsr_process_cast): Modify first_interp field for each new
2108 (slsr_process_copy): Likewise.
2109 (dump_candidate): Dump first_interp field.
2110 (replace_mult_candidate): Process all interpretations, not just
2112 (replace_rhs_if_not_dup): Likewise.
2113 (replace_one_candidate): Likewise.
2115 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2117 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2119 (aarch64_needs_frame_chain): New function.
2120 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2122 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2125 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2126 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2127 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2128 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2129 as true for strict-align.
2130 (aarch64_can_inline_p): Perform checks even when callee has no
2131 attributes to check for strict alignment.
2132 * doc/extend.texi (AArch64 Function Attributes): Document
2134 * doc/invoke.texi: (AArch64 Options): Likewise.
2136 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2138 PR tree-optimization/85853
2139 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2140 the handling of the root of the node to...
2141 (vect_slp_analyze_node_operations_1): ...this new function,
2142 and run the whole thing with the child nodes' def types
2143 set according to their SLP node's def type.
2145 2018-05-23 Richard Biener <rguenther@suse.de>
2148 * tree-data-ref.c (create_runtime_alias_checks): Defer
2149 and ignore overflow warnings.
2151 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2153 PR tree-optimization/85822
2154 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2157 2018-05-23 Richard Biener <rguenther@suse.de>
2159 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2160 memset constants via native_interpret_expr.
2162 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2165 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2167 (cgraph_node::create_alias): Likewise.
2168 (cgraph_node::get_availability): Check ifunc_resolver instead
2169 of looking up ifunc attribute.
2170 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2171 * varasm.c (do_assemble_alias): Likewise.
2172 (assemble_alias): Likewise.
2173 (default_binds_local_p_3): Likewise.
2174 * cgraph.h (cgraph_node): Add ifunc_resolver.
2175 (cgraph_node::only_called_directly_or_aliased_p): Return false
2177 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2179 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2180 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2181 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2182 instead of looking up ifunc attribute.
2184 2018-05-22 Luis Machado <luis.machado@linaro.org>
2186 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2188 2018-05-22 Martin Sebor <msebor@redhat.com>
2191 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2192 only when expasion succeeds.
2193 (expand_builtin_strcmp): Same.
2194 (expand_builtin_strncmp): Same.
2196 2018-05-22 Martin Sebor <msebor@redhat.com>
2198 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2200 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2201 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2203 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2204 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2206 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2207 Add check for legitimate_address.
2208 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2209 (aarch64_swap_ldrstr_operands): New.
2210 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2213 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2214 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2216 * config/aarch64/aarch64.md: New patterns to generate stp
2218 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2219 single words and double words.
2220 (load_pair_sw, load_pair_dw): Likewise.
2221 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2223 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2225 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2226 for different mode ldpstp and add peephole for merged zero stores.
2228 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2230 (aarch64_gen_store_pair): Rename calls to match new patterns.
2231 (aarch64_gen_load_pair): Rename calls to match new patterns.
2232 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2233 (load_pair<DREG:mode><DREG2:mode>): ... This.
2234 (store_pair<mode>): Rename to...
2235 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2236 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2238 (V_INT_EQUIV): Handle SImode.
2239 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2242 2018-05-22 Martin Sebor <msebor@redhat.com>
2245 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2246 or ajust the presumed bound on an operation to avoid unnecessary
2249 2018-05-22 Martin Sebor <msebor@redhat.com>
2251 PR tree-optimization/85826
2252 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2253 assuming that a DECL necesarily has a constant size.
2255 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2258 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2260 2018-05-22 Richard Biener <rguenther@suse.de>
2262 PR tree-optimization/85834
2263 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2264 non-constant and non-zero memset arguments.
2266 2018-05-22 Martin Liska <mliska@suse.cz>
2269 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2271 2018-05-22 Richard Biener <rguenther@suse.de>
2273 PR tree-optimization/85863
2274 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2275 comparisons when vectype is specified.
2276 (vectorizable_condition): Do not specify vectype for
2277 vect_is_simple_cond when SLP vectorizing.
2279 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2282 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2283 define __ibm128 as long double.
2284 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2285 as a distinct type when IEEE 128-bit support is enabled.
2286 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2287 128-bit types to use the correct functions.
2288 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2289 convert between 128-bit floating point types that have different
2290 modes but the same representation, instead of using gen_lowpart to
2292 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2294 (IFKF_reg): New attributes to give the register constraints for
2296 (extend<mode>tf2_internal): New insns to mark an explicit
2297 conversion between 128-bit floating point types that have a
2298 different mode but share the same representation.
2300 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2302 PR tree-optimization/85814
2303 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2304 a null return from get_strinfo when unsharing the next
2305 strinfo in the chain.
2307 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2310 * varasm.c (weak_finish): Clean up weak_decls.
2312 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2314 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2315 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2316 UNSPEC_UADALP values.
2317 * config/aarch64/iterators.md (ABAL): New int iterator.
2320 (sur): Add mappings for the above.
2321 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2323 (aarch64_<sur>abal<mode>_4): Likewise.
2324 (aarch64_<sur>adalp<mode>_3): Likewise.
2325 (<sur>sadv16qi): New define_expand.
2327 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2329 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2330 (*movdf_internal): Ditto.
2331 (*rcpsf2_sse): Ditto.
2332 (*rsqrtsf2_sse): Ditto.
2333 (*sqrt<mode>2_sse): Ditto.
2335 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2337 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2339 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2340 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2341 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2342 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2344 * config/aarch64/arm_neon.h: Likewise.
2345 * config/aarch64/iterators.md (VQ_I): New.
2347 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2349 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2351 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2354 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2356 * config/nds32/constraints.md (S): New constraint.
2357 * config/nds32/nds32.md (call_internal): Use constraint S.
2358 (call_value_internal): Likewise.
2359 (sibcall_internal): Likewise.
2360 (sibcall_value_internal): Likewise.
2362 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2363 Chung-Ju Wu <jasonwucj@gmail.com>
2365 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2368 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2369 Chung-Ju Wu <jasonwucj@gmail.com>
2371 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2372 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2373 (nds32_rtx_costs_impl): Simplify.
2374 (nds32_address_cost_impl): Simplify.
2375 (nds32_init_rtx_costs): New function.
2376 (nds32_rtx_costs_speed_prefer): Likewise.
2377 (nds32_rtx_costs_size_prefer): Likewise.
2378 (nds32_address_cost_speed_prefer): Likewise.
2379 (nds32_address_cost_speed_fwprop): Likewise.
2380 (nds32_address_cost_size_prefer): Likewise.
2381 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2382 * config/nds32/nds32.c (nds32_option_override): Use
2383 nds32_init_rtx_costs function.
2385 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2387 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2388 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2389 (TARGET_PIPELINE_N8): Likewise.
2390 (TARGET_PIPELINE_N10): Likewise.
2391 (TARGET_PIPELINE_N13): Likewise.
2392 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2394 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2396 * config/nds32/nds32-fpu.md: Update copyright year.
2398 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2400 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2402 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2404 * config/nds32/nds32.c
2405 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2406 * config/nds32/nds32.opt (minline-asm-r15): New option.
2408 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2410 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2412 * config/nds32/nds32.md (abssi2): New pattern.
2414 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2416 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2417 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2418 Merge insn pattern from sse_cvtsi2ss<round_name> and
2419 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2420 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2421 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2422 using SWI48 mode iterator.
2423 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2424 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2425 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2426 pattern from sse_cvttss2si<round_saeonly_name>
2427 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2428 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2429 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2430 using SWI48 mode iterator.
2431 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2432 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2433 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2434 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2435 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2436 using SWI48 mode iterator.
2437 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2438 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2439 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2440 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2441 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2442 SWI48 mode iterator.
2443 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2444 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2445 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2446 pattern from sse_cvttsd2si<round_saeonly_name>
2447 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2449 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2451 * config/nds32/nds32-md-auxiliary.c
2452 (nds32_valid_smw_lwm_base_p): Refine.
2453 (nds32_output_smw_single_word): Refine.
2454 (nds32_output_smw_double_word): New.
2455 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2457 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2459 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2460 (nds32_output_stack_pop): Refine.
2461 (nds32_expand_unaligned_load): Refine.
2462 (nds32_expand_unaligned_store): Refine.
2464 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2465 Chung-Ju Wu <jasonwucj@gmail.com>
2467 * config/nds32/constants.md: Add TP_REGNUM constant.
2468 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2469 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2471 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2472 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2473 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2474 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2475 and PIC code generation.
2476 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2478 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2480 * config/nds32/nds32.md: Support TLS and PIC.
2481 * config/nds32/nds32.c: Support TLS and PIC.
2482 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2483 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2486 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2488 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2489 mode with E_ prefix.
2491 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2492 Chung-Ju Wu <jasonwucj@gmail.com>
2494 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2495 * config/nds32/nds32-md-auxiliary.c
2496 (symbolic_reference_mentioned_p): New.
2497 (nds32_legitimize_ict_address): New.
2498 (nds32_expand_ict_move): New.
2499 (nds32_indirect_call_referenced_p): New.
2500 (nds32_symbol_binds_local_p): Delete.
2501 (nds32_long_call_p): Modify.
2502 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2503 * config/nds32/nds32-protos.h
2504 (symbolic_reference_mentioned_p): Declare.
2505 (nds32_legitimize_ict_address): Declare.
2506 (nds32_expand_ict_move): Declare.
2507 (nds32_indirect_call_referenced_p): Declare.
2508 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2509 (nds32_relax_group): Use nds32_ict_const_p as condition.
2510 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2511 (nds32_asm_file_start): Output ict_model directive in asm code.
2512 (nds32_legitimate_address_p): Consider indirect call.
2513 (nds32_print_operand): Consider indirect call.
2514 (nds32_print_operand_address): Consider indirect call.
2515 (nds32_insert_attributes): Handle "indirect_call" attribute.
2516 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2517 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2518 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2519 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2520 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2521 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2522 (TARGET_ICT_MODEL_SMALL): Define.
2523 (TARGET_ICT_MODEL_LARGE): Define.
2524 * config/nds32/nds32.md (movsi): Consider ict model.
2525 (call, call_value): Consider ict model.
2526 (sibcall, sibcall_value): Consider ict model.
2527 * config/nds32/nds32.opt (mict-model): New option.
2528 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2531 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2532 Monk Chiang <sh.chiang04@gmail.com>
2533 Jim Wilson <jimw@sifive.com>
2535 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2536 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2537 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2538 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2539 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2540 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2541 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2542 compute save_libcall_adjustment properly.
2543 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2544 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2545 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2546 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2547 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2548 (ABI_SPEC): Handle mabi=ilp32e.
2549 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2550 (RVE): Add RVE mask.
2551 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2552 <-march>: Add rv32e as an example.
2554 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2557 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2558 (intra_create_variable_infos): Handle C++ constructors.
2560 2018-05-18 Martin Liska <mliska@suse.cz>
2562 * passes.def: Remove a redundant pass.
2564 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2567 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2569 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2571 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2573 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2575 (armv2, armv2a, armv3, armv3m): Delete architectures.
2576 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2577 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2578 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2580 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2581 (*mulsidi3adddi): Likewise.
2582 (mulsidi3): Likewise.
2583 (*mulsidi3_nov6): Likewise.
2584 (umulsidi3): Likewise.
2585 (umulsidi3_nov6): Likewise.
2586 (umaddsidi4): Likewise.
2587 (*umulsidi3adddi): Likewise.
2588 (smulsi3_highpart): Likewise.
2589 (*smulsi3_highpart_nov6): Likewise.
2590 (umulsi3_highpart): Likewise.
2591 (*umulsi3_highpart_nov6): Likewise.
2592 * config/arm/arm.h (arm_arch3m): Delete.
2593 * config/arm/arm.c (arm_arch3m): Delete.
2594 (arm_option_override_internal): Update armv3-related comment.
2595 (arm_configure_build_target): Delete use of isa_bit_mode32.
2596 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2597 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2598 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2600 (mulusa3): Likewise.
2601 * config/arm/arm-protos.h (arm_arch3m): Delete.
2602 * config/arm/arm-tables.opt: Regenerate.
2603 * config/arm/arm-tune.md: Likewise.
2604 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2605 deleted architectures.
2607 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2609 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2610 (armv5t, armv5te): New features.
2611 (ARMv5, ARMv5e): Delete fgroups.
2612 (ARMv5t, ARMv5te): Adjust for above changes.
2614 (armv5, armv5e): Delete arches.
2615 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2617 (*call_reg_arm): Likewise.
2618 (*call_value_reg_armv5): Likewise.
2619 (*call_value_reg_arm): Likewise.
2620 (*call_symbol): Likewise.
2621 (*call_value_symbol): Likewise.
2622 (*sibcall_insn): Likewise.
2623 (*sibcall_value_insn): Likewise.
2625 (prefetch): Likewise.
2626 (define_split and define_peephole2 dependent on arm_arch5):
2628 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2630 (TARGET_ARM_QBIT): Likewise.
2631 (TARGET_DSP_MULTIPLY): Likewise.
2632 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2633 (arm_arch5, arm_arch5e): Delete.
2634 (arm_arch5t, arm_arch5te): Declare.
2635 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2636 (arm_arch5t): Declare.
2637 (arm_option_reconfigure_globals): Update for the above.
2638 (arm_options_perform_arch_sanity_checks): Update comment, replace
2639 use of arm_arch5 with arm_arch5t.
2640 (use_return_insn): Likewise.
2641 (arm_emit_call_insn): Likewise.
2642 (output_return_instruction): Likewise.
2643 (arm_final_prescan_insn): Likewise.
2644 (arm_coproc_builtin_available): Likewise.
2645 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2646 arm_arch5e with arm_arch5t and arm_arch5te.
2647 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2648 (arm_arch5t, arm_arch5te): Declare.
2649 * config/arm/arm-tables.opt: Regenerate.
2650 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2651 * config/arm/t-multilib: Likewise.
2652 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2653 instead of arm_arch5.
2654 (*call_reg_thumb1): Likewise.
2655 (*call_value_reg_thumb1_v5): Likewise.
2656 (*call_value_reg_thumb1): Likewise.
2657 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2659 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2661 2018-05-18 Martin Liska <mliska@suse.cz>
2663 PR gcov-profile/84846
2664 * doc/gcov.texi: Document -t option of gcov tool.
2666 2018-05-18 Martin Liska <mliska@suse.cz>
2668 PR gcov-profile/84846
2669 * gcov.c (print_usage): Add new -t option.
2670 (process_args): Handle the option.
2671 (generate_results): Use stdout as output when requested by
2674 2018-05-18 Martin Liska <mliska@suse.cz>
2676 PR gcov-profile/84846
2677 * coverage.c (coverage_init): Write PWD to .gcno file.
2678 * doc/gcov.texi: Document how working directory is printed.
2679 * gcov-dump.c (dump_gcov_file): Print PWD.
2680 * gcov.c (output_intermediate_file): Likewise.
2681 (read_graph_file): Read PWD string.
2682 (output_lines): Print PWD.
2684 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2687 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2688 for retval and return false if all args to phi are zero.
2690 2018-05-18 Richard Biener <rguenther@suse.de>
2692 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2694 (evrp_dom_walker::before_dom_children): Call it.
2696 2018-05-18 Richard Biener <rguenther@suse.de>
2698 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2699 results when processing array refs with variable index.
2701 2018-05-18 Toon Moene <toon@moene.org>
2703 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2704 directly after that of -floop-interchange. Indicate that both
2705 options are enabled by default when specifying -O3.
2707 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2709 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2710 iterator. Delete separate integer-mode vec_set<mode> expander.
2711 (aarch64_simd_vec_setv2di): Delete.
2712 (vec_setv2di): Delete.
2713 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2714 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2715 the "w, r" alternative.
2717 2018-05-18 Martin Liska <mliska@suse.cz>
2719 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2720 * tree-pass.h (make_pass_lower_switch_O0): New function.
2721 * tree-switch-conversion.c (node_has_low_bound): Remove.
2722 (node_has_high_bound): Likewise.
2723 (node_is_bounded): Likewise.
2724 (class pass_lower_switch): Make it a template type and create
2726 (pass_lower_switch::execute): Add template argument.
2727 (make_pass_lower_switch): New function.
2728 (make_pass_lower_switch_O0): New function.
2729 (do_jump_if_equal): Remove.
2730 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2731 all the hard work to tree optimization passes.
2733 2018-05-18 Martin Liska <mliska@suse.cz>
2735 * dbgcnt.c (limit_low): Renamed from limit.
2736 (limit_high): New variable.
2737 (dbg_cnt_is_enabled): Check for upper limit.
2738 (dbg_cnt): Adjust dumping.
2739 (dbg_cnt_set_limit_by_index): Add new argument for high
2741 (dbg_cnt_set_limit_by_name): Likewise.
2742 (dbg_cnt_process_single_pair): Parse new format.
2743 (dbg_cnt_process_opt): Use strtok.
2744 (dbg_cnt_list_all_counters): Remove 'value' and add
2746 * doc/invoke.texi: Document changes.
2748 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2750 * doc/sourcebuild.texi (scalar_all_fma): Document.
2751 * tree.def (FMA_EXPR): Delete.
2752 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2753 * internal-fn.c (ternary_direct): New macro.
2754 (expand_ternary_optab_fn): Likewise.
2755 (direct_ternary_optab_supported_p): Likewise.
2756 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2757 * builtins.c (fold_builtin_fma): Delete.
2758 (fold_builtin_3): Don't call it.
2759 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2760 * expr.c (expand_expr_real_2): Likewise.
2761 * fold-const.c (operand_equal_p): Likewise.
2762 (fold_ternary_loc): Likewise.
2763 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2764 * gimple.c (DEFTREECODE): Likewise.
2765 * gimplify.c (gimplify_expr): Likewise.
2766 * optabs-tree.c (optab_for_tree_code): Likewise.
2767 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2768 * tree-eh.c (operation_could_trap_p): Likewise.
2769 (stmt_could_throw_1_p): Likewise.
2770 * tree-inline.c (estimate_operator_cost): Likewise.
2771 * tree-pretty-print.c (dump_generic_node): Likewise.
2772 (op_code_prio): Likewise.
2773 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2774 * tree-ssa-operands.c (get_expr_operands): Likewise.
2775 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2776 * fold-const-call.h (fold_fma): Delete.
2777 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2778 CFN_FNMA and CFN_FNMS.
2780 * genmatch.c (combined_fn): New enum.
2781 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2782 (commutative_op): New function.
2783 (commutate): Use it. Handle more than 2 operands.
2784 (dt_operand::gen_gimple_expr): Use commutative_op.
2785 (parser::parse_expr): Allow :c to be used with non-binary
2786 operators if the commutative operand is known.
2787 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2788 CFN_FMS, CFN_FNMA and CFN_FNMS.
2789 (backprop::process_assign_use): Remove FMA_EXPR handling.
2790 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2791 (gen_hsa_fma): New function.
2792 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2793 IFN_FNMA and IFN_FNMS.
2794 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2795 * gimple-fold.h (follow_all_ssa_edges): Declare.
2796 * gimple-fold.c (follow_all_ssa_edges): New function.
2797 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2798 gimple_build interface and use follow_all_ssa_edges to fold the result.
2799 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2800 instead of checking for optabs directly.
2801 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2802 rather than FMA_EXPRs.
2803 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2804 call to IFN_FMA instead of an FMA_EXPR.
2806 2018-05-17 Jim Wilson <jimw@sifive.com>
2808 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2809 sign extended promoted subreg, and the range does not have the sign bit
2810 set, then do a sign extend.
2812 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2813 test, check for sign extended subreg and/or constant operands, and
2814 do a sign extend in that case.
2816 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2818 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2819 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2821 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2822 Change logics_shift_reg to logics_shift_imm.
2823 (thunderx2t99_fp_loadpair_basic): Delete.
2824 (thunderx2t99_fp_storepair_basic): Delete.
2825 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2826 (thunderx2t99_asimd_polynomial): Delete.
2827 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2828 and neon_fp_mul_d_scalar_q.
2829 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2830 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2831 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2832 (thunderx2t99_asimd_lut): Add missing tbl types.
2833 (thunderx2t99_asimd_ext): Delete.
2834 (thunderx2t99_asimd_load1_1_mult): Delete.
2835 (thunderx2t99_asimd_load1_2_mult): Delete.
2836 (thunderx2t99_asimd_load1_ldp): New.
2837 (thunderx2t99_asimd_load1): New.
2838 (thunderx2t99_asimd_load2): Add missing *load2* types.
2839 (thunderx2t99_asimd_load3): New.
2840 (thunderx2t99_asimd_load4): New.
2841 (thunderx2t99_asimd_store1_1_mult): Delete.
2842 (thunderx2t99_asimd_store1_2_mult): Delete.
2843 (thunderx2t99_asimd_store2_mult): Delete.
2844 (thunderx2t99_asimd_store2_onelane): Delete.
2845 (thunderx2t99_asimd_store_stp): New.
2846 (thunderx2t99_asimd_store1): New.
2847 (thunderx2t99_asimd_store2): New.
2848 (thunderx2t99_asimd_store3): New.
2849 (thunderx2t99_asimd_store4): New.
2851 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2853 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2854 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2856 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2857 Segher Boessenkool <segher@kernel.crashing.org>
2860 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2863 2018-05-17 Richard Biener <rguenther@suse.de>
2865 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2866 for pruning loop and prune defs feeding only already visited PHIs.
2868 2018-05-17 Richard Biener <rguenther@suse.de>
2870 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2872 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2873 Richard Biener <rguenther@suse.de>
2875 PR tree-optimization/85793
2876 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2877 for VMAT_ELEMENTWISE.
2879 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2881 * internal-fn.h (lookup_internal_fn): Declare
2882 * internal-fn.c (lookup_internal_fn): New function.
2883 * gimple.c (gimple_build_call_from_tree): Handle calls to
2885 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2886 internal function names.
2887 * tree-pretty-print.c (dump_generic_node): Likewise.
2888 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2890 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2892 * gimple-fold.h (gimple_build): Make the function forms take
2893 combined_fn rather than built_in_function.
2894 (gimple_simplify): Likewise.
2895 * gimple-match-head.c (gimple_simplify): Likewise.
2896 * gimple-fold.c (gimple_build): Likewise.
2897 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2898 rather than gimple_build_call_internal.
2899 (get_initial_defs_for_reduction): Likewise.
2900 (vect_create_epilog_for_reduction): Likewise.
2901 (vectorizable_live_operation): Likewise.
2903 2018-05-17 Martin Liska <mliska@suse.cz>
2905 * gimple-ssa-sprintf.c (format_directive): Do not use
2906 space in between 'G_' and '('.
2908 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2911 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2912 even if the mask is not all ones.
2915 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2917 (ix86_gimple_fold_builtin): Likewise.
2920 * config/i386/i386.c: Include tree-vector-builder.h.
2921 (ix86_vector_shift_count): New function.
2922 (ix86_fold_builtin): Fold shift builtins by scalar count.
2923 (ix86_gimple_fold_builtin): Likewise.
2925 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2926 _mm512_setzero): New intrinsics.
2928 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2929 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2931 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2932 code generation for cases where splatting a value is not useful.
2933 * simplify-rtx.c (simplify_ternary_operation): Simplify
2934 vec_merge across a vec_duplicate and a paradoxical subreg forming
2935 a vector mode to a vec_concat.
2937 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2939 * config.gcc: Support "goldmont-plus".
2940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2942 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2943 PROCESSOR_GOLDMONT_PLUS.
2944 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2945 (processor_target_table): Add "goldmont-plus".
2946 (PTA_GOLDMONT_PLUS): Define.
2947 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2948 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2949 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2950 (fold_builtin_cpu): Add "goldmont-plus".
2951 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2952 (ix86_option_override_internal): Add "goldmont-plus".
2953 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2954 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2955 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2956 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2958 2018-05-17 Richard Biener <rguenther@suse.de>
2960 PR tree-optimization/85757
2961 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2962 remove defs that only feed that PHI from further processing.
2964 2018-05-16 Jim Wilson <jimw@sifive.com>
2966 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2968 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2970 2018-05-16 Mark Wielaard <mark@klomp.org>
2972 * dwarf2out.c (count_index_strings): New function.
2973 (output_indirect_strings): Call count_index_strings and generate
2974 header for dwarf_version >= 5.
2976 2018-05-16 Mark Wielaard <mark@klomp.org>
2978 * dwarf2out.c (dwarf_FORM): New function.
2979 (set_indirect_string): Use dwarf_FORM.
2980 (reset_indirect_string): Likewise.
2981 (size_of_die): Likewise.
2982 (value_format): Likewise.
2983 (output_die): Likewise.
2984 (add_skeleton_AT_string): Likewise.
2985 (output_macinfo_op): Likewise.
2986 (index_string): Likewise.
2987 (output_index_string_offset): Likewise.
2988 (output_index_string): Likewise.
2989 (count_index_strings): Likewise.
2991 2018-05-16 Carl Love <cel@us.ibm.com>
2993 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2994 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2996 2018-05-16 Martin Jambor <mjambor@suse.cz>
2998 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2999 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3001 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3003 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3004 (fnma<mode>4): Likewise.
3005 (fms<mode>4): Likewise.
3006 (fnms<mode>4): Likewise.
3007 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3008 (aarch64_fnma<mode>4): Likewise.
3009 (aarch64_fms<mode>4): Likewise.
3010 (aarch64_fnms<mode>4): Likewise.
3011 (aarch64_fnmadd<mode>4): Likewise.
3013 2018-05-16 Jason Merrill <jason@redhat.com>
3015 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3017 2018-05-16 Richard Biener <rguenther@suse.de>
3019 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3020 (dump_stmt_cost): Declare.
3021 (add_stmt_cost): Dump cost we add.
3022 (add_stmt_costs): New function.
3023 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3025 (vect_analyze_stmt): Adjust prototype.
3026 (vectorizable_condition): Likewise.
3027 (vectorizable_live_operation): Likewise.
3028 (vectorizable_reduction): Likewise.
3029 (vectorizable_induction): Likewise.
3030 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3031 cost vector to pass to vectorizable_ and record afterwards.
3032 (vect_model_reduction_cost): Take cost vector argument and adjust.
3033 (vect_model_induction_cost): Likewise.
3034 (vectorizable_reduction): Likewise.
3035 (vectorizable_induction): Likewise.
3036 (vectorizable_live_operation): Likewise.
3037 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3038 SLP_TREE_NUMBER_OF_VEC_STMTS.
3039 (vect_analyze_slp_cost_1): Remove.
3040 (vect_analyze_slp_cost): Likewise.
3041 (vect_slp_analyze_node_operations): Take visited args and
3042 a target cost vector. Avoid processing already visited stmt sets.
3043 (vect_slp_analyze_operations): Use a local cost vector to gather
3044 costs and register those of non-discarded instances.
3045 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3046 (vect_schedule_slp_instance): Remove copying of
3047 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3049 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3050 adding cost. Record cost entry location.
3051 (vect_prologue_cost_for_slp_op): Function to compute cost of
3052 a constant or invariant generated for SLP vect in the prologue,
3053 split out from vect_analyze_slp_cost_1.
3054 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3055 (vect_model_promotion_demotion_cost): Likewise.
3056 (vect_model_store_cost): Likewise, make static.
3057 (vect_model_load_cost): Likewise.
3058 (vectorizable_bswap): Add cost vector arg and adjust.
3059 (vectorizable_call): Likewise.
3060 (vectorizable_simd_clone_call): Likewise.
3061 (vectorizable_conversion): Likewise.
3062 (vectorizable_assignment): Likewise.
3063 (vectorizable_shift): Likewise.
3064 (vectorizable_operation): Likewise.
3065 (vectorizable_store): Likewise.
3066 (vectorizable_load): Likewise.
3067 (vectorizable_condition): Likewise.
3068 (vectorizable_comparison): Likewise.
3069 (can_vectorize_live_stmts): Likewise.
3070 (vect_analyze_stmt): Likewise.
3071 (vect_transform_stmt): Adjust calls to vectorizable_*.
3072 * tree-vectorizer.c: Include gimple-pretty-print.h.
3073 (dump_stmt_cost): New function.
3075 2018-05-16 Richard Biener <rguenther@suse.de>
3077 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3078 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3079 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3080 (check_name): New callback.
3081 (dse_classify_store): Track cycles via a visited bitmap of PHI
3082 defs and simplify handling of in-loop and across loop dead stores
3083 and properly fail for loop-variant refs. Handle byte-tracking with
3084 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3087 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3089 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3090 (vect_get_mask_type_for_stmt): Likewise.
3091 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3093 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3094 to determine the statement's vector type and the vector type that
3095 should be used for calculating nunits. Deal with cases in which
3096 the type has to be deferred.
3097 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3098 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3099 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3100 (vect_determine_vf_for_stmt): New functions, split out from...
3101 (vect_determine_vectorization_factor): ...here.
3102 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3103 (vect_get_mask_type_for_stmt): New functions, split out from
3104 vect_determine_vectorization_factor.
3106 2018-05-16 Richard Biener <rguenther@suse.de>
3108 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3109 verify the [VEC_]COND_EXPR embedded comparison.
3111 2018-05-15 Martin Sebor <msebor@redhat.com>
3113 PR tree-optimization/85753
3114 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3115 RECORD_TYPE in addition to ARRAY_TYPE.
3117 2018-05-15 Martin Sebor <msebor@redhat.com>
3120 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3122 2018-05-15 Richard Biener <rguenther@suse.de>
3124 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3125 add by_clobber_p one. Change algorithm to collect all defs
3126 representing uses we need to walk and try reducing them to
3127 a single one before failing.
3128 (dse_dom_walker::dse_optimize_stmt): Adjust.
3130 2018-05-13 Mark Wielaard <mark@klomp.org>
3132 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3133 (size_of_loc_descr): Likewise.
3134 (output_loc_operands): Likewise.
3135 (output_loc_operands_raw): Likewise.
3136 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3137 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3138 (hash_loc_operands): Likewise.
3139 (compare_loc_operands): Likewise.
3141 2018-05-14 Mark Wielaard <mark@klomp.org>
3143 * dwarf2out.c (count_index_addrs): New function.
3144 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3146 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3148 PR tree-optimization/83648
3149 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3150 return value as malloc candidate.
3152 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3155 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3156 param as true in call to suggest_attribute.
3158 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3160 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3161 -mreadonly-in-sdata.
3163 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3165 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3167 (aarch64_crypto_aesd_fused): Likewise.
3169 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3171 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3172 (movsi_aarch64): Likewise.
3173 (load_pairsi): Likewise.
3174 (load_pairdi): Likewise.
3175 (store_pairsi): Likewise.
3176 (store_pairdi): Likewise.
3177 (load_pairsf): Likewise.
3178 (load_pairdf): Likewise.
3179 (store_pairsf): Likewise.
3180 (store_pairdf): Likewise.
3181 (zero_extend): Likewise.
3182 (trunc): Swap alternatives.
3183 (fcvt_target): Add '?' to prefer w over r.
3185 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3188 * config/i386/i386.md: Disallow non-commutative arithmetics in
3189 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3190 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3191 in the peephole2 before it.
3193 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3195 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3196 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3197 (ix86_handle_option): Handle -mcldemote.
3198 * config.gcc: New header.
3199 * config/i386/cldemoteintrin.h: New file.
3200 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3201 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3203 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3204 OPTION_MASK_ISA_CLDEMOTE.
3205 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3206 (ix86_valid_target_attribute_inner_p): Ditto.
3207 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3208 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3209 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3210 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3211 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3213 * config/i386/i386.opt: Add -mcldemote.
3214 * config/i386/x86intrin.h: New header.
3215 * doc/invoke.texi: Add -mcldemote.
3217 2018-05-14 Richard Biener <rguenther@suse.de>
3219 * doc/match-and-simplify.texi: Adjust :s documentation.
3221 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3223 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3224 intended memcpy size.
3225 (REORDER_45): Likewise.
3227 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3229 * sort.cc: New file.
3230 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3231 * vec.c (qsort_chk): Use gcc_qsort.
3232 * Makefile.in (OBJS-libcommon): Add sort.o.
3233 (build/sort.o): New target. Use it...
3234 (BUILD_RTL): ... here, and...
3235 (build/gencfn-macros): ... here, and...
3236 (build/genmatch): ... here.
3238 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3239 Chung-Ju Wu <jasonwucj@gmail.com>
3241 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3242 * config/nds32/nds32-graywolf.md: New file.
3243 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3244 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3246 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3247 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3248 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3249 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3250 * config/nds32/pipelines.md: Include n15 settings.
3252 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3253 Chung-Ju Wu <jasonwucj@gmail.com>
3255 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3256 * config/nds32/nds32-n13.md: New file.
3257 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3258 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3260 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3261 * config/nds32/nds32.md (pipeline_model): Add n13.
3262 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3263 * config/nds32/pipelines.md: Include n13 settings.
3265 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3266 Chung-Ju Wu <jasonwucj@gmail.com>
3268 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3269 * config/nds32/nds32-n10.md: New file.
3270 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3271 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3273 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3274 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3275 * config/nds32/nds32.md (pipeline_model): Add n10.
3276 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3277 * config/nds32/pipelines.md: Include n10 settings.
3279 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3280 Kito Cheng <kito.cheng@gmail.com>
3281 Chung-Ju Wu <jasonwucj@gmail.com>
3283 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3284 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3285 Add enum values for DSP extension instructions.
3286 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3288 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3289 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3291 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3292 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3293 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3294 * config/nds32/nds32-intrinsic.md: Likewise.
3295 * config/nds32/nds32_intrinsic.h: Likewise.
3296 * config/nds32/nds32-md-auxiliary.c: Likewise.
3297 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3298 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3299 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3300 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3301 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3302 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3303 TYPE_DMAC in switch statement.
3304 * config/nds32/nds32.c: New checking and implementation for DSP
3305 extension instructions.
3306 * config/nds32/nds32.h: Likewise.
3307 * config/nds32/nds32.md: Likewise.
3308 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3309 * config/nds32/predicates.md: Implement new predicates for DSP
3312 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3314 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3315 Reformat alternatives and attributes so it is easier to identify
3316 which constraints/attributes go with which instruction.
3317 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3318 (mov<mode>_softfloat32, FMOVE64): Likewise.
3319 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3320 (mov<mode>_softfloat64, FMOVE64): Likewise.
3322 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3324 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3326 (Basic PowerPC Built-in Functions): The new name of the
3327 subsection previously known as "PowerPC Built-in Functions".
3328 (Basic PowerPC Built-in Functions Available on all Configurations):
3330 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3331 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3332 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3333 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3335 2018-05-11 Martin Jambor <mjambor@suse.cz>
3338 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3341 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3344 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3346 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3348 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3349 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3350 (ix86_handle_option): Handle -mwaitpkg.
3351 * config.gcc: New header.
3352 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3353 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3354 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3356 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3357 OPTION_MASK_ISA_WAITPKG.
3358 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3359 (ix86_option_override_internal): Add PTA_WAITPKG.
3360 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3361 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3362 IX86_BUILTIN_TPAUSE.
3363 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3364 __builtin_ia32_umwait and __builtin_ia32_tpause.
3365 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3366 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3367 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3368 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3369 UNSPECV_TPAUSE): New.
3370 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3371 * config/i386/i386.opt: Add -mwaitpkg.
3372 * config/i386/waitpkgintrin.h: New file.
3373 * config/i386/x86intrin.h: New header.
3374 * doc/invoke.texi: Add -mwaitpkg.
3376 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3379 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3381 (cortex-m0): Use armv6s-m isa.
3382 (cortex-m0plus): Likewise.
3383 (cortex-m1): Likewise.
3384 (cortex-m0.small-multiply): Likewise.
3385 (cortex-m0plus.small-multiply): Likewise.
3386 (cortex-m1.small-multiply): Likewise.
3388 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3389 Jakub Jelinek <jakub@redhat.com>
3391 PR tree-optimization/85692
3392 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3393 source permute as well.
3395 2018-05-11 Martin Liska <mliska@suse.cz>
3398 * doc/extend.texi: Document LLVM style format for no_sanitize
3401 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3403 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3404 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3405 (mode_supports_vsx_dform_quad): Likewise.
3406 (mode_supports_vmx_dform): Move these functions to be next to the
3407 other mode_supports functions.
3408 (mode_supports_dq_form): Likewise.
3409 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3410 mode_supports_dq_form.
3411 (reg_offset_addressing_ok_p): Likewise.
3412 (offsettable_ok_by_alignment): Likewise.
3413 (rs6000_legitimate_offset_address_p): Likewise.
3414 (legitimate_lo_sum_address_p): Likewise.
3415 (rs6000_legitimize_address): Likewise.
3416 (rs6000_legitimize_reload_address): Likewise.
3417 (rs6000_secondary_reload_inner): Likewise.
3418 (rs6000_preferred_reload_class): Likewise.
3419 (rs6000_output_move_128bit): Likewise.
3421 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3423 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3424 Generate SImode target register for null target.
3425 <case IX86_BUILTIN_XGETBV>: Ditto.
3426 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3427 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3429 2018-05-10 Carl Love <cel@us.ibm.com>
3431 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3432 dcbtt and dcbtstt if operands[2] is 0.
3434 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3437 * config/i386/sse.md (usadv64qi): New expander.
3439 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3441 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3442 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3443 -maltivec=be support.
3444 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3445 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3446 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3447 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3448 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3449 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3450 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3451 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3452 altivec_vsumsws): Adjust.
3453 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3454 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3455 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3457 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3458 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3459 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3460 (altivec_lve<VI_char>x): Delete expand.
3461 (*altivec_lve<VI_char>x_internal): Rename to...
3462 (altivec_lve<VI_char>x): ... this.
3463 (altivec_lvxl_<mode>): Delete expand.
3464 (*altivec_lvxl_<mode>_internal): Rename to ...
3465 (altivec_lvxl_<mode>): ... this.
3466 (altivec_stvxl_<mode>): Delete expand.
3467 (*altivec_stvxl_<mode>_internal): Rename to ...
3468 (altivec_stvxl_<mode>): ... this.
3469 (altivec_stve<VI_char>x): Delete expand.
3470 (*altivec_stve<VI_char>x_internal): Rename to ...
3471 (altivec_stve<VI_char>x): ... this.
3472 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3473 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3474 reduc_plus_scal_<mode>): Adjust.
3475 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3477 (rs6000_cpu_cpp_builtins): Adjust.
3478 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3479 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3480 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3481 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3482 -maltivec=be support.
3483 (rs6000_split_vec_extract_var): Adjust.
3484 (rs6000_split_v4si_init): Adjust.
3485 (swap_selector_for_mode): Delete.
3486 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3487 altivec_expand_stvex_be): Delete.
3488 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3489 -maltivec=be support.
3490 (rs6000_gimple_fold_builtin): Ditto.
3491 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3493 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3494 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3495 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3496 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3497 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3498 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3499 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3500 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3501 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3502 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3503 anonymous split): Adjust.
3504 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3505 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3507 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3509 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3510 when --with-gxx-include-dir is also specified.
3511 * configure: Regenerate.
3513 2018-05-09 Jim Wilson <jimw@sifive.com>
3516 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3517 * config/riscv/t-withmultilib: New.
3518 * config/riscv/withmultilib.h: New.
3519 * doc/install.texi: Document RISC-V --with-multilib-list support.
3521 2018-05-09 Richard Biener <rguenther@suse.de>
3523 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3525 (vect_bb_vectorization_profitable_p): Adjust. Compute
3526 actual scalar cost using the cost vector and the add_stmt_cost
3529 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3531 PR rtl-optimization/85645
3532 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3533 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3535 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3537 PR rtl-optimization/85645
3538 * shrink-wrap.c (spread_components): Return a boolean saying if
3539 anything was changed.
3540 (try_shrink_wrapping_separate): Iterate spread_components until
3541 nothing changes anymore.
3543 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3545 PR rtl-optimization/85645
3546 * regrename.c (build_def_use): Also kill the chains that include the
3547 destination of a REG_CFA_REGISTER note.
3549 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3551 PR rtl-optimization/85645
3552 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3553 insn that has a REG_CFA_REGISTER note.
3555 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3557 * cfgexpand.c (expand_clobber): New function.
3558 (expand_gimple_stmt_1): Use it.
3559 * tree-vect-stmts.c (vect_clobber_variable): New function,
3561 (vectorizable_simd_clone_call): ...here.
3562 (vectorizable_store): Emit a clobber either side of an
3563 IFN_STORE_LANES sequence.
3564 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3566 2018-05-09 Tom de Vries <tom@codesourcery.com>
3569 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3570 (define_insn "trap_if_false"): Add exit after trap.
3572 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3574 PR rtl-optimization/85638
3575 * bb-reorder.c: Include common/common-target.h.
3576 (create_forwarder_block): New function extracted from...
3577 (fix_up_crossing_landing_pad): ...here. Rename into...
3578 (dw2_fix_up_crossing_landing_pad): ...this.
3579 (sjlj_fix_up_crossing_landing_pad): New function.
3580 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3581 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3582 from both partitions and exit the loop after one iteration.
3584 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3587 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3589 (Basic PowerPC Built-in Functions): The new name of the
3590 subsection previously known as "PowerPC Built-in Functions".
3591 (Basic PowerPC Built-in Functions Available on all Configurations):
3593 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3595 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3597 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3599 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3602 2018-05-08 Jim Wilson <jimw@sifive.com>
3604 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3605 (LD_EMUL_SUFFIX): New.
3606 (LINK_SPEC): Use it.
3608 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3610 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3612 (Basic PowerPC Built-in Functions): The new name of the
3613 subsection previously known as "PowerPC Built-in Functions".
3614 (Basic PowerPC Built-in Functions Available on all Configurations):
3616 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3618 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3620 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3622 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3625 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3628 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3629 after cmpelim optimization.
3631 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3633 * config.gcc: Support "goldmont".
3634 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3635 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3637 * config/i386/i386.c (m_GOLDMONT): Define.
3638 (processor_target_table): Add "goldmont".
3639 (PTA_GOLDMONT): Define.
3640 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3641 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3642 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3643 (fold_builtin_cpu): Add "goldmont".
3644 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3645 (ix86_option_override_internal): Add "goldmont".
3646 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3647 (processor_type): Add PROCESSOR_GOLDMONT.
3648 * config/i386/i386.md: Add CPU "glm".
3649 * config/i386/glm.md: New file.
3650 * config/i386/x86-tune.def: Add m_GOLDMONT.
3651 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3653 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3656 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3658 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3659 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3660 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3663 * config/i386/i386.c (ix86_fold_builtin): Handle
3664 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3667 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3668 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3670 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3673 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3674 (check_arch): Likewise.
3675 (check_fpu): Return the result rather than printing it.
3676 (end arch): Fix operator precedence.
3677 (end cpu): Likewise.
3678 (END): Print the result from check_fpu.
3680 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3681 Alan Hayward <alan.hayward@arm.com>
3682 David Sherwood <david.sherwood@arm.com>
3684 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3685 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3686 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3687 (*fcmuo<mode>_and): New patterns.
3689 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3691 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3692 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3693 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3694 (cmp_op, sve_imm_con): New code attributes.
3695 (SVE_COND_INT_CMP, imm_con): Delete.
3696 (cmp_op): Remove above unspecs from int attribute.
3697 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3699 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3700 comparison-specific unspecs.
3701 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3702 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3703 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3704 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3705 (*vec_fcm<cmp_op><mode>): Rename to...
3706 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3707 (*vec_fcmuo<mode>): Rename to...
3708 (*fcmuo<mode>): ...this and adjust likewise.
3709 (*pred_fcm<cmp_op><mode>): New pattern.
3710 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3711 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3713 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3715 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3716 (aarch64_emit_sve_predicated_cond): New function.
3717 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3718 (aarch64_emit_unspec_cond_or): Replace with...
3719 (aarch64_emit_sve_or_conds): ...this new function. Use
3720 aarch64_emit_sve_ptrue_op for the individual comparisons and
3721 aarch64_emit_binop to OR them together.
3722 (aarch64_emit_inverted_unspec_cond): Replace with...
3723 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3724 aarch64_emit_sve_ptrue_op for the comparison and
3725 aarch64_emit_unop to invert the result.
3726 (aarch64_expand_sve_vec_cmp_float): Update after the above
3727 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3729 2018-05-07 Nathan Sidwell <nathan@acm.org>
3731 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3732 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3733 (Backwards Compatibility): Likewise.
3735 2018-05-07 Luis Machado <luis.machado@linaro.org>
3739 2018-05-07 Luis Machado <luis.machado@linaro.org>
3741 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3742 <prefetch_dynamic_strides>: New const bool field.
3743 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3744 prefetch_dynamic_strides.
3745 (exynosm1_prefetch_tune): Likewise.
3746 (thunderxt88_prefetch_tune): Likewise.
3747 (thunderx_prefetch_tune): Likewise.
3748 (thunderx2t99_prefetch_tune): Likewise.
3749 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3751 (aarch64_override_options_internal): Update to set
3752 PARAM_PREFETCH_DYNAMIC_STRIDES.
3753 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3754 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3755 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3756 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3757 prefetch-dynamic-strides setting.
3759 2018-05-07 Luis Machado <luis.machado@linaro.org>
3761 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3762 <minimum_stride>: New const int field.
3763 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3764 minimum_stride field.
3765 (exynosm1_prefetch_tune): Likewise.
3766 (thunderxt88_prefetch_tune): Likewise.
3767 (thunderx_prefetch_tune): Likewise.
3768 (thunderx2t99_prefetch_tune): Likewise.
3769 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3770 (aarch64_override_options_internal): Update to set
3771 PARAM_PREFETCH_MINIMUM_STRIDE.
3772 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3773 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3774 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3775 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3776 stride is constant and is below the minimum stride threshold.
3778 2018-05-07 Luis Machado <luis.machado@linaro.org>
3780 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3783 2018-05-07 Luis Machado <luis.machado@linaro.org>
3785 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3786 <prefetch_dynamic_strides>: New const bool field.
3787 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3788 prefetch_dynamic_strides.
3789 (exynosm1_prefetch_tune): Likewise.
3790 (thunderxt88_prefetch_tune): Likewise.
3791 (thunderx_prefetch_tune): Likewise.
3792 (thunderx2t99_prefetch_tune): Likewise.
3793 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3795 (aarch64_override_options_internal): Update to set
3796 PARAM_PREFETCH_DYNAMIC_STRIDES.
3797 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3798 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3799 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3800 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3801 prefetch-dynamic-strides setting.
3803 2018-05-07 Luis Machado <luis.machado@linaro.org>
3805 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3806 <minimum_stride>: New const int field.
3807 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3808 minimum_stride field.
3809 (exynosm1_prefetch_tune): Likewise.
3810 (thunderxt88_prefetch_tune): Likewise.
3811 (thunderx_prefetch_tune): Likewise.
3812 (thunderx2t99_prefetch_tune): Likewise.
3813 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3814 (aarch64_override_options_internal): Update to set
3815 PARAM_PREFETCH_MINIMUM_STRIDE.
3816 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3817 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3818 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3819 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3820 stride is constant and is below the minimum stride threshold.
3822 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3825 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3826 the type is addressable. Don't force op into register if it has
3829 2018-05-05 Roland McGrath <mcgrathr@google.com>
3832 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3833 any section for which we don't know a specific type it should have,
3834 regardless of name. Previously this was done only for the exact
3835 names ".init_array", ".fini_array", and ".preinit_array".
3836 (default_elf_asm_named_section): Add comment about
3837 relationship with default_section_type_flags and SECTION_NOTYPE.
3838 (get_section): Don't consider it a type conflict if one side has
3839 SECTION_NOTYPE and the other doesn't, as long as neither has the
3840 SECTION_BSS et al used in the default_section_type_flags logic.
3842 2018-05-05 Tom de Vries <tom@codesourcery.com>
3845 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3846 (workaround_barsyncs): New function.
3847 (nvptx_reorg): Use workaround_barsyncs.
3848 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3849 (define_expand "nvptx_membar_cta"): New define_expand.
3850 (define_insn "*nvptx_membar_cta"): New insn.
3852 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3854 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3855 To improve optimization opportunities.
3856 * builtin-types.def: The new needed builtin types for the above.
3858 2018-05-04 Richard Biener <rguenther@suse.de>
3860 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3861 * gimple-ssa-store-merging.c
3862 (imm_store_chain_info::output_merged_store): Remove redundant create,
3863 release split_store vector contents on failure.
3864 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3865 scalar stmt vector on cache hit.
3867 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3869 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3871 * config.gcc (powerpc-xilinx-eabi*): Remove.
3872 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3874 (fusion_addis_mem_combo_load): Ditto.
3875 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3877 (rs6000_cpu_cpp_builtins): Ditto.
3878 * config/rs6000/rs6000-linux.c
3879 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3880 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3881 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3883 (rs6000_setup_reg_addr_masks): Ditto.
3884 (rs6000_init_hard_regno_mode_ok): Ditto.
3885 (rs6000_option_override_internal): Ditto.
3886 (legitimate_lo_sum_address_p): Ditto.
3887 (rs6000_legitimize_address): Ditto.
3888 (rs6000_legitimize_reload_address): Ditto.
3889 (rs6000_legitimate_address_p): Ditto.
3890 (abi_v4_pass_in_fpr): Ditto.
3891 (setup_incoming_varargs): Ditto.
3892 (rs6000_gimplify_va_arg): Ditto.
3893 (rs6000_split_multireg_move): Ditto.
3894 (rs6000_savres_strategy): Ditto.
3895 (rs6000_emit_prologue_components): Ditto.
3896 (rs6000_emit_epilogue_components): Ditto.
3897 (rs6000_emit_prologue): Ditto.
3898 (rs6000_emit_epilogue): Ditto.
3899 (rs6000_elf_file_end): Ditto.
3900 (rs6000_function_value): Ditto.
3901 (rs6000_libcall_value): Ditto.
3902 * config/rs6000/rs6000.h: Ditto.
3903 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3904 (TARGET_MINMAX): ... this. New.
3905 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3906 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3907 (*movsi_internal1_single): Delete.
3908 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3909 mfpu=, mxilinx-fpu): Delete.
3910 * config/rs6000/singlefp.h: Delete.
3911 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3912 * config/rs6000/t-rs6000: Ditto.
3913 * config/rs6000/t-xilinx: Delete.
3914 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3915 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3916 (VStype_simple): Delete.
3917 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3918 * config/rs6000/xfpu.h: Delete.
3919 * config/rs6000/xfpu.md: Delete.
3920 * config/rs6000/xilinx.h: Delete.
3921 * config/rs6000/xilinx.opt: Delete.
3922 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3923 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3925 2018-05-04 Tom de Vries <tom@codesourcery.com>
3928 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3931 2018-05-04 Richard Biener <rguenther@suse.de>
3934 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3935 (expand_complex_div_wide): Likewise.
3936 (expand_complex_operations_1): Likewise.
3937 (expand_complex_libcall): Preserve EH info of the original stmt.
3938 (tree_lower_complex): Handle removed blocks.
3939 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3940 on complex multiplication and division libcall builtins.
3942 2018-05-04 Richard Biener <rguenther@suse.de>
3945 * fold-const.c (negate_expr_p): Restrict negation of operand
3946 zero of a division to when we know that can happen without
3948 (fold_negate_expr_1): Likewise.
3950 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3953 * real.h (real_nextafter): Declare.
3954 * real.c (real_nextafter): New function.
3955 * fold-const-call.c (fold_const_nextafter): New function.
3956 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3957 CASE_CFN_NEXTTOWARD.
3958 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3959 even when arg1_mode is different from arg0_mode.
3961 2018-05-03 Nathan Sidwell <nathan@acm.org>
3963 * doc/extend.texi (Deprecated Features): Remove
3965 (Backwards Compatibility): Likewise.
3966 * doc/invoke.texi (C++ Language Options): Likewise.
3967 (C++ Dialect Options): Likewise.
3969 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3972 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3973 _mm512_mask_mullox_epi64): New intrinsics.
3975 2018-05-03 Tom de Vries <tom@codesourcery.com>
3978 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3979 dump files): Add offload-tree.
3981 2018-05-03 Richard Biener <rguenther@suse.de>
3983 PR tree-optimization/85615
3984 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3985 to loops not nested in BBs loop father to avoid creating multi-entry
3988 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3990 PR tree-optimization/70291
3991 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3992 arguments. Change return type to tree. Emit libcall as a new
3993 statement rather than replacing existing one when inplace_p is true.
3994 (expand_complex_multiplication_components): New function.
3995 (expand_complex_multiplication): Expand floating-point complex
3996 multiplication using the above.
3997 (expand_complex_division): Rename inner_type parameter to type.
3998 Update expand_complex_libcall call-site.
3999 (expand_complex_operations_1): Update expand_complex_multiplication
4000 and expand_complex_division call-sites.
4002 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4005 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4006 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4007 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4008 the highest significant bit of the shift count mask is clear. In
4009 check whether and[sq]i3 is needed verify that all significant bits
4010 of the shift count other than the highest are set.
4012 2018-05-02 Tom de Vries <tom@codesourcery.com>
4015 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4016 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4017 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4018 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4019 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4020 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4021 __builtin_goacc_parlevel_size.
4023 2018-05-02 Richard Biener <rguenther@suse.de>
4025 PR tree-optimization/85597
4026 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4027 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4030 2018-05-02 Tom de Vries <tom@codesourcery.com>
4033 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4034 dump files): Add ltrans-tree.
4036 2018-05-02 Tom de Vries <tom@codesourcery.com>
4039 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4040 dump files): Add wpa-ipa.
4042 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4044 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4045 powerpc*-*-linux*paired* target.
4046 * config/rs6000/750cl.h: Delete.
4047 * config/rs6000/paired.h: Delete.
4048 * config/rs6000/paired.md: Delete.
4049 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4051 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4052 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4053 comment. Remove paired float support.
4054 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4055 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4057 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4058 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4060 * config/rs6000/rs6000.c: Remove paired float support.
4061 (paired_expand_vector_init, paired_expand_vector_move,
4062 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4063 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4064 paired_expand_builtin, paired_expand_predicate_builtin,
4065 paired_init_builtins): Delete.
4066 * config/rs6000/rs6000.h: Remove paired float support.
4067 * config/rs6000/rs6000.md: Remove paired float support.
4068 (move_from_CR_ov_bit): Delete.
4069 * config/rs6000/rs6000.opt (mpaired): Delete.
4070 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4071 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4073 2018-05-02 Richard Biener <rguenther@suse.de>
4076 * gimplify.c (gimplify_save_expr): When in SSA form allow
4077 SAVE_EXPRs to compute to SSA vars.
4079 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4082 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4083 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4084 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4085 clobber operands[2], instead use a new pseudo. Formatting fixes.
4087 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4089 PR tree-optimization/85586
4090 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4091 exit early for statements in the same group if the accesses are
4094 2018-05-02 Tom de Vries <tom@codesourcery.com>
4097 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4100 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4102 PR tree-optimization/85143
4103 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4105 2018-05-01 Tom de Vries <tom@codesourcery.com>
4108 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4109 not found" error message.
4111 2018-05-01 Tom de Vries <tom@codesourcery.com>
4114 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4115 * vec.c (test_ordered_remove_if): New function.
4116 (vec_c_tests): Call test_ordered_remove_if.
4117 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4118 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4119 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4120 VEC_ORDERED_REMOVE_IF.
4122 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4124 PR tree-optimization/82665
4125 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4126 pointer subtraction where arguments come from a memchr call.
4128 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4130 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4131 --push-state --as-needed and --pop-state instead of --as-needed and
4132 --no-as-needed if ld supports it.
4133 * configure: Regenerated.
4136 * doc/install.texi2html: Replace _002d with - and _002a with * in
4137 generated html files using sed.
4139 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4142 * gcc-rich-location.c (blank_line_before_p): New function.
4143 (use_new_line): New function.
4144 (gcc_rich_location::add_fixit_insert_formatted): New function.
4145 * gcc-rich-location.h
4146 (gcc_rich_location::add_fixit_insert_formatted): New function.
4148 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4150 * selftest.c (assert_streq): Rename "expected" and "actual" to
4151 "val1" and "val2". Extend NULL-handling to cover both inputs
4152 symmetrically, while still requiring both to be non-NULL for a pass.
4153 * selftest.h (assert_streq): Rename "expected" and "actual" to
4155 (ASSERT_EQ): Likewise.
4156 (ASSERT_EQ_AT): Likewise.
4157 (ASSERT_KNOWN_EQ): Likewise.
4158 (ASSERT_KNOWN_EQ_AT): Likewise.
4159 (ASSERT_NE): Likewise.
4160 (ASSERT_MAYBE_NE): Likewise.
4161 (ASSERT_MAYBE_NE_AT): Likewise.
4162 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4163 the assertion to pass.
4164 (ASSERT_STREQ_AT): Likewise.
4166 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4168 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4169 interaction with -pie.
4171 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4173 * selftest.h: Fix alphabetization of per-source-file selftest
4176 2018-04-30 Jason Merrill <jason@redhat.com>
4178 PR c++/61982 - dead stores to destroyed objects.
4179 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4182 2018-04-30 Jason Merrill <jason@redhat.com>
4184 * tree.c (build_clobber): New.
4185 * tree.h: Declare it.
4186 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4188 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4190 * diagnostic-show-locus.c (layout::layout): Update for
4191 location_get_source_line returning a char_span.
4192 (struct char_span): Move to input.h.
4193 (struct correction): Update for fields in char_span becoming
4195 (struct source_line): Update for location_get_source_line
4196 returning a char_span.
4197 (layout::print_line): Likewise.
4198 * edit-context.c (edited_file::print_content): Likewise.
4199 (edited_file::print_diff_hunk): Likewise.
4200 (edited_file::print_run_of_changed_lines): Likewise.
4201 (edited_file::get_num_lines): Likewise.
4202 (edited_line::edited_line): Likewise.
4203 * final.c (asm_show_source): Likewise.
4204 * input.c (location_get_source_line): Convert return type
4205 from const char * to char_span, losing the final "line_len"
4207 (dump_location_info): Update for the above.
4208 (get_substring_ranges_for_loc): Likewise. Use a char_span
4209 when handling the literal within the line.
4210 (test_reading_source_line): Update for location_get_source_line
4211 returning a char_span.
4212 * input.h (class char_span): Move here from
4213 diagnostic-show-locus.c, converting from a struct to a class.
4214 Make data members private.
4215 (char_span::operator bool): New.
4216 (char_span::length): New.
4217 (char_span::get_buffer): New.
4218 (char_span::operator[]): New.
4219 (char_span::subspan): Make const.
4220 (char_span::xstrdup): New.
4221 (location_get_source_line): Convert return type from const char *
4222 to char_span, losing the final "line_size" param.
4224 2018-04-30 Jan Hubicka <jh@suse.cz>
4226 * lto-wrapper.c (ltrans_priorities): New static var.
4227 (cmp_priority): New.
4228 (run_gcc): Read priorities and if doing parallel build order
4229 the Makefile by them.
4231 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4233 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4235 2018-04-30 Richard Biener <rguenther@suse.de>
4237 * tree-cfg.c (verify_address): Remove base argument, add
4238 flag whether to check TREE_ADDRESSABLE and do that.
4239 (verify_expr): Remove.
4240 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4241 (verify_gimple_assign_single): Likewise.
4242 (verify_gimple_switch): Likewise.
4243 (verify_expr_location_1): Dereference tp once. Add (disabled)
4244 piece from verify_expr.
4245 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4247 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4249 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4251 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4253 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4254 (small_data_pattern): Likewise.
4255 (arc_rewrite_small_data): Likewise.
4256 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4257 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4258 (get_symbol_alignment): New function.
4259 (legitimate_small_data_address_p): Likewise.
4260 (legitimate_scaled_address): Update, call
4261 legitimate_small_data_address_p.
4262 (output_sdata): New static variable.
4263 (arc_print_operand): Update how we handle small data operands.
4264 (arc_print_operand_address): Likewise.
4265 (arc_legitimate_address_p): Update, use
4266 legitimate_small_data_address_p.
4267 (arc_rewrite_small_data_p): Remove.
4268 (arc_rewrite_small_data_1): Likewise.
4269 (arc_rewrite_small_data): Likewise.
4270 (small_data_pattern): Likewise.
4271 (compact_sda_memory_operand): Update to use
4272 legitimate_small_data_address_p and get_symbol_alignment.
4273 (prepare_move_operands): Don't rewite sdata pattern.
4274 (prepare_extend_operands): Remove.
4275 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4277 (zero_extendqisi2): Likewise.
4278 (zero_extendhisi2): Likewise.
4279 (extendqihi2): Likewise.
4280 (extendqisi2): Likewise.
4281 (extendhisi2): Likewise.
4285 * config/arc/constraints.md (Usd): Change it to memory constraint.
4287 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4289 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4290 as source of std instructions.
4291 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4292 allow 6-bit constants as source for store instructions.
4293 (movdi_insn): Update instruction pattern to allow 6-bit constants
4294 as source for store instructions.
4296 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4298 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4300 2018-04-30 Nathan Sidwell <nathan@acm.org>
4301 Sandra Loosemore <sandra@codesourcery.com>
4303 * dumpfile.c (dump_open): Allow '-' for stdout.
4304 * doc/invoke.texi (Developer Options): Document dump filename
4305 determination early. Document stdin/stdout selection.
4307 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4309 Microblaze Target: PIC data text relative
4311 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4312 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4314 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4315 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4316 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4317 New addressing mode for data-text relative position indepenedent code.
4318 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4319 'ADDRESS_SYMBOLIC_TXT_REL'.
4320 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4321 (microblaze_legitimate_pic_operand): Exclude function calls from
4322 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4323 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4325 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4326 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4327 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4328 for 'address + offset'.
4329 (microblaze_expand_prologue): Add new function prologue call for
4331 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4332 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4333 table in case of TARGET_PIC_DATA_TEXT_REL.
4334 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4335 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4336 Add new macros 'UNSPEC_TEXT',
4337 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4338 + exclude function calls from 'UNSPEC_PLT' in case of data text
4340 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4341 new target hook for generating address diff vector tables in case of
4343 * doc/tm.texi : Regenerate.
4344 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4345 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4346 of addr diff vector generation.
4347 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4348 target hook definition.
4349 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4350 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4351 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4352 option for data text relative.
4354 2018-04-30 Richard Biener <rguenther@suse.de>
4356 * tree-chrec.h (evolution_function_is_constant_p): Remove
4358 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4360 2018-04-30 Richard Biener <rguenther@suse.de>
4363 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4365 2018-04-30 Richard Biener <rguenther@suse.de>
4367 PR tree-optimization/28364
4368 PR tree-optimization/85275
4369 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4370 copying first exit test.
4372 2018-04-28 Mark Wielaard <mark@klomp.org>
4374 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4376 (dwarf_AT): Handle DW_AT_addr_base.
4377 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4379 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4382 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4383 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4384 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4385 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4387 2018-04-28 Richard Biener <rguenther@suse.de>
4389 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4390 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4391 to reflect use. Only add interesting stmts.
4393 2018-04-27 Martin Jambor <mjambor@suse.cz>
4396 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4397 the jump function allows for passing through aggregate values.
4399 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4401 * input.h (in_system_header_at): Convert from macro to inline
4403 (from_macro_expansion_at): Likewise.
4404 (from_macro_definition_at): Likewise.
4406 2018-04-27 Jeff Law <law@redhat.com>
4408 * config.gcc: Mark tile* targets as deprecated/obsolete.
4410 2018-04-27 Richard Biener <rguenther@suse.de>
4412 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4415 2018-04-27 Richard Biener <rguenther@suse.de>
4417 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4419 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4421 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4422 with Yd constraint. Set "preferred_for_speed" attribute from
4423 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4425 (*movdi_internal): Ditto.
4426 (movti_interunit splitters): Remove
4427 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4428 (movdi_interunit splitters): Ditto.
4429 * config/i386/constraints.md (Ye): Remove.
4430 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4432 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4435 * config/aarch64/constraints.md (Usg): Limit to 31.
4438 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4440 PR tree-optimization/85529
4441 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4442 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4443 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4444 zero extension or masking of the MSB bit.
4445 (optimize_range_tests): Add FIRST_BB argument, pass it through
4446 to optimize_range_tests_var_bound.
4447 (maybe_optimize_range_tests, reassociate_bb): Adjust
4448 optimize_range_tests callers.
4450 2018-04-26 Richard Biener <rguenther@suse.de>
4451 Jakub Jelinek <jakub@redhat.com>
4453 * cgraph.h (symbol_table): Just declare debug method here.
4454 * symtab.c (symbol_table::debug): Define.
4456 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4458 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4460 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4462 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4463 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4464 (*movdi_internal): Substitute Yi and Yj constraint with x
4465 and Ym and Yn constraint with y constraint. Update "isa"
4466 attribute and set "preferred_for_speed" attribute from
4467 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4468 (*movsi_internal): Ditto.
4469 (*movdf_internal): Ditto.
4470 (*movsf_internal): Ditto.
4471 (*zero_extendsidi2): Ditto.
4472 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4473 (sse2_loadld): Ditto.
4474 (*vec_extract<ssevecmodelower>_0): Ditto.
4475 (*vec_extractv4si_0_zext_sse4): Ditto.
4476 (vec_concatv2di): Ditto.
4477 (*vec_dup<mode>): Ditto.
4478 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4479 * config/i386/constraints.md (Yi): Remove.
4484 2018-04-26 Nathan Sidwell <nathan@acm.org>
4486 * dumpfile.c (dump_open): New.
4487 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4488 (dump_finish): Detect stdio/stderr by value not name.
4490 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4492 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4494 2018-04-26 Tom de Vries <tom@codesourcery.com>
4497 * config/nvptx/nvptx.c (verify_neutering_jumps)
4498 (verify_neutering_labels): New function
4499 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4501 2018-04-26 Tom de Vries <tom@codesourcery.com>
4504 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4505 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4508 2018-04-26 Richard Biener <rguenther@suse.de>
4509 Tom de Vries <tom@codesourcery.com>
4512 * lto-streamer-out.c (output_function): Fixup loops if required to match
4513 discovery done in the reader.
4515 2018-04-26 Richard Biener <rguenther@suse.de>
4517 PR tree-optimization/85116
4518 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4519 have a loop exit from the single latch predecessor. Remove
4520 case of header with just condition.
4521 (ch_base::copy_headers): Exclude infinite loops from any
4523 (pass_ch::execute): Record exits.
4525 2018-04-26 Richard Biener <rguenther@suse.de>
4527 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4528 prologue cost vector and pass it to vect_get_load_cost.
4529 (vect_get_peeling_costs_all_drs): Likewise.
4530 (vect_peeling_hash_get_lowest_cost): Likewise.
4531 (vect_enhance_data_refs_alignment): Likewise.
4533 2018-04-26 Richard Biener <rguenther@suse.de>
4536 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4537 checking of integer<->pointer conversions.
4538 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4539 sign-/zero-extending pointer types.
4540 (expand_omp_for_static_chunk): Likewise.
4542 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4543 Jean Lee <xiaoyur347@gmail.com>
4545 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4546 (TARGET_ASAN_SHADOW_OFFSET): Define.
4547 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4548 true for -fsanitize=address.
4550 2018-04-25 Mark Wielaard <mark@klomp.org>
4552 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4555 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4557 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4558 than "alu", remove explicit "memory" and "imm_disp" attributes.
4559 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4562 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4563 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4564 gen_lowpart_no_emit.
4566 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4569 * config/i386/i386.c (ix86_expand_builtin): Change memory
4570 operand to XI, extend p0 to Pmode.
4571 * config/i386/i386.md: Change unspec volatile and operand
4572 1 mode to XI, change operand 0 mode to P.
4574 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4576 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4577 GET_MODE_MASK before any checking.
4578 (nds32_can_use_bset_p): Likewise.
4579 (nds32_can_use_btgl_p): Likewise.
4581 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4583 * config/nds32/nds32-doubleword.md: New define_split pattern for
4584 illegal register number.
4586 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4588 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4590 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4592 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4594 2018-04-25 Richard Biener <rguenther@suse.de>
4596 * lto-streamer.h (LTO_major_version): Bump to 8.
4598 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4600 * BASE-VER: Set to 9.0.0.
4602 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4604 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4605 in __abskf2 and __powikf2.
4607 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4610 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4611 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4612 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4613 Use the above on operand 2. Reindent.
4614 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4616 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4619 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4621 * config/i386/i386.opt (mcet): Removed.
4622 * doc/install.texi: Remove -mcet documentation.
4623 * doc/invoke.texi: Likewise.
4625 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4628 * doc/install.texi: Remove -mcet from bootstrap-cet.
4630 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4633 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4634 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4638 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4639 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4640 containing a CONST_VECTOR.
4642 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4644 * doc/install.texi: Update newlib dependency for nvptx.
4646 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4649 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4650 instead of INTVAL when shifting x left.
4652 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4654 PR tree-optimization/85478
4655 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4656 vect_grouped_store_supported for single element vectors.
4658 2018-04-24 Richard Biener <rguenther@suse.de>
4661 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4662 load cost increase to the case of non-constant step.
4664 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4667 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4668 destination if any_malformed_asm.
4670 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4673 * expr.c (store_field): In the bitfield case, if the value comes from
4674 a function call and is returned in registers by means of a PARALLEL,
4675 do not change the mode of the temporary unless BLKmode and VOIDmode.
4677 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4679 PR rtl-optimization/85423
4680 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4681 dependencies to debug insns when the previous insn is non-debug.
4683 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4685 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4686 enums into a single definition.
4687 (fls): Fix predicates and printing.
4690 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4692 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4693 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4694 and short u6 immediate.
4695 (check_if_valid_sleep_operand): Remove.
4696 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4698 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4700 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4701 flag_always_save_lp condition.
4702 * config/nds32/nds32.opt (malways-save-lp): New option.
4704 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4706 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4707 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4708 * config/nds32/nds32.h
4709 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4710 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4712 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4714 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4715 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4717 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4718 Chung-Ju Wu <jasonwucj@gmail.com>
4720 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4722 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4723 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4725 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4727 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4729 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4731 * config/nds32/nds32-protos.h (nds32_data_alignment,
4732 nds32_local_alignment): Declare.
4733 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4734 nds32_local_alignment): New functions.
4735 (TARGET_CONSTANT_ALIGNMENT): Define.
4736 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4738 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4740 * config/nds32/nds32.c
4741 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4742 (TARGET_MODES_TIEABLE_P): Likewise.
4744 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4746 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4749 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4750 Chung-Ju Wu <jasonwucj@gmail.com>
4752 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4753 for unaligned access.
4754 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4756 * config/nds32/nds32-intrinsic.md: Likewise.
4757 * config/nds32/nds32_intrinsic.h: Likewise.
4758 * config/nds32/nds32.h (nds32_builtins): Likewise.
4759 * config/nds32/nds32.opt (munaligned-access): New option.
4760 * config/nds32/nds32.c (nds32_asm_file_start): Display
4761 flag_unaligned_access status.
4763 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4765 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4766 -mno-relax is present.
4767 * config/riscv/linux.h (LINK_SPEC): Ditto.
4769 2018-04-20 Martin Sebor <msebor@redhat.com>
4772 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4774 (gimple_fold_builtin_stxcpy_chk): Same.
4775 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4777 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4780 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4781 __powikf2 when long double is IEEE 128-bit.
4783 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4785 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4786 step to make sure stack always aligned.
4788 2018-04-20 Carl Love <cel@us.ibm.com>
4791 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4792 size check for arg0.
4794 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4795 Tom de Vries <tom@codesourcery.com>
4798 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4799 Emit insns for calls too.
4800 (nvptx_find_par): Always look for worker-level predecessor insn.
4801 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4803 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4804 (nvptx_process_pars): Propagate frames for calls.
4806 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4809 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4811 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4812 (ix86_handle_option): Don't handle OPT_mibt.
4813 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4815 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4817 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4818 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4819 (ix86_target_macros): Define __CET__ with flag_cf_protection
4820 for -fcf-protection.
4821 * config/i386/i386.c (isa2_opts): Remove -mibt.
4822 * config/i386/i386.h (TARGET_IBT): Removed.
4823 (TARGET_IBT_P): Likewise.
4824 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4825 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4826 * config/i386/i386.opt (mcet): Update help message.
4829 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4830 -mcet as an alias for -mshstk.
4832 2018-04-20 Richard Biener <rguenther@suse.de>
4835 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4836 complexity by forcing a single use of the multiply operand.
4838 2018-04-20 Martin Jambor <mjambor@suse.cz>
4841 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4842 recursion dependency to only apply to non-clones.
4844 2018-04-20 Martin Jambor <mjambor@suse.cz>
4847 * ipa-cp.c (create_specialized_node): Check that clones of
4848 self-recursive edges exist during IPA-CP.
4850 2018-04-19 Toon Moene <toon@moene.org>
4852 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4855 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4857 PR tree-optimization/85467
4858 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4859 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4860 VECTOR_CST element to type.
4862 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4865 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4866 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4867 (builtin_longjmp): Likewise.
4868 (save_stack_nonlocal): New pattern.
4869 (restore_stack_nonlocal): Likewise.
4871 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4874 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4875 Replace ASM_OUTPUT_LABEL with fprintf.
4877 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4880 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4881 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4882 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4883 define __IBT__ and __SHSTK__ for -fcf-protection.
4884 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4886 (ix86_trampoline_init): Likewise.
4887 (x86_output_mi_thunk): Likewise.
4888 (ix86_notrack_prefixed_insn_p): Likewise.
4889 (ix86_option_override_internal): Don't disallow -fcf-protection.
4890 * config/i386/i386.md (rdssp<mode>): Also enable for
4892 (incssp<mode>): Likewise.
4893 (nop_endbr): Likewise.
4894 * config/i386/i386.opt (mcet): Change help message to built-in
4898 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4899 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4900 enable CET built-in functions.
4902 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4904 * common/config/i386/i386-common.c
4905 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4906 OPTION_MASK_ISA_MOVDIRI_UNSET,
4907 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4908 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4909 * config.gcc (movdirintrin.h): New header.
4910 * config/i386/cpuid.h (bit_MOVDIRI,
4911 bit_MOVDIR64B): New bits.
4912 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4914 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4915 (VOID, PVOID, PCVOID)): New function types.
4916 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4917 __builtin_ia32_directstoreu_u64,
4918 __builtin_ia32_movdir64b): New builtins.
4919 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4920 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4922 (ix86_valid_target_attribute_inner_p): Ditto.
4923 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4924 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4925 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4926 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4927 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4928 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4929 (movdiri<mode>, movdir64b_<mode>): New.
4930 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4931 * config/i386/immintrin.h: Include movdirintrin.h.
4932 * config/i386/movdirintrin.h: New file.
4933 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4935 2018-04-19 Richard Biener <rguenther@suse.de>
4938 * cfg.c (clear_bb_flags): When loop state says we have
4939 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4941 2018-04-19 Richard Biener <rguenther@suse.de>
4943 PR tree-optimization/84737
4944 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4945 copying restrict info.
4946 (vect_setup_realignment): Use it.
4947 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4948 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4949 the first DR to all generated stores.
4950 (vectorizable_load): Likewise for loads.
4952 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4954 PR tree-optimization/85446
4955 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4956 the integral and pointer types to have the same precision.
4958 * doc/install.texi: Document --disable-cet being the default and
4961 2018-04-18 Martin Liska <mliska@suse.cz>
4963 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4966 2018-04-18 Martin Liska <mliska@suse.cz>
4969 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4972 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4973 arguments if they are comparable.
4975 2018-04-18 Martin Liska <mliska@suse.cz>
4978 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4981 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4984 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4987 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4988 ENDBR after calling __morestack.
4990 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4993 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4994 by using gcc_base_ver to generate a gcc_driver_version, and use
4995 it when generating GCC_DRIVER_NAME.
4996 * configure: Regenerate.
4998 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5001 * config.gcc: Obsolete powerpc*-*-*spe*.
5003 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5006 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5007 (stabstr_D): Change type of unum from unsigned int to
5008 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5011 2018-04-17 Jim Wilson <jimw@sifive.com>
5014 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5015 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5016 Set arg_pointer_offset after using pretend_args_size.
5018 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5020 PR rtl-optimization/85431
5021 * dse.c (record_store): Ignore zero width stores.
5024 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5025 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5026 __builtin_stack_restore rather than after it.
5027 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5028 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5029 argument instead of virtual_dynamic_stack_rtx.
5031 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5033 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5035 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5036 Add note to error message to explain internal mapping of overloaded
5037 built-in function name to non-overloaded built-in function name.
5038 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5041 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5044 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5045 where the inputs overlap with the output.
5047 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5049 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5050 (=v, v) alternative and explicit "memory" attribute.
5051 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5052 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5054 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5055 "sselog1" type instead of "sselog".
5056 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5057 "sselog". Remove explicit "memory" attribute.
5058 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5059 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5061 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5062 "sselog1" type instead of "sselog". Remove explicit "memory"
5064 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5065 use "sselog1" type instead of "sselog". Remove explicit "memory"
5067 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5068 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5070 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5071 "sselog1" type instead of "sselog". Remove explicit "memory"
5073 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5074 use "sselog1" type instead of "sselog". Remove explicit "memory"
5078 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5081 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5084 2018-04-17 Martin Jambor <mjambor@suse.cz>
5087 * ipa-cp.c (create_specialized_node): Call
5088 expand_all_artificial_thunks if necessary.
5090 2018-04-17 Martin Liska <mliska@suse.cz>
5093 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5094 in message, remote space in between '_G' and '('.
5096 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5099 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5100 avx512f_vmcmp<mode>3<round_saeonly_name>,
5101 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5102 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5103 avx512f_rndscale<mode><round_saeonly_name>,
5104 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5105 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5106 Use %<iptr>2 instead of %2 for -masm=intel.
5107 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5108 avx512f_vcvttss2usi<round_saeonly_name>,
5109 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5111 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5112 avx512f_vcvttsd2usi<round_saeonly_name>,
5113 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5114 Use %q1 instead of %1 for -masm=intel.
5115 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5116 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5117 of %3 for -masm=intel.
5118 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5120 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5122 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5124 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5125 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5127 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5129 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5130 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5131 %g1 and one with %0 and %1.
5132 (avx512er_vmrcp28<mode><round_saeonly_name>,
5133 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5135 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5136 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5137 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5138 of %0 and %{%4%} for -masm=intel.
5139 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5140 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5141 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5142 order of %0 and %{%5%}%{z%} for -masm=intel.
5144 2018-04-17 Jan Hubicka <jh@suse.cz>
5147 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5149 2018-04-17 Martin Liska <mliska@suse.cz>
5152 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5153 for target_clone error message. Make default implementation
5154 clone to be a local declaration.
5155 (separate_attrs): Add new argument and check for an empty
5157 (expand_target_clones): Handle it.
5158 (ipa_target_clone): Make redirection just for target_clones
5161 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5162 Tom de Vries <tom@codesourcery.com>
5165 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5166 tiled basic blocks without omp continue statements.
5168 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5171 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5172 vec_extract expression as having side effects to make sure it gets
5175 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5178 * config/i386/i386.c (get_builtin_code_for_version): Check
5181 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5184 * gcc/config.gcc: Support "skylake".
5185 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5187 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5188 (processor_target_table): Add "skylake".
5189 (ix86_option_override_internal): Add "skylake".
5190 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5191 PROCESSOR_CANNONLAKE.
5192 (get_builtin_code_for_version): Fix priority for
5193 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5194 PROCESSOR_SKYLAKE-AVX512.
5195 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5196 (processor_type): Add PROCESSOR_SKYLAKE.
5198 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5199 Jason Merrill <jason@redhat.com>
5202 * convert.c (convert_to_integer_1): Use direct recursion for
5203 enumeral types and types with a precision less than the number
5204 of bits in their mode.
5206 2018-04-16 Julia Koval <julia.koval@intel.com>
5209 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5210 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5212 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5215 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5216 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5217 and -mno-direct-move.
5219 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5222 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5223 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5224 Ensure negative shifts result in {0}.
5226 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5228 PR rtl-optimization/79916
5229 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5230 regs (if any) to define how to gnerate SD moves when LRA is in
5233 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5235 PR rtl-optimization/85393
5236 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5237 * except.c (expand_dw2_landing_pad_for_region): Make static.
5238 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5239 a label and unconditional jump to old_bb, rather than
5240 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5243 PR rtl-optimization/85376
5244 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5245 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5246 instead of a specific value.
5248 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5249 Bin Cheng <bin.cheng@arm.com>
5251 PR tree-optimization/82965
5252 PR tree-optimization/83991
5253 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5254 by_profile_only parameter.
5255 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5256 information if the loop was predicted to iterate too many times.
5257 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5259 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5262 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5265 2018-04-13 Martin Liska <mliska@suse.cz>
5266 Jakub Jelinek <jakub@redhat.com>
5269 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5270 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5271 * builtins.c (expand_builtin_memory_copy_args): Use
5272 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5273 handle dest_addr == pc_rtx.
5275 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5278 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5279 asked to not generate direct moves.
5280 (fix_trunc<mode>si2_stfiwx): Similar.
5281 (fix_trunc<mode>si2_internal): Similar.
5283 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5286 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5287 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5288 lookup if dest in some wider mode is known to be const0_rtx and
5289 if so, record permanent equivalence for it to be ZERO_EXTEND of
5290 the narrower mode destination.
5292 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5294 * lto-streamer-out.c (output_function): Revert 259346.
5295 * omp-expand.c (expand_oacc_for): Likewise.
5297 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5299 PR rtl-optimization/85354
5300 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5301 * sel-sched.c (sel_global_init): ... here.
5303 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5306 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5307 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5308 mode for PE-COFF targets.
5309 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5310 (i386_pe_asm_lto_end): Likewise.
5311 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5312 (TARGET_ASM_LTO_END): Likewise.
5313 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5314 (i386_pe_asm_lto_start): New function.
5315 (i386_pe_asm_lto_end): Likewise.
5317 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5318 Richard Biener <rguenther@suse.de>
5321 * lto-streamer-out.c (output_function): Fix CFG loop state before
5323 * omp-expand.c (expand_oacc_for): Handle calls to internal
5324 functions like regular functions.
5326 2018-04-12 Richard Biener <rguenther@suse.de>
5329 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5330 for the early LTO debug to properly generate references to it
5331 during DIE emission. Do not re-use that for the skeleton for
5333 (dwarf2out_early_finish): Likewise.
5335 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5338 * config/i386/sse.md
5339 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5340 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5341 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5342 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5343 and output is a reg, avoid creating invalid lowpart subreg, but
5344 instead split into a 512-bit move. Don't split if not AVX512VL,
5345 input is xmm16+ reg and output is a mem.
5346 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5347 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5348 xmm16+ reg and output is a mem.
5350 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5352 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5353 also for flag_dwarf2_cfi_asm.
5355 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5357 PR rtl-optimization/85342
5358 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5359 a bool scalar var inside of the loop instead. Don't try to update
5360 recog_data.operand after failed apply_change_group.
5362 2018-04-12 Tom de Vries <tom@codesourcery.com>
5365 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5366 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5367 array with flexible array member as array without given dimension.
5368 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5369 argument for undefined param to true.
5371 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5374 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5375 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5376 from PowerPC section.
5377 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5378 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5379 help text that is too long.
5380 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5381 help text that is too long.
5382 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5383 help text that is too long.
5385 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5387 * config/alpha/alpha.md (stack_probe_internal): Rename
5388 from "probe_stack". Update all callers.
5390 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5392 PR rtl-optimization/84566
5393 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5394 sched_macro_fuse_insns.
5396 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5399 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5400 (compute_block_dependences): ... from here.
5402 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5404 PR tree-optimization/85331
5405 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5406 from int to HOST_WIDE_INT.
5408 2018-04-11 Martin Jambor <mjambor@suse.cz>
5411 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5412 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5413 not the same as the source val.
5414 (cgraph_edge_brings_value_p): New parameter.
5415 (gather_edges_for_value): Pass destination value to
5416 cgraph_edge_brings_value_p.
5417 (perhaps_add_new_callers): Likewise.
5418 (get_info_about_necessary_edges): Likewise and exclude values brought
5419 only by self-recursive edges.
5420 (create_specialized_node): Redirect only clones of self-calling edges.
5421 (+self_recursive_pass_through_p): New function.
5422 (find_more_scalar_values_for_callers_subset): Use it.
5423 (find_aggregate_values_for_callers_subset): Likewise.
5424 (known_aggs_to_agg_replacement_list): Removed.
5425 (decide_whether_version_node): Re-calculate known constants for all
5426 remaining context clones.
5428 2018-04-11 Richard Biener <rguenther@suse.de>
5431 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5432 from early DWARF output.
5433 (dwarf2out_early_finish): Output line info unconditionally into
5434 early DWARF and add reference to it.
5436 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5439 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5440 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5441 other than V2DFmode using iptr mode attribute.
5442 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5444 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5446 PR rtl-optimization/84659
5447 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5449 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5452 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5454 (output_loc_list): Pass address of a dummy size variable even in the
5455 locview handling loop.
5456 (index_location_lists): Add comment on why skip_loc_list_entry can't
5459 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5462 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5465 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5468 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5469 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5470 and -mstring-compare-inline-limit.
5472 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5475 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5476 for stack clash protection in a register whenever we need it to be in
5479 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5481 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5482 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5484 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5487 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5489 (mlong-double-): Ditto.
5490 * config/rs6000/sysv4.opt (msdata=): Ditto.
5491 (mtls-size=): Ditto.
5493 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5495 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5496 erroneous entries for
5497 "vector int vec_ldl (int, long int *)", and
5498 "vector unsigned int vec_ldl (int, unsigned long int *)".
5499 Add comments and entries for
5500 "vector bool char vec_ldl (int, bool char *)",
5501 "vector bool short vec_ldl (int, bool short *)",
5502 "vector bool int vec_ldl (int, bool int *)",
5503 "vector bool long long vec_ldl (int, bool long long *)",
5504 "vector pixel vec_ldl (int, pixel *)",
5505 "vector long long vec_ldl (int, long long *)",
5506 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5507 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5508 type tree bool_long_long_type_node and correct definition of
5509 bool_V2DI_type_node to make reference to this new type tree.
5510 (rs6000_mangle_type): Replace erroneous reference to
5511 bool_long_type_node with bool_long_long_type_node.
5512 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5513 comments to emphasize sign distinctions for char and int types and
5514 replace RS6000_BTI_bool_long constant with
5515 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5516 use of RS6000_BTI_pixel.
5517 (bool_long_type_node): Remove this macro definition.
5518 (bool_long_long_type_node): New macro definition
5520 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5522 PR rtl-optimization/85300
5523 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5524 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5525 simplify_unary_operation fails.
5527 2018-04-10 Martin Liska <mliska@suse.cz>
5529 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5530 cgraph_edge and ipa_ref.
5532 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5536 * config/i386/sse.md
5537 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5538 computation of the VEC_MERGE selector from mask.
5539 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5540 Fix decoding of the VEC_MERGE selector into mask.
5542 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5544 PR tree-optimization/85286
5545 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5547 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5549 * final.c (final_1): Set insn_last_address as well as
5550 insn_current_address.
5552 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5555 * explow.c (emit_stack_probe): Call validize_mem on memory location
5556 before passing it to gen_probe_stack. Create address operand and
5557 legitimize it for the probe_stack_address case.
5559 2018-04-09 Jan Hubicka <jh@suse.cz>
5562 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5563 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5564 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5566 (free_lang_data): Rebuild type inheritance graph.
5568 2018-04-09 Martin Sebor <msebor@redhat.com>
5570 * invoke.texi (-finline-small-functions): Mention other optimization
5572 (-findirect-inlining, -fpartial-inlining): Same.
5573 (-finline-functions-called-once): Same.
5574 (-freorder-blocks-and-partition): Same.
5576 2018-04-09 Jan Hubicka <jh@suse.cz>
5579 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5580 jumps; choose last target that matches the criteria (i.e.
5581 no partition changes for non-crossing jumps).
5582 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5583 support for redirecting crossing jumps to non-crossing.
5585 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5587 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5588 also for naked functions.
5590 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5592 * config/arc/arc.md (add_shift): New pattern.
5593 (add_shift2): Likewise.
5594 (sub_shift): Likewise.
5595 (sub_shift_cmp0_noout): Likewise.
5596 (compare_si_ashiftsi): Likewise.
5597 (xbfu_cmp0_noout): New combine pattern.
5598 (xbfu_cmp0"): Likewise.
5599 (movsi_set_cc_insn): Place the predicable variant first.
5600 (commutative_binary_cmp0_noout): Remove clobber.
5601 (commutative_binary_cmp0): New pattern.
5602 (noncommutative_binary_cmp0): Likewise.
5603 (noncommutative_binary_cmp0_noout): Likewise.
5604 (noncommutative_binary_comparison_result_used): Removed.
5605 (rsub_cmp0): New pattern.
5606 (rsub_cmp0_noout): Likewise.
5607 (extzvsi): Changed, keep only meaningful variants.
5608 (SQH, SEZ): New iterators.
5609 (SQH_postfix): New mode attribute.
5610 (SEZ_prefix): New code attribute.
5611 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5612 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5613 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5615 (noncommutative_operator): Check the availability of barrel
5618 2018-04-09 Richard Biener <rguenther@suse.de>
5620 PR tree-optimization/85284
5621 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5622 Only use the niter constraining form of simple_iv when the exit
5625 2018-04-09 Tom de Vries <tom@codesourcery.com>
5628 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5629 (define_expand "*memory_barrier"): New define_expand.
5630 (define_insn "memory_barrier"): New insn.
5632 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5634 PR rtl-optimization/80463
5635 PR rtl-optimization/83972
5636 PR rtl-optimization/83480
5638 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5639 correct producer for the insn.
5640 (tidy_control_flow): Fixup seqnos in case of debug insns.
5642 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5644 PR rtl-optimization/83913
5646 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5647 different sched-times when merging exprs.
5649 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5651 PR rtl-optimization/83962
5653 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5654 tidy_fallthru_edge and tidy_control_flow.
5656 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5658 PR rtl-optimization/83530
5660 * sel-sched.c (force_next_insn): New global variable.
5661 (remove_insn_for_debug): When force_next_insn is true, also leave only
5662 next insn in the ready list.
5663 (sel_sched_region): When the region wasn't scheduled, make another pass
5664 over it with force_next_insn set to 1.
5666 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5668 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5670 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5671 for interrupt control.
5672 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5673 functions for interrupt control.
5674 * config/nds32/nds32-intrinsic.md: Likewise.
5675 * config/nds32/nds32_intrinsic.h: Likewise.
5676 * config/nds32/nds32.h (nds32_builtins): Likewise.
5678 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5680 * config/nds32/nds32.c (nds32_init_machine_status,
5681 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5682 strict_aligned_p field.
5683 (nds32_expand_to_rtl_hook): New function.
5684 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5685 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5687 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5688 Chung-Ju Wu <jasonwucj@gmail.com>
5690 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5691 * config/nds32/nds32-n7.md: New file.
5692 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5693 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5695 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5696 * config/nds32/nds32.md (pipeline_model): Add n7.
5697 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5698 * config/nds32/pipelines.md: Include n7 settings.
5700 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5701 Chung-Ju Wu <jasonwucj@gmail.com>
5703 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5704 * config/nds32/nds32-e8.md: New file.
5705 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5706 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5708 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5709 * config/nds32/nds32.md (pipeline_model): Add e8.
5710 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5711 * config/nds32/pipelines.md: Include e8 settings.
5713 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5714 Chung-Ju Wu <jasonwucj@gmail.com>
5716 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5717 * config/nds32/nds32-n8.md: New file.
5718 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5719 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5721 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5722 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5723 * config/nds32/nds32.md (pipeline_model): Add n8.
5724 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5725 * config/nds32/pipelines.md: Include n8 settings.
5727 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5728 Chung-Ju Wu <jasonwucj@gmail.com>
5730 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5731 * config/nds32/nds32-n9-2r1w.md: New file.
5732 * config/nds32/nds32-n9-3r2w.md: New file.
5733 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5734 nds32_register_ports): New or modify for cpu n9.
5735 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5737 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5738 * config/nds32/nds32-utils.c: New file.
5739 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5740 TARGET_MUL_SLOW): Define.
5741 * config/nds32/nds32.md (pipeline_model): New attribute.
5742 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5743 New options that support cpu n9.
5744 * config/nds32/pipelines.md: Include n9 settings.
5745 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5747 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5749 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5750 information if necessary.
5751 (output_cond_branch_compare_zero): Likewise.
5752 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5753 (nds32_target_alignment): Refine for alignment.
5754 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5755 (FUNCTION_BOUNDARY): Modify.
5756 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5758 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5760 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5762 * config/nds32/constants.md (unspec_volatile_element): Add values for
5763 TLB operation and data prefetch.
5764 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5765 functions for TLB operation and data prefetch.
5766 * config/nds32/nds32-intrinsic.md: Likewise.
5767 * config/nds32/nds32_intrinsic.h: Likewise.
5768 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5769 (nds32_print_operand): Likewise.
5770 * config/nds32/nds32.h (nds32_builtins): Likewise.
5772 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5773 Andrew Pinski <pinsika@gcc.gnu.org>
5776 * match.pd: Use constant_boolean_node of correct type instead of
5777 boolean_true_node or boolean_false_node for simplifying
5778 pointer comparisons to zero.
5780 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5782 PR tree-optimization/80021
5783 * tree.c (verify_type_variant): Make error call in verify_variant_match
5784 translatable and remove final full stop.
5786 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5788 * config/nds32/constants.md (unspec_volatile_element): Add
5789 UNSPEC_VOLATILE_EH_RETURN.
5790 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5791 nds32_output_stack_pop): Support dwarf exception handling process.
5792 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5793 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5794 exception handling process.
5795 (nds32_compute_stack_frame): Likewise.
5796 (nds32_return_addr_rtx): Likewise.
5797 (nds32_initial_elimination_offset): Likewise.
5798 (nds32_expand_prologue): Likewise.
5799 (nds32_expand_epilogue): Likewise.
5800 (nds32_dynamic_chain_address): New function.
5801 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5803 (DYNAMIC_CHAIN_ADDRESS): Define.
5804 (EH_RETURN_DATA_REGNO): Define.
5805 (EH_RETURN_STACKADJ_RTX): Define.
5806 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5807 patterns for dwarf exception handling.
5809 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5811 * config/nds32/nds32.h: Clean up obsolete macros.
5813 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5815 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5816 Add enum values for particular instructions.
5817 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5818 particular intrinsic functions.
5819 * config/nds32/nds32-intrinsic.md: Likewise.
5820 * config/nds32/nds32_intrinsic.h: Likewise.
5821 * config/nds32/nds32.h (nds32_builtins): Likewise.
5822 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5823 (btst, ave): New patterns for particular instructions.
5825 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5827 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5828 Add enum values for atomic load/store and memory sync.
5829 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5831 * config/nds32/nds32-intrinsic.md: Likewise.
5832 * config/nds32/nds32_intrinsic.h: Likewise.
5833 * config/nds32/nds32.h (nds32_builtins): Likewise.
5835 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5837 PR tree-optimization/85257
5838 * fold-const.c (native_encode_vector): If not all elts could fit
5839 and off is -1, return 0 rather than offset.
5840 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5841 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5842 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5843 adjust buffer in native_interpret_expr call.
5845 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5847 * config/nds32/constants.md (unspec_volatile_element): Add cache
5848 control enum values.
5849 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5850 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5851 * config/nds32/nds32.c (nds32_cctl_names): New.
5852 (nds32_print_operand): Handle cache control register names.
5853 * config/nds32/nds32.h (nds32_builtins): New enum values.
5854 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5856 * config/nds32/nds32.md (type): Add mmu.
5857 * config/nds32/pipelines.md (simple_insn): Add mmu.
5859 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5861 * config/nds32/nds32.md (type): Remove call.
5862 * config/nds32/pipelines.md (simple_insn): Likewise.
5864 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5866 * config/nds32/constants.md (unspec_volatile_element): Add
5867 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5868 UNSPEC_VOLATILE_FMFCFG.
5869 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5870 description for fmfcfg and fmfcsr.
5871 (bdesc_1arg): Add fmtcsr.
5872 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5873 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5874 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5875 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5876 unspec_fmfcfg): New patterns.
5877 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5878 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5879 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5880 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5881 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5882 __nds32__fmfcfg): Define.
5884 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5886 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5887 intrinsic register names.
5888 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5889 intrinsic register enum values and macros.
5891 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5893 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5894 for load/store addressing form.
5895 (nds32_print_operand_address): Likewise.
5897 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5900 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5901 based on LABEL_REF. Remove useless assertion.
5902 (pic_address_needs_scratch): Fix formatting.
5903 (sparc_legitimize_pic_address): Minor tweaks.
5904 (sparc_delegitimize_address): Adjust assertion accordingly.
5905 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5906 into symbolic_operand.
5907 (movsi_high_pic_label_ref): Likewise.
5908 (movsi_lo_sum_pic_label_ref): Likewise.
5909 (movdi_pic_label_ref): Likewise.
5910 (movdi_high_pic_label_ref): Likewise.
5911 (movdi_lo_sum_pic_label_ref): Likewise.
5913 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5915 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5916 custom LIB_SPEC setup.
5918 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5919 Kito Cheng <kito.cheng@gmail.com>
5921 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5922 * config/riscv/freebsd.h: New.
5924 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5926 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5927 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5930 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5931 Kito Cheng <kito.cheng@gmail.com>
5933 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5934 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5935 * config/nds32/nds32-protos.h (nds32_output_call,
5936 nds32_output_return): Declare.
5937 * config/nds32/nds32.md: Refine all the call and return patterns.
5939 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5942 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5943 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5945 PR rtl-optimization/84872
5946 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5947 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5950 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5952 * expr.c (copy_blkmode_to_reg): Revert 254862.
5953 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5955 2018-04-06 Richard Biener <rguenther@suse.de>
5958 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5959 after seeing a component reference with an adjacent field. Treat
5960 refs to arrays at struct end of external decls similar to
5961 refs to unconstrained commons.
5963 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5966 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5967 look through SAVE_EXPRs with non-side-effects argument. Adjust
5969 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5972 2018-04-06 Richard Biener <rguenther@suse.de>
5975 * alias.c (find_base_term): New wrapper around find_base_term
5976 unwinding CSELIB_VAL_PTR changes.
5977 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5980 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5982 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5984 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5985 constant definitions.
5986 ("nop"): lr 0,0 -> nopr r0
5987 ("nop_lr0", "nop_lr1"): New insn definitions.
5989 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5991 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5992 NDS32_V3PUSH_AVAILABLE_P macro.
5994 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5995 Chung-Ju Wu <jasonwucj@gmail.com>
5997 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5998 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5999 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6000 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6001 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6002 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6003 * config/nds32/constraints.md: New constraints and checking for hard
6004 float configuration.
6005 * config/nds32/iterators.md: New mode iterator and attribute for hard
6006 float configuration.
6007 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6009 * config/nds32/nds32-fpu.md: New file.
6010 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6011 deal with hard float code generation.
6012 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6014 (abi_type, float_reg_number): New enum type.
6015 * config/nds32/nds32-predicates.c: New predicates for hard float.
6016 * config/nds32/nds32-protos.h: Declare functions for hard float.
6017 * config/nds32/nds32.c: Implementation for hard float configuration.
6018 * config/nds32/nds32.h: Definitions for hard float configuration.
6019 * config/nds32/nds32.md: Include hard float machine description and
6020 modify patterns for hard float configuration.
6021 * config/nds32/nds32.opt: New options for hard float configuration.
6022 * config/nds32/predicates.md: New predicates for hard float
6025 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6027 * common/config/nds32/nds32-common.c
6028 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6030 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6033 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6034 CONSTRUCTOR_ELT (ctor, ...)->value.
6036 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6039 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6041 2018-04-05 Tom de Vries <tom@codesourcery.com>
6044 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6047 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6048 Kito Cheng <kito.cheng@gmail.com>
6050 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6051 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6052 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6053 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6055 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6056 Kito Cheng <kito.cheng@gmail.com>
6058 * config/nds32/constraints.md (Ufe): New memory constraint.
6059 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6060 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6061 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6063 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6064 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6066 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6068 * config/nds32/nds32.md: Use optimize_size in the condition for
6069 alu-shift instructions.
6071 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6073 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6075 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6077 * config/nds32/nds32.md (negsi2): Refine pattern.
6079 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6080 Chung-Ju Wu <jasonwucj@gmail.com>
6082 * config/nds32/iterators.md (shift_rotate): New code iterator.
6083 (shift): New code attribute.
6084 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6085 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6086 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6087 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6088 bit-wise operations.
6089 (andsi3, *andsi3): Ditto.
6090 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6091 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6092 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6093 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6094 nds32_ior_operand, nds32_xor_operand): New predicates.
6096 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6099 (addsi3, subsi3): ... this.
6101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6103 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6105 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6107 * config/nds32/nds32.md: Adjust indention.
6109 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6111 * config/nds32/nds32.md (feature): New attribute.
6113 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6115 * config/nds32/nds32.md (subtype): New attribute.
6117 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6120 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6121 expansion to perform a bitwise AND of the argument followed by a
6122 boolean negation of the result.
6124 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6126 PR rtl-optimization/84878
6127 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6128 the basic block. Assert the use reference is not artificial and that
6129 it has an associated insn.
6131 2018-04-04 Michael Matz <matz@suse.de>
6133 * builtins.c (compute_objsize): Pass correct operand
6134 to array_at_struct_end_p.
6136 2018-04-04 Richard Biener <rguenther@suse.de>
6139 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6140 from contexts for DINFO_LEVEL_TERSE and below.
6142 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6144 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6145 resiter_operand condition.
6146 * config/nds32/nds32.md (*move<mode>): Ditto.
6148 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6149 Monk Chiang <sh.chiang04@gmail.com>
6151 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6153 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6155 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6157 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6158 Kito Cheng <kito.cheng@gmail.com>
6160 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6161 nds32_cond_code_str, output_cond_branch,
6162 output_cond_branch_compare_zero, nds32_expand_cbranch,
6163 nds32_expand_cstore, nds32_expand_movcc,
6164 nds32_output_cbranchsi4_equality_zero,
6165 nds32_output_cbranchsi4_equality_reg,
6166 nds32_output_cbranchsi4_equality_reg_or_const_int,
6167 nds32_output_cbranchsi4_greater_less_zero: New functions.
6168 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6169 nds32_expand_cstore, nds32_expand_movcc,
6170 nds32_output_cbranchsi4_equality_zero,
6171 nds32_output_cbranchsi4_equality_reg,
6172 nds32_output_cbranchsi4_equality_reg_or_const_int,
6173 nds32_output_cbranchsi4_greater_less_zero): Declare.
6174 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6175 nds32_rimm11s_operand): New predicates.
6176 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6177 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6180 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6182 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6183 * config/nds32/nds32.md: Ditto.
6184 * config/nds32/pipelines.md: Ditto.
6186 2018-04-04 Richard Biener <rguenther@suse.de>
6188 PR tree-optimization/85168
6189 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6190 propagating abnormals.
6192 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6194 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6196 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6197 Kito Cheng <kito.cheng@gmail.com>
6199 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6200 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6201 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6202 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6203 * config/nds32/nds32.md (sibcall_internal): New.
6204 (sibcall_register): Remove.
6205 (sibcall_immediate): Remove.
6206 (sibcall_value_internal): New.
6207 (sibcall_value_register): Remove.
6208 (sibcall_value_immediate): Remove.
6209 * config/nds32/predicates.md (nds32_general_register_operand): New.
6210 (nds32_call_address_operand): New.
6212 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6214 PR rtl-optimization/85167
6215 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6216 bb_defs if *split_p, instead preinitialize it to NULL.
6218 PR tree-optimization/85156
6219 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6220 evaluating the argument multiple times.
6222 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6224 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6226 (_mm_cvtpd_ps): Likewise.
6227 (_mm_cvttpd_epi32): Likewise.
6228 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6229 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6230 vector, pixel, and bool following altivec.h include.
6232 2018-04-03 Martin Sebor <msebor@redhat.com>
6234 * doc/extend.texi (Common Function Attributes): Clarify.
6235 (const attribute): Likewise.
6236 (pure attribute): Likewise.
6238 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6241 * config/i386/i386.c (ix86_expand_vector_set): Use
6242 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6244 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6246 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6247 instructions when changing rounding bits to preserve precision bits
6248 in the x87 control word.
6250 2018-04-03 Martin Liska <mliska@suse.cz>
6252 PR tree-optimization/82491
6253 * rtl.h (strip_offset_and_add): Replace += suboffset with
6254 poly_uint64 () + suboffset.
6256 2018-03-29 Martin Liska <mliska@suse.cz>
6257 Martin Jambor <mjambor@suse.cz>
6260 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6261 param_type is not an integral or pointer type.
6263 2018-04-03 Richard Biener <rguenther@suse.de>
6265 * sese.h (recompute_all_dominators): Remove.
6267 2018-04-02 Martin Sebor <msebor@redhat.com>
6269 * doc/invoke.texi (-Wrestrict): Fix typos.
6271 2018-04-02 Jim Wilson <jimw@sifive.com>
6273 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6274 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6275 (<optab>di3, <optab>si3_extend): Likewise.
6276 (<optab>si3_mask, <optab>si3_mask_1): New.
6277 (<optab>di3_mask, <optab>di3_mask_1): New.
6278 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6279 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6280 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6282 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6284 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6287 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6289 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6290 (nds32_canonicalize_comparison): New function.
6292 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6293 Kito Cheng <kito.cheng@gmail.com>
6294 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6296 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6297 * config/nds32/constants.md (unspec_volatile_element): Add
6298 UNSPEC_VOLATILE_RELAX_GROUP.
6299 * config/nds32/nds32-relax-opt.c: New file.
6300 * config/nds32/nds32-predicates.c
6301 (nds32_symbol_load_store_p): New function.
6302 * config/nds32/nds32-protos.h
6303 (nds32_symbol_load_store_p): Declare function.
6304 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6305 * config/nds32/nds32.c
6306 (nds32_register_pass): New function to register pass.
6307 (nds32_register_passes): New function to register passes.
6308 * config/nds32/nds32.md (relax_group): New pattern.
6309 * config/nds32/nds32.opt (mrelax-hint): New option.
6310 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6312 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6314 * config/nds32/t-nds32: Modify files dependency.
6316 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6318 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6319 (PROFILE_HOOK): Define its implementation.
6321 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6323 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6324 type and 32-bit size.
6326 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6329 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6330 (V_128_256): New mode iterator.
6331 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6332 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6333 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6335 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6336 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6338 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6341 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6342 NaN inputs correctly.
6344 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6347 * config/rs6000/vsx.md (??r): New mode attribute.
6348 (*vsx_mov<mode>_64bit): Use it.
6349 (*vsx_mov<mode>_32bit): Likewise.
6351 2018-03-30 Martin Sebor <msebor@redhat.com>
6353 PR tree-optimization/84818
6354 * builtins.c (check_access): Use warning_n.
6356 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6359 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6361 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6364 2018-03-30 Julia Koval <julia.koval@intel.com>
6367 * x86-tune.def (movx, partial_reg_dependency): Enable for
6370 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6373 * lra-constraints.c (process_alt_operands): Move setting
6374 this_alternative_matches below.
6376 2018-03-29 Martin Liska <mliska@suse.cz>
6379 * doc/invoke.texi: Document how LTO works with debug info.
6380 Describe auto-load support of binutils. Mention 'x86-64'
6381 as valid option value of -march option.
6383 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6385 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6388 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6389 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6390 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6393 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6396 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6397 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6398 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6399 for RS6000_BTM_POWERPC64.
6400 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6401 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6402 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6407 2018-03-28 Carl Love <cel@us.ibm.com>
6410 2017-09-27 Carl Love <cel@us.ibm.com>
6412 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6413 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6414 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6417 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6419 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6420 instead of __vector bool.
6421 (_mm_max_pu8): Likewise.
6422 (_mm_min_pi16): Likewise.
6424 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6427 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6428 (DIVWEUO): Likewise.
6430 (DIVDEUO): Likewise.
6431 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6432 DIVWEUO and DIVDEUO.
6433 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6434 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6435 (div_extend): Likewise.
6436 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6438 (__builtin_divweuo): Likewise.
6439 (__builtin_divdeo): Likewise.
6440 (__builtin_divdeuo): Likewise.
6442 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6445 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6446 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6448 PR tree-optimization/82004
6449 * gimple-match-head.c (optimize_pow_to_exp): New function.
6450 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6451 Don't fold to exp if optimize_pow_to_exp is false.
6453 2018-03-28 Martin Liska <mliska@suse.cz>
6456 * calls.c (initialize_argument_information): Fix trailing space.
6457 * common.opt: Fix typo and provide better explanation for
6458 -fsanitize-coverage option.
6459 * config/i386/i386.opt: Fix typo.
6461 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6462 Martin Liska <mliska@suse.cz>
6465 * gimplify.c (asan_poison_variable): Don't do the check for
6466 gimplify_omp_ctxp here.
6467 (gimplify_decl_expr): Do it here.
6468 (gimplify_target_expr): Likewise.
6470 2018-03-28 Martin Liska <mliska@suse.cz>
6473 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6474 chkp_type_bounds_count if MPX is not enabled.
6476 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6478 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6480 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6483 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6484 function to create the function decl for complex long double
6485 multiply and divide for -mabi=ieeelongdouble.
6486 (init_float128_ieee): Call it.
6488 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6491 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6492 -fcf-protection=branch -mibt.
6493 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6495 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6498 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6499 arm_word_relocations.
6501 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6504 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6505 extern array declarations.
6507 2018-03-27 Richard Biener <rguenther@suse.de>
6510 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6511 explicit single_use checks.
6513 2018-03-27 Richard Biener <rguenther@suse.de>
6515 PR tree-optimization/85082
6516 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6519 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6521 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6522 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6523 Turn on fasynchronous-unwind-tables and funwind-tables.
6525 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6528 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6529 (*bmi_blsr_<mode>_ccz): Ditto.
6531 2018-03-26 Tom de Vries <tom@codesourcery.com>
6533 PR tree-optimization/85063
6534 * omp-general.c (offloading_function_p): New function. Factor out
6536 * omp-offload.c (pass_omp_target_link::gate): ... here.
6537 * omp-general.h (offloading_function_p): Declare.
6538 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6539 with attribute omp declare target for offloading functions.
6541 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6543 PR tree-optimization/84005
6544 * tree-data-ref.h (get_base_for_alignment): Declare.
6545 * tree-data-ref.c (get_base_for_alignment_1): New function.
6546 (get_base_for_alignment): Likewise.
6547 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6548 get_base_for_alignment to find a suitable base object, instead
6549 of always using drb->base_address.
6551 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6554 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6555 known size by default.
6557 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6560 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6561 and non BLKmode operands.
6563 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6566 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6567 Clean up attributes.
6569 2018-03-23 Richard Biener <rguenther@suse.de>
6572 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6573 we are going to emit early debug for LTO.
6575 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6578 * function.c (match_asm_constraints_1): Don't optimize if input
6579 doesn't satisfy general_operand predicate for output's mode.
6582 * alias.c (write_dependence_p): Don't require for x_canonicalized
6583 non-VOIDmode if x has VOIDmode.
6586 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6587 just don't try to optimize it rather than assert it never happens.
6589 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6591 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6592 macro expansions for definition of ST_INTERNAL_<mode> and
6593 LD_INTERNAL_<mode> builtins.
6594 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6596 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6598 (altivec_expand_st_builtin): Likewise.
6599 (altivec_expand_builtin): Remove calls to deleted functions.
6600 (rs6000_address_for_altivec): Delete this function.
6601 * config/rs6000/vector.md: Remove expands for
6602 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6604 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6607 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6608 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6609 re-computing once computed.
6610 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6611 (arm_init_machine_status): Initialize
6612 machine->static_chain_stack_bytes.
6614 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6617 * doc/extend.texi: Add four new prototypes for vec_ld.
6618 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6619 definitions for more logical presentation.
6620 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6621 entries for V1TI variants of __builtin_altivec_ld builtin.
6622 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6623 handling of V1TI variant of LVX icode pattern.
6624 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6625 (rs6000_gimple_fold_builtin): Likewise.
6626 (altivec_init_builtins): Add code to define
6627 __builtin_altivec_lvx_v1ti function.
6629 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6632 * function.c (match_asm_constraints_1): Don't do the optimization
6633 if input isn't a REG, SUBREG, MEM or constant.
6635 2018-03-22 Tom de Vries <tom@codesourcery.com>
6637 PR tree-optimization/84956
6638 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6639 bb_has_abnormal_pred.
6641 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6644 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6645 DECL_INITIAL (decl) to decl at the end.
6646 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6649 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6651 * doc/extend.texi (__builtin_tgmath): Document when complex
6652 integer types are treated as _Complex _Float64.
6654 2018-03-21 Tom de Vries <tom@codesourcery.com>
6656 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6658 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6660 PR tree-optimization/84960
6661 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6662 if it is ENTRY block, move them into single succ of ENTRY in that case.
6664 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6666 PR tree-optimization/84811
6667 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6668 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6669 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6670 (known_subrange_p): Update accordingly. Cast each value involved
6671 in the size comparison, rather than casting the result of the
6674 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6676 PR tree-optimization/84982
6677 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6678 by flipping the least significant bit rather than all bits from
6679 bitpos to bitpos + bitsize - 1.
6681 2018-03-21 Nathan Sidwell <nathan@acm.org>
6683 * doc/extend.texi (Deprecated Features): Remove mention of
6684 long-deleted deprecations.
6686 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6689 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6690 * configure: Regenerate.
6692 2018-03-21 Tom de Vries <tom@codesourcery.com>
6694 PR tree-optimization/83126
6695 * tree-parloops.c (num_phis): New function.
6696 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6698 2018-03-21 Nathan Sidwell <nathan@acm.org>
6700 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6701 mention anon-struct/union members and trailing attributes.
6703 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6705 PR tree-optimization/84969
6706 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6707 builtin memset partitions if they set different rhs values.
6709 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6711 PR rtl-optimization/84989
6712 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6713 VEC_DUPLICATE with scalar result mode.
6715 2018-03-21 Martin Liska <mliska@suse.cz>
6718 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6719 not intended return statement.
6721 2018-03-21 Martin Liska <mliska@suse.cz>
6724 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6725 (chkp_find_bound_slots_1): Limit number of iterations.
6727 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6730 * Minor grammar fixes for x86 options.
6732 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6735 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6736 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6738 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6741 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6742 (altivec_lvx_<mode>_1op): Likewise.
6743 (altivec_stvx_<mode>_2op): Likewise.
6744 (altivec_stvx_<mode>_1op): Likewise.
6745 (altivec_lvx_<VM2:mode>): New define_expand.
6746 (altivec_stvx_<VM2:mode>): Likewise.
6747 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6748 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6749 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6750 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6751 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6752 (rs6000_gen_lvx): Likewise.
6753 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6754 (altivec_expand_stv_builtin): Likewise.
6755 (altivec_expand_builtin): Likewise.
6756 * config/rs6000/vector.md: Likewise.
6758 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6761 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6764 2018-03-20 Richard Biener <rguenther@suse.de>
6767 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6768 sign-conversions as zero, fall back to standard scalar_stmt
6771 2018-03-20 Martin Liska <mliska@suse.cz>
6774 * predict.c (rebuild_frequencies): Handle case when we have
6775 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6777 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6780 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6781 flag_section_anchors.
6782 * varasm.c (use_blocks_for_decl_p): Remove hack for
6783 dw2_force_const_mem.
6786 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6788 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6789 be created, use lowpart_subreg of operands[0] rather than operands[0]
6791 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6792 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6793 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6794 and n constraint instead of aarch64_shift_imm_di and Usd.
6795 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6796 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6798 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6801 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6802 to favor GPR over NEON registers.
6803 (<shift>di3_neon): Likewise.
6805 2018-03-20 Tom de Vries <tom@codesourcery.com>
6808 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6809 (nvptx_process_pars): Emit bar.sync asap and alap.
6811 2018-03-20 Tom de Vries <tom@codesourcery.com>
6814 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6815 seen_label if seen_label is already set.
6817 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6820 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6821 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6822 Use 1U instead of 1. Formatting fixes.
6825 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6826 instead of TREE_TYPE (s1) for the return value.
6828 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6830 PR tree-optimization/84946
6831 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6832 bitsize + bitsize in poly_uint64 rather than poly_int64.
6835 * dwarf2asm.c: Include fold-const.c.
6836 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6837 of decl rather than decl itself.
6839 PR rtl-optimization/84643
6840 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6842 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6845 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6846 calling assemble_variable.
6848 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6851 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6852 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6854 2018-03-19 Jim Wilson <jimw@sifive.com>
6857 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6858 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6859 (riscv_first_stack_step): Likewise.
6860 (riscv_option_override): Use STACK_BOUNDARY instead of
6862 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6864 (BIGGEST_ALIGNMENT): Set to 128.
6865 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6866 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6869 2018-03-19 Richard Biener <rguenther@suse.de>
6871 PR tree-optimization/84933
6872 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6873 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6875 2018-03-19 Richard Biener <rguenther@suse.de>
6877 PR tree-optimization/84859
6878 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6879 (cond_if_else_store_replacement): Perform sinking operation on
6880 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6881 Generalize what a BB with a single eligible store is.
6883 2018-03-19 Richard Biener <rguenther@suse.de>
6885 PR tree-optimization/84929
6886 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6887 chrec_is_positive against non-chrec arg.
6889 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6892 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6894 2018-03-18 Martin Liska <mliska@suse.cz>
6896 PR rtl-optimization/84635
6897 * regrename.c (build_def_use): Use matches_mode only when
6900 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6902 PR tree-optimization/84913
6903 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6904 vectorize chains of COND_EXPRs.
6906 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6908 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6910 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6912 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6914 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6916 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6918 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6919 Kito Cheng <kito.cheng@gmail.com>
6921 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6922 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6923 (nds32_adjust_reg_alloc_order): New function.
6924 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6926 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6928 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6929 nds32_print_operand, nds32_print_operand_address): Use
6930 HOST_WIDE_INT_PRINT_DEC instead.
6932 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6934 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6936 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6939 * config/i386/i386.c (initial_ix86_tune_features,
6940 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6942 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6943 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6944 rather than 1u << ix86_tune. Formatting fix.
6945 (ix86_option_override_internal): Change ix86_arch_mask from
6946 unsigned int to unsigned HOST_WIDE_INT, initialize to
6947 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6948 (ix86_function_specific_restore): Likewise.
6950 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6953 * postreload.c (reload_combine_recognize_pattern): Perform
6954 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6955 truncate_int_for_mode the result for the destination's mode.
6958 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6959 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6961 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6964 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6967 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6968 regno_allocno_class_array and sorted_pseudos.
6969 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6970 insns where regno is used.
6972 2018-03-16 Martin Liska <mliska@suse.cz>
6975 * multiple_target.c (create_dispatcher_calls): Redirect
6976 reference in the symbol table.
6978 2018-03-16 Martin Liska <mliska@suse.cz>
6981 * multiple_target.c (create_dispatcher_calls): Redirect also
6984 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6988 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6989 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6990 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6992 2018-03-16 Julia Koval <julia.koval@intel.com>
6994 * doc/invoke.texi (Skylake Server): Add CLWB.
6995 Cannonlake): Remove CLWB.
6997 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6999 PR tree-optimization/84841
7000 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7002 (FLOAT_ONE_CONST_TYPE): Define.
7003 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7004 (sort_by_operand_rank): Put entries with higher constant_type last
7005 rather than first to match comments.
7007 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7009 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7012 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7015 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7016 check and use address of target always.
7018 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7021 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7022 (indirect_thunk_bnd_needed): Likewise.
7023 (indirect_thunks_used): Likewise.
7024 (indirect_thunks_bnd_used): Likewise.
7025 (indirect_return_needed): New.
7026 (indirect_return_bnd_needed): Likewise.
7027 (output_indirect_thunk_function): Add a bool argument for
7029 (output_indirect_thunk_function): Don't generate alias for
7030 function return thunk.
7031 (ix86_code_end): Call output_indirect_thunk_function to generate
7032 function return thunks.
7033 (ix86_output_function_return): Set indirect_return_bnd_needed
7034 and indirect_return_needed instead of indirect_thunk_bnd_needed
7035 and indirect_thunk_needed.
7037 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7039 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7040 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7041 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7043 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7044 Paul Hua <paul.hua.gm@gmail.com>
7047 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7049 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7051 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7052 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7055 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7058 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7059 instead of GET_MODE_SIZE when comparing Units.
7061 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7064 * varasm.c (hash_section): Return an unchangeble hash value
7065 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7066 Return !aarch64_can_use_per_function_literal_pools_p ().
7068 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7071 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7072 rather than address of cmode as last argument to prepare_cmp_insn.
7074 2018-03-15 Julia Koval <julia.koval@intel.com>
7076 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7077 F_AVX512VNNI, F_AVX512BITALG): New.
7079 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7082 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7083 insn for floating-point loads and stores.
7085 2018-03-14 Carl Love <cel@us.ibm.com>
7087 * config/rs6000/rs6000-c.c: Add macro definitions for
7088 ALTIVEC_BUILTIN_VEC_PERMXOR.
7089 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7090 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7091 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7092 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7094 * config/doc/extend.texi: Add prototypes for vec_permxor.
7096 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7099 * diagnostic-show-locus.c (class layout_point): Convert m_line
7100 from int to linenum_type.
7101 (line_span::comparator): Use linenum "compare" function when
7102 comparing line numbers.
7103 (test_line_span): New function.
7104 (layout_range::contains_point): Convert param "row" from int to
7106 (layout_range::intersects_line_p): Likewise.
7107 (layout::will_show_line_p): Likewise.
7108 (layout::print_source_line): Likewise.
7109 (layout::should_print_annotation_line_p): Likewise.
7110 (layout::print_annotation_line): Likewise.
7111 (layout::print_leading_fixits): Likewise.
7112 (layout::annotation_line_showed_range_p): Likewise.
7113 (struct line_corrections): Likewise for field m_row.
7114 (line_corrections::line_corrections): Likewise for param "row".
7115 (layout::print_trailing_fixits): Likewise.
7116 (layout::get_state_at_point): Likewise.
7117 (layout::get_x_bound_for_row): Likewise.
7118 (layout::print_line): Likewise.
7119 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7120 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7121 * input.c (selftest::test_linenum_comparisons): New function.
7122 (selftest::input_c_tests): Call it.
7123 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7124 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7125 * selftest.h (ASSERT_GT): New macro.
7126 (ASSERT_GT_AT): New macro.
7127 (ASSERT_LT): New macro.
7128 (ASSERT_LT_AT): New macro.
7130 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7132 PR rtl-optimization/84780
7133 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7135 2018-03-14 Martin Liska <mliska@suse.cz>
7137 * tree.c (record_node_allocation_statistics): Use
7138 get_stats_node_kind.
7139 (get_stats_node_kind): New function extracted from
7140 record_node_allocation_statistics.
7141 (free_node): Use get_stats_node_kind.
7143 2018-03-14 Richard Biener <rguenther@suse.de>
7145 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7146 that the value-set of ANTIC_IN doesn't grow.
7149 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7151 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7152 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7154 2018-03-14 Julia Koval <julia.koval@intel.com>
7156 * config.gcc (icelake-client, icelake-server): New.
7158 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7159 (initial_ix86_arch_features): Ditto.
7160 (PTA_SKYLAKE): Add SGX.
7161 (PTA_ICELAKE): Remove.
7162 (PTA_ICELAKE_CLIENT): New.
7163 (PTA_ICELAKE_SERVER): New.
7164 (ix86_option_override_internal): Split up icelake on icelake client and
7166 (get_builtin_code_for_version): Ditto.
7167 (fold_builtin_cpu): Ditto.
7168 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7169 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7170 * config/i386/i386.h (processor_type): Ditto.
7171 * doc/invoke.texi: Ditto.
7173 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7176 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7177 INTEGER_CST offset, add it together with bitpos / 8 and
7178 sign extend based on POINTER_SIZE.
7182 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7185 * config/i386/constraints.md (Yc): New register constraint.
7186 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7187 Use Yc constraint for alternative 2 of operand 0. Remove
7188 preferred_for_speed attribute.
7190 2018-03-14 Richard Biener <rguenther@suse.de>
7192 PR tree-optimization/84830
7193 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7194 with the old one to avoid oscillations.
7196 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7199 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7201 (assign_by_spills): Return a flag of reload assignment failure.
7202 Do not process the reload assignment failures. Do not spill other
7203 reload pseudos if they has the same reg class. Update n if
7205 (lra_assign): Add a return arg. Set up from the result of
7206 assign_by_spills call.
7207 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7208 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7209 usage_insns if it is not NULL.
7210 (spill_hard_reg_in_range): New function.
7211 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7212 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7213 function prototypes.
7214 (lra_assign): Change prototype.
7215 * lra.c (lra): Add code to deal with fails by splitting hard reg
7218 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7220 * config/riscv/riscv.opt (mrelax): New option.
7221 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7222 "norelax" when riscv_mrelax is disabled.
7223 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7225 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7228 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7229 reassociation for int modes.
7231 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7233 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7234 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7236 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7237 * config/aarch64/aarch64-sve.md
7238 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7239 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7240 (*extend<mode><Vwide>2): Rename to...
7241 (aarch64_sve_extend<mode><Vwide>2): ...this.
7242 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7243 renaming the old pattern to...
7244 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7246 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7247 define_expand, renaming the old pattern to...
7248 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7249 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7250 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7251 account when deciding which SVE instruction the optab should use.
7252 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7254 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7256 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7257 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7258 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7259 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7260 (tlsdesc_small_<mode>): Turn a define_expand and use
7261 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7262 (tlsdesc_small_advsimd_<mode>): ...this.
7263 (tlsdesc_small_sve_<mode>): New pattern.
7265 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7267 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7268 (UNSPEC_UMUL_HIGHPART): New constants.
7269 (MUL_HIGHPART): New int iteraor.
7270 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7271 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7273 (*<su>mul<mode>3_highpart): New define_insn.
7275 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7278 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7281 2018-03-13 Martin Liska <mliska@suse.cz>
7284 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7286 (sem_item_optimizer::~sem_item_optimizer): Release it.
7287 (sem_item_optimizer::merge_classes): Register variable aliases.
7288 (sem_item_optimizer::fixup_pt_set): New function.
7289 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7290 * ipa-icf.h: Declare new variables and functions.
7292 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7295 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7296 integer_pow2p@2 and test integer_pow2p in condition.
7297 (A < 0 ? C : 0): Similarly for @1.
7300 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7301 characters starting at p contain '\0' character, don't look beyond
7305 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7306 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7309 * reg-stack.c (change_stack): Change update_end var from int to
7310 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7311 also call set_block_for_insn on the newly added insns and rescan.
7314 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7315 on the last operand.
7318 * tree.c (stabilize_reference_1): Return save_expr (e) for
7319 STATEMENT_LIST even if it doesn't have side-effects.
7321 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7323 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7325 2018-03-12 Renlin Li <renlin.li@arm.com>
7327 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7328 aarch64_output_scalar_simd_mov_immediate.
7330 2018-03-12 Martin Sebor <msebor@redhat.com>
7332 PR tree-optimization/83456
7333 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7334 for perfectly overlapping calls to memcpy.
7335 (gimple_fold_builtin_memory_chk): Same.
7336 (gimple_fold_builtin_strcpy): Handle no-warning.
7337 (gimple_fold_builtin_stxcpy_chk): Same.
7338 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7340 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7342 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7343 parameter. Use it for SFmode.
7344 (rs6000_function_arg_advance_1): Adjust.
7345 (rs6000_function_arg): Adjust.
7346 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7348 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7350 PR rtl-optimization/84169
7351 PR rtl-optimization/84780
7352 * combine.c (can_combine_p): Check for a 2-insn combination whether
7353 the destination register is used between the two insns, too.
7355 2018-03-12 Richard Biener <rguenther@suse.de>
7357 PR tree-optimization/84803
7358 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7359 for refs DR analysis didn't process.
7361 2018-03-12 Richard Biener <rguenther@suse.de>
7363 PR tree-optimization/84777
7364 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7365 force-vectorize loops ignore whether we are optimizing for size.
7367 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7369 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7370 (TARGET_MD_ASM_ADJUST): Define.
7372 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7373 Kito Cheng <kito.cheng@gmail.com>
7374 Chung-Ju Wu <jasonwucj@gmail.com>
7376 * config/nds32/nds32.c (nds32_compute_stack_frame,
7377 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7378 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7379 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7380 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7381 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7382 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7383 * config/nds32/nds32.md (prologue, epilogue): Use macro
7384 NDS32_V3PUSH_AVAILABLE_P to do checking.
7386 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7389 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7390 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7391 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7392 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7393 addition of most attributes on !orig_type_die or the attribute not
7394 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7396 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7397 Chung-Ju Wu <jasonwucj@gmail.com>
7399 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7401 * config/nds32/nds32.opt (mvh): New option.
7403 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7404 Chung-Ju Wu <jasonwucj@gmail.com>
7406 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7408 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7409 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7412 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7413 Chung-Ju Wu <jasonwucj@gmail.com>
7415 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7417 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7418 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7420 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7421 Kito Cheng <kito.cheng@gmail.com>
7422 Chung-Ju Wu <jasonwucj@gmail.com>
7424 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7425 UNSPEC_FFMISM and UNSPEC_FLMISM.
7426 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7427 for ffb, ffmism and flmism.
7428 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7429 (unspec_ffmism): Ditto.
7430 (unspec_flmism): Ditto.
7431 (nds32_expand_builtin_impl): Check if string extension is available.
7432 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7433 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7435 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7438 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7441 * lra-assigns.c (assign_by_spills): Return a flag of reload
7442 assignment failure. Do not process the reload assignment
7443 failures. Do not spill other reload pseudos if they has the same
7445 (lra_assign): Add a return arg. Set up from the result of
7446 assign_by_spills call.
7447 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7448 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7449 usage_insns if it is not NULL.
7450 (spill_hard_reg_in_range): New function.
7451 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7452 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7453 function prototypes.
7454 (lra_assign): Change prototype.
7455 * lra.c (lra): Add code to deal with fails by splitting hard reg
7458 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7461 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7463 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7466 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7467 (dw_val_node): Add val_symbolic_view.
7468 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7469 (symview_upper_bound): New.
7470 (new_line_info_table): Initialize symviews_since_reset.
7471 (dwarf2out_source_line): Count symviews_since_reset and set
7472 symview_upper_bound.
7473 (dw_val_equal_p): Handle symview.
7474 (add_AT_symview): New.
7475 (print_dw_val): Handle symview.
7476 (attr_checksum, attr_checksum_ordered): Likewise.
7477 (same_dw_val_p, size_of_die): Likewise.
7478 (value_format, output_die): Likewise.
7479 (add_high_low_attributes): Use add_AT_symview for entry_view.
7480 (dwarf2out_finish): Reset symview_upper_bound, clear
7483 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7486 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7487 Add strict argument and use it.
7488 (rs6000_split_multireg_move): Update for new strict argument.
7489 (mem_operand_gpr): Disallow all non-offsettable addresses.
7490 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7492 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7495 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7496 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7497 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7500 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7501 decl, use remap_type if we want to use the type.
7503 2018-03-09 Martin Sebor <msebor@redhat.com>
7505 PR tree-optimization/84526
7506 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7508 (builtin_access::generic_overlap): Be prepared to handle non-array
7511 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7513 PR rtl-optimization/84682
7514 * lra-constraints.c (process_address_1): Check is_address flag
7515 for address constraints.
7516 (process_alt_operands): Likewise.
7517 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7518 preprocess_constraints.
7519 * recog.h (preprocess_constraints): Add oploc parameter.
7521 * recog.c (preprocess_constraints): Test address_operand for
7522 CT_ADDRESS constraints.
7524 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7527 * lra-assigns.c (assign_by_spills): Return a flag of reload
7528 assignment failure. Do not process the reload assignment
7529 failures. Do not spill other reload pseudos if they has the same
7531 (lra_assign): Add a return arg. Set up from the result of
7532 assign_by_spills call.
7533 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7534 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7535 usage_insns if it is not NULL.
7536 (spill_hard_reg_in_range): New function.
7537 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7538 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7539 function prototypes.
7540 (lra_assign): Change prototype.
7541 * lra.c (lra): Add code to deal with fails by splitting hard reg
7544 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7547 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7548 Accept complain bool parameter. Only emit errors if it is true.
7549 (arm_parse_cpu_option_name): Likewise.
7550 (arm_target_thumb_only): Adjust callers of the above.
7551 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7552 prototype to take a default true bool parameter.
7553 (arm_parse_arch_option_name): Likewise.
7555 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7556 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7560 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7561 * configure: Regenerate.
7562 * configure.ac ("linker --version-script option"): New.
7563 ("linker soname option"): New.
7565 2018-03-09 Richard Biener <rguenther@suse.de>
7567 PR tree-optimization/84775
7568 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7569 immediate uses of predicate stmts and mark them modified.
7572 PR tree-optimization/84178
7573 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7575 (version_loop_for_if_conversion): Delay update_ssa call.
7576 (tree_if_conversion): Delay update_ssa until after predicate
7579 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7582 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7583 when the function accesses prior frames.
7585 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7588 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7589 gen_llsym, otherwise call maybe_gen_llsym.
7592 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7593 has ',' character inside of it.
7595 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7598 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7599 as clobbering CC_REGNUM.
7601 2018-03-08 Richard Biener <rguenther@suse.de>
7604 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7605 (follow_copies_to_constant): Do not follow SSA names registered
7608 2018-03-08 Richard Biener <rguenther@suse.de>
7610 PR tree-optimization/84178
7611 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7613 (version_loop_for_if_conversion): Delay update_ssa call.
7614 (tree_if_conversion): Delay update_ssa until after predicate
7617 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7619 PR tree-optimization/84178
7620 * tree-if-conv.c (release_bb_predicate): Remove the
7621 the assertion that the stmts have NULL use_ops.
7622 Discard the statements, asserting that they haven't
7623 yet been added to a BB.
7625 2018-03-08 Richard Biener <rguenther@suse.de>
7627 PR tree-optimization/84746
7628 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7629 (phi_translate): Pass in destination ANTIC_OUT set.
7630 (phi_translate_1): Likewise. For a simplified result lookup
7631 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7632 (phi_translate_set): Adjust.
7633 (do_pre_regular_insertion): Likewise.
7634 (do_pre_partial_partial_insertion): Likewise.
7636 2018-03-08 Martin Liska <mliska@suse.cz>
7638 PR gcov-profile/84735
7639 * doc/gcov.texi: Document usage of profile files.
7640 * gcov-io.h: Document changes in the format.
7642 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7646 * dwarf2out.c (struct dw_line_info_table): Update comments for
7648 (FORCE_RESET_NEXT_VIEW): New.
7649 (FORCE_RESETTING_VIEW_P): New.
7650 (RESETTING_VIEW_P): Check for -1 too.
7651 (ZERO_VIEW_P): Likewise.
7652 (new_line_info_table): Force-reset next view.
7653 (dwarf2out_begin_function): Likewise.
7654 (dwarf2out_source_line): Simplify zero_view_p initialization.
7655 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7656 view directly. Omit view when omitting .loc at line 0.
7658 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7660 PR tree-optimization/84740
7661 * tree-switch-conversion.c (process_switch): Call build_constructors
7662 only if info.phi_count is non-zero.
7664 PR tree-optimization/84739
7665 * tree-tailcall.c (find_tail_calls): Check call arguments against
7666 DECL_ARGUMENTS (current_function_decl) rather than
7667 DECL_ARGUMENTS (func) when checking for tail recursion.
7669 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7671 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7672 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7673 Volker Reichelt's entry and add entries for people that perform
7674 GCC fuzzy testing and report numerous bugs.
7676 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7679 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7680 readonly data in sdata, if that is disabled.
7681 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7682 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7683 -mreadonly-in-sdata option.
7685 2018-03-07 Martin Sebor <msebor@redhat.com>
7687 PR tree-optimization/84468
7688 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7689 basic block when looking for nul assignment.
7691 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7694 * except.h (output_function_exception_table): Adjust prototype.
7695 * except.c (output_function_exception_table): Remove FNNAME parameter
7696 and add SECTION parameter. Ouput one part of the table at a time.
7697 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7698 the first part of the exception table and emit unwind directives.
7699 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7700 (i386_pe_seh_cold_init): Likewise.
7701 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7702 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7703 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7704 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7705 * config/i386/winnt.c: Include except.h.
7706 (struct seh_frame_state): Add reg_offset, after_prologue and
7707 in_cold_section fields.
7708 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7709 (i386_pe_seh_cold_init): New function.
7710 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7711 to seh->in_cold_section.
7712 (seh_emit_push): Record the offset of the push.
7713 (seh_emit_save): Record the offet of the save.
7714 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7715 Test seh->after_prologue to disregard the epilogue.
7716 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7717 (i386_pe_end_cold_function): New function.
7719 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7722 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7723 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7726 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7728 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7729 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7732 * multiple_target.c: Include tree-inline.h and intl.h.
7733 (expand_target_clones): Diagnose and fail if node->definition and
7734 !tree_versionable_function_p (node->decl).
7736 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7738 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7740 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7741 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7742 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7744 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7747 * combine.c (try_combine): Use reg_or_subregno instead of handling
7748 just paradoxical SUBREGs and REGs.
7750 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7752 * config/arc/arc.c (arc_finalize_pic): Remove function.
7753 (arc_must_save_register): We use single base PIC register, remove
7754 checks to save/restore the PIC register.
7755 (arc_expand_prologue): Likewise.
7756 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7758 (arc_verify_short): Likewise.
7759 (arc_attr_type): Likewise.
7760 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7761 (walk_stores): Likewise.
7762 (arc_address_cost): Make it static.
7763 (arc_verify_short): Likewise.
7764 (branch_dest): Likewise.
7765 (arc_attr_type): Likewise.
7766 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7767 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7768 (arc_final_prescan_insn): Remove inserting the nops due to
7769 hardware hazards. It is done in reorg step.
7770 (insn_length_variant_t): Remove.
7771 (insn_length_parameters_t): Likewise.
7772 (arc_insn_length_parameters): Likewise.
7773 (arc_get_insn_variants): Likewise.
7774 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7776 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7779 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7782 PR tree-optimization/84687
7783 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7785 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7787 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7789 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7790 Rename to ppc_speculation_barrier.
7791 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7792 __builtin_ppc_speculation_barrier.
7794 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7797 * combine.c (combine_simplify_rtx): Don't try to simplify if
7798 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7801 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7803 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7804 to 32 bytes when compiling for POWER9.
7806 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7809 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7810 regparm >= 3 with no arg reg available also for calls with
7811 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7814 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7816 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7818 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7821 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7823 2018-03-05 Richard Biener <rguenther@suse.de>
7825 PR tree-optimization/84486
7826 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7827 When inserting a __builtin_assume_aligned call set the LHS
7828 SSA name alignment info accordingly.
7830 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7832 PR tree-optimization/84114
7833 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7834 Avoid reassociation of FLOAT_MODE addition.
7836 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7838 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7839 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7840 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7841 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7842 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7843 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7844 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7846 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7847 __builtin_ia32_wbinvd): New builtins.
7848 (SPECIAL_ARGS2): New.
7849 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7850 (SPECIAL_ARGS2): New.
7851 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7852 (ix86_valid_target_attribute_inner_p): Ditto.
7853 (ix86_init_mmx_sse_builtins): Add special_args2.
7854 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7855 TARGET_WBNOINVD_P): New.
7856 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7857 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7858 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7859 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7860 * config/i386/pconfigintrin.h: New file.
7861 * config/i386/wbnoinvdintrin.h: Ditto.
7862 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7864 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7866 2018-03-05 Richard Biener <rguenther@suse.de>
7868 PR tree-optimization/84670
7869 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7871 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7872 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7873 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7874 doesn't grow if all successors (recursively) were visited at least
7877 2018-03-05 Richard Biener <rguenther@suse.de>
7879 PR tree-optimization/84650
7880 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7881 if executed in the loop pipeline.
7883 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7885 * doc/configfiles.texi (Configuration Files): Move info about
7886 conditionalizing $target-protos.h to...
7887 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7888 differs from $target-protos.h.
7890 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7891 Chung-Ju Wu <jasonwucj@gmail.com>
7893 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7894 * config/nds32/nds32-multiple.md (setmemsi): Define.
7895 * config/nds32/nds32-memory-manipulation.c
7896 (nds32_gen_dup_4_byte_to_word_value): New.
7897 (emit_setmem_word_loop): New.
7898 (emit_setmem_byte_loop): New.
7899 (nds32_expand_setmem_loop): New.
7900 (nds32_expand_setmem_loop_v3m): New.
7901 (nds32_expand_setmem_unroll): New.
7902 (nds32_expand_setmem): New.
7904 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7905 Chung-Ju Wu <jasonwucj@gmail.com>
7907 * config/nds32/nds32-memory-manipulation.c
7908 (nds32_emit_load_store): New.
7909 (nds32_emit_post_inc_load_store): New.
7910 (nds32_emit_mem_move): New.
7911 (nds32_emit_mem_move_block): New.
7912 (nds32_expand_movmemsi_loop_unknown_size): New.
7913 (nds32_expand_movmemsi_loop_known_size): New.
7914 (nds32_expand_movmemsi_loop): New.
7915 (nds32_expand_movmemsi_unroll): New.
7916 (nds32_expand_movmemqi): Rename ...
7917 (nds32_expand_movmemsi): ... to this.
7918 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7919 (movmemsi): ... to this.
7920 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7921 (nds32_expand_movmemsi): ... to this.
7923 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7924 Monk Chiang <sh.chiang04@gmail.com>
7925 Chung-Ju Wu <jasonwucj@gmail.com>
7927 * config/nds32/nds32-protos.h
7928 (nds32_expand_load_multiple): New arguments.
7929 (nds32_expand_store_multiple): Ditto.
7930 (nds32_valid_multiple_load_store): Rename ...
7931 (nds32_valid_multiple_load_store_p): ... to this.
7932 * config/nds32/nds32-memory-manipulation.c
7933 (nds32_expand_load_multiple): Refine implementation.
7934 (nds32_expand_store_multiple): Ditto.
7935 * config/nds32/nds32-multiple.md
7936 (load_multiple): Update nds32_expand_load_multiple interface.
7937 (store_multiple): Update nds32_expand_store_multiple interface.
7938 * config/nds32/nds32-predicates.c
7939 (nds32_valid_multiple_load_store): Rename ...
7940 (nds32_valid_multiple_load_store_p): ... to this and refine
7942 * config/nds32/predicates.md
7943 (nds32_load_multiple_and_update_address_operation): New predicate.
7944 (nds32_store_multiple_and_update_address_operation): New predicate.
7946 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7947 Chung-Ju Wu <jasonwucj@gmail.com>
7949 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7950 (combo): New attribute.
7951 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7953 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7955 * config/nds32/nds32.opt: Change -mcmodel= default value.
7957 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7958 Monk Chiang <sh.chiang04@gmail.com>
7959 Chung-Ju Wu <jasonwucj@gmail.com>
7961 * config/nds32/constants.md (unspec_element): New enum.
7962 * config/nds32/constraints.md (Umw): New constraint.
7963 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7964 * config/nds32/nds32-intrinsic.md: Likewise.
7965 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7966 (nds32_valid_smw_lwm_base_p): New.
7967 (nds32_output_smw_single_word): New.
7968 (nds32_output_lmw_single_word): New.
7969 (nds32_expand_unaligned_load): New.
7970 (nds32_expand_unaligned_store): New.
7971 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7972 (nds32_output_smw_single_word): Declare.
7973 (nds32_output_lmw_single_word): Declare.
7974 (nds32_expand_unaligned_load): Declare.
7975 (nds32_expand_unaligned_store): Declare.
7976 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7977 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7978 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7979 NDS32_BUILTIN_UASTORE_DW.
7980 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7983 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7984 Kito Cheng <kito.cheng@gmail.com>
7985 Chung-Ju Wu <jasonwucj@gmail.com>
7987 * config/nds32/nds32-intrinsic.c
7988 (nds32_expand_builtin_null_ftype_reg): Delete.
7989 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7990 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7991 (nds32_read_argument): New.
7992 (nds32_legitimize_target): Ditto.
7993 (nds32_legitimize_argument): Ditto.
7994 (nds32_check_constant_argument): Ditto.
7995 (nds32_expand_unop_builtin): Ditto.
7996 (nds32_expand_unopimm_builtin): Ditto.
7997 (nds32_expand_binop_builtin): Ditto.
7998 (nds32_builtin_decl_impl): Ditto.
7999 (builtin_description): Ditto.
8000 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8001 (nds32_init_builtins_impl): Ditto.
8002 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8003 (nds32_builtin_decl): New.
8004 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8005 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8007 2018-03-02 Jeff Law <law@redhat.com>
8009 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8010 (redundant_insn, fill_simple_delay_slots): Likewise.
8011 (fill_slots_from_thread): Likewise.
8012 * resource.c (mark_referenced_resources): Likewise.
8013 (mark_set_resources, find_dead_or_set_registers): Likewise.
8015 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8017 * substring-locations.h (format_warning_va): Formatting fix for
8019 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8021 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8022 * substring-locations.c: Include intl.h.
8023 (format_warning_va): Turned into small wrapper around
8024 format_warning_n_va, renamed to ...
8025 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8026 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8028 (format_warning_at_substring_n): New function.
8029 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8030 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8031 format_warning_at_substring with just a shorter name instead of
8032 const function pointer.
8033 (fmtwarn_n): New function.
8034 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8035 appropriate, get rid of all the fmtstr temporaries, move conditionals
8036 with G_() wrapped string literals directly into fmtwarn arguments,
8037 cast dir.len to (int), formatting fixes.
8039 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8041 * doc/invoke.texi: Remove "Cilk Plus" references.
8043 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8044 Richard Biener <rguenther@suse.de>
8047 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8048 for error or warning attributes if CALL_FROM_THUNK_P is set.
8051 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8054 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8055 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8058 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8059 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8061 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8062 instead of -1U in last predictors element's probability member.
8064 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8067 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8068 arguments if they are comparable.
8070 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8072 PR tree-optimization/84634
8073 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8074 masks and masked_loop_p with a single loop_masks, making sure it's
8075 null for bb vectorization.
8077 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8079 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8080 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8081 loop->force_vectorize to check whether there is no alias.
8083 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8086 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8088 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8090 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8091 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8092 instead of a loop around prev_real_insn.
8093 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8097 * config/i386/i386.c (ix86_print_operand): Use conditional
8098 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8101 2018-03-02 Richard Biener <rguenther@suse.de>
8103 PR tree-optimization/84427
8104 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8105 (bitmap_set_subtract_values): Rewrite to handle multiple
8108 (prune_clobbered_mems): Likewise.
8109 (phi_translate): Take edge instead of pred/phiblock.
8110 (phi_translate_1): Likewise.
8111 (phi_translate_set): Likewise. Insert all translated
8112 exprs for a value into the set, keeping possibly multiple
8113 expressions per value.
8114 (compute_antic_aux): Adjust for phi_translate changes.
8115 When intersecting union the expressions and prune those
8116 not in the final value set, keeping possibly multiple
8117 expressions per value. Do not use value-insertion
8118 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8119 all expressions. Add verification that the value-sets
8120 only shrink during iteration.
8121 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8122 (do_pre_regular_insertion): Likewise.
8123 (do_pre_partial_partial_insertion): Likewise.
8125 2018-03-02 Richard Biener <rguenther@suse.de>
8128 * config/darwin.c (saved_debug_info_level): New static global.
8129 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8130 (darwin_asm_lto_end): Restore debug info generation settings.
8132 2018-03-01 Martin Liska <mliska@suse.cz>
8135 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8138 2018-03-01 Richard Biener <rguenther@suse.de>
8141 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8142 type annotation with LTO when debug was disabled at compile-time.
8144 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8146 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8148 (mips_final_postscan_insn): Likewise.
8150 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8152 PR rtl-optimization/84528
8153 * alias.c (init_alias_target): Add commentary.
8154 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8155 a unique base value if the frame pointer is not eliminated
8156 to the stack pointer.
8158 2018-03-01 Tom de Vries <tom@codesourcery.com>
8160 PR rtl-optimization/83327
8161 * lra-int.h (hard_regs_spilled_into): Declare.
8162 * lra.c (hard_regs_spilled_into): Define.
8163 (init_reg_info): Init hard_regs_spilled_into.
8164 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8165 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8166 (process_bb_lives): Handle hard_regs_spilled_into.
8167 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8168 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8170 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8172 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8173 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8174 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8175 * config/rs6000/aix72.h: New file.
8177 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8179 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8180 instead of warning_at with conditional singular and plural messages
8184 * stor-layout.c (update_alignment_for_field): For
8185 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8186 && !DECL_PACKED (field), do the alignment update, just use
8187 only desired_align instead of MAX (type_align, desired_align)
8189 (place_field): Don't do known_align < desired_align handling
8190 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8191 is non-NULL, instead do it after rli->prev_field handling and
8192 only if not within a bitfield word. For DECL_PACKED (field)
8193 use type_align of BITS_PER_UNIT.
8195 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8197 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8198 superfluous parentheses and trailing spaces.
8200 2018-02-28 Richard Biener <rguenther@suse.de>
8202 PR tree-optimization/84584
8203 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8204 SCoPs with fake exit edge.
8206 2018-02-28 Martin Liska <mliska@suse.cz>
8209 * timevar.c (timer::print): Fix format to properly print 100%
8212 2018-02-28 Richard Biener <rguenther@suse.de>
8215 * genmatch.c (capture_info::walk_match): Do not mark
8216 captured expressions without operands as expr_p given
8217 they act more like predicates and should be subject to
8218 "lost tail" side-effect preserving.
8220 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8222 PR rtl-optimization/81611
8223 * auto-inc-dec.c (attempt_change): Move dead note from
8224 mem_insn if it's the next use of regno
8225 (find_address): Take address use of reg holding
8226 non-incremented value. Add parm to limit search to the named
8228 (merge_in_block): Attempt to use a mem insn that is the next
8229 use of the original regno.
8231 2018-02-27 Martin Sebor <msebor@redhat.com>
8234 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8235 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8237 2018-02-27 Martin Sebor <msebor@redhat.com>
8239 PR translation/84207
8240 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8241 n argument to unsigned HOST_WIDE_INT.
8242 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8243 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8244 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8245 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8247 2018-02-27 Richard Biener <rguenther@suse.de>
8249 PR tree-optimization/84512
8250 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8251 Do not use the estimate returned from record_stmt_cost for
8252 the scalar iteration cost but sum properly using add_stmt_cost.
8254 2018-02-27 Richard Biener <rguenther@suse.de>
8256 PR tree-optimization/84466
8257 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8258 Adjust last change to less strictly validate use operands.
8260 2018-02-27 Martin Liska <mliska@suse.cz>
8262 PR gcov-profile/84548
8263 * gcov.c (process_file): Allow partial overlap and consider it
8264 also as group functions.
8265 (output_lines): Properly calculate range of lines for a group.
8267 2018-02-27 Martin Liska <mliska@suse.cz>
8269 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8270 'ggc' suffixes. Change first column width.
8271 (timer::print): Fix formatting of the column.
8273 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8275 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8276 preserve inline entry blocks for the sake of debug inline
8277 entry point markers alone.
8278 (remove_unused_locals): Suggest in comments a better place to
8279 force the preservation of inline entry blocks that are
8280 otherwise unused, but do not preserve them.
8282 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8284 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8286 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8289 * config/i386/constraints.md (Bs): Replace
8290 ix86_indirect_branch_register with
8291 TARGET_INDIRECT_BRANCH_REGISTER.
8293 * config/i386/i386.md (indirect_jump): Likewise.
8294 (tablejump): Likewise.
8295 (*sibcall_memory): Likewise.
8296 (*sibcall_value_memory): Likewise.
8297 Peepholes of indirect call and jump via memory: Likewise.
8298 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8299 (*sibcall_value_GOT_32): Likewise.
8300 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8301 (GOT_memory_operand): Likewise.
8302 (call_insn_operand): Likewise.
8303 (sibcall_insn_operand): Likewise.
8304 (GOT32_symbol_operand): Likewise.
8305 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8307 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8309 PR rtl-optimization/83496
8310 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8311 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8312 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8313 redundant insn, if any.
8314 (relax_delay_slots): Likewise.
8315 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8317 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8319 PR tree-optimization/83965
8320 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8321 that grouped statements are part of a reduction chain. Return
8322 true if the statement is not marked as a reduction itself but
8324 (vect_recog_dot_prod_pattern): Don't check whether the statement
8325 is part of a group here.
8326 (vect_recog_sad_pattern): Likewise.
8327 (vect_recog_widen_sum_pattern): Likewise.
8329 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8332 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8334 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8337 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8339 (ix86_output_indirect_function_return): New prototype.
8340 (ix86_split_simple_return_pop_internal): Likewise.
8341 * config/i386/i386.c (indirect_return_via_cx): New.
8342 (indirect_return_via_cx_bnd): Likewise.
8343 (indirect_thunk_name): Handle return va CX_REG.
8344 (output_indirect_thunk_function): Create alias for
8345 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8346 (ix86_output_indirect_jmp): Remove the bool argument.
8347 (ix86_output_indirect_function_return): New function.
8348 (ix86_split_simple_return_pop_internal): Likewise.
8349 * config/i386/i386.md (*indirect_jump): Don't pass false
8350 to ix86_output_indirect_jmp.
8351 (*tablejump_1): Likewise.
8352 (simple_return_pop_internal): Change it to define_insn_and_split.
8353 Call ix86_split_simple_return_pop_internal to split it for
8355 (simple_return_indirect_internal): Call
8356 ix86_output_indirect_function_return instead of
8357 ix86_output_indirect_jmp.
8359 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8362 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8363 memset and value initialization afterwards.
8365 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8367 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8369 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8372 * common/config/aarch64/aarch64-common.c
8373 (aarch_option_optimization_table[]): Switch
8374 off fomit-frame-pointer
8376 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8377 Chung-Ju Wu <jasonwucj@gmail.com>
8379 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8381 (store_multiple): Ditto.
8383 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8385 * config.gcc: Add --with-cpu support for nds32 target.
8386 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8387 * config/nds32/nds32.opt: Add -mcpu= option.
8389 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8391 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8392 isel=yes): Warn for these deprecated options.
8394 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8396 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8397 ISA_2_5_MASKS_EMBEDDED.
8399 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8401 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8402 p->max as pointers rather than using iterative_hash_expr.
8404 2018-02-23 Carl Love <cel@us.ibm.com>
8406 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8407 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8409 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8410 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8411 P8V_BUILTIN_VEC_VUNSIGNED2.
8413 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8416 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8417 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8419 * lra-constraints.c (curr_insn_transform): Set up
8420 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8422 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8423 * lra-eliminations.c (spill_pseudos): Ditto.
8424 (process_insn_for_elimination): Ditto.
8425 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8426 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8427 LRA_NON_CLOBBERED_ALT.
8429 2018-02-22 Martin Sebor <msebor@redhat.com>
8431 PR tree-optimization/84480
8432 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8433 to maybe_diag_stxncpy_trunc. Call it.
8434 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8435 from gimple_fold_builtin_strcpy. Print inlining stack.
8436 (handle_builtin_stxncpy): Print inlining stack.
8437 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8439 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8442 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8443 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8444 and -fcheck-pointer-bounds are used together.
8445 (indirect_thunk_prefix): New enum.
8446 (indirect_thunk_need_prefix): New function.
8447 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8448 "_nt" instead of "_bnd" for NOTRACK prefix.
8449 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8450 (output_indirect_thunk_function): Likewise.
8452 (ix86_code_end): Update output_indirect_thunk_function calls.
8453 (ix86_output_indirect_branch_via_reg): Replace
8454 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8455 (ix86_output_indirect_branch_via_push): Likewise.
8456 (ix86_output_function_return): Likewise.
8457 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8458 incompatible with -fcf-protection=branch and
8459 -fcheck-pointer-bounds.
8461 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8464 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8465 Change gcc_assert call to output_operand_lossage.
8467 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8469 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8471 2018-02-22 DJ Delorie <dj@redhat.com>
8472 Sebastian Perta <sebastian.perta@renesas.com>
8473 Oleg Endo <olegendo@gcc.gnu.org>
8475 * config/rx/rx.c (rx_rtx_costs): New function.
8476 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8478 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8480 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8482 2018-02-22 Martin Liska <mliska@suse.cz>
8485 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8486 Add "native" as a possible value.
8488 2018-02-22 Martin Liska <mliska@suse.cz>
8491 * config/i386/i386.c (ix86_option_override_internal):
8492 Add "native" as a possible value for -march and -mtune.
8494 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8497 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8498 to all type variants.
8500 PR tree-optimization/84503
8501 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8502 width as info->bitpos + info->bitsize - start.
8503 (merged_store_group::merge_overlapping): Simplify width computation.
8504 (check_no_overlap): New function.
8505 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8506 start + width and last_order of the group, fail if check_no_overlap
8508 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8509 to group if check_no_overlap fails.
8511 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8513 * config/rs6000/altivec.md: Delete contraint arguments to
8514 define_expand, define_split, and define_peephole2, and in
8515 define_insn_and_split if always unused.
8516 * config/rs6000/darwin.md: Ditto.
8517 * config/rs6000/dfp.md: Ditto.
8518 * config/rs6000/rs6000.md: Ditto.
8519 * config/rs6000/sync.md: Ditto.
8520 * config/rs6000/vector.md: Ditto.
8521 * config/rs6000/vsx.md: Ditto.
8523 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8525 * config/rs6000/altivec.md: Write output control strings as braced
8526 blocks instead of double-quoted strings.
8527 * config/rs6000/darwin.md: Ditto.
8528 * config/rs6000/rs6000.md: Ditto.
8529 * config/rs6000/vector.md: Ditto.
8530 * config/rs6000/vsx.md: Ditto.
8532 2018-02-21 Jason Merrill <jason@redhat.com>
8534 PR c++/84314 - ICE with templates and fastcall attribute.
8535 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8537 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8539 * ipa-cp.c (determine_versionability): Fix comment typos.
8541 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8544 * ipa-cp.c (determine_versionability): Do not version functions caling
8547 2018-02-21 Martin Liska <mliska@suse.cz>
8550 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8551 Add "native" as a possible value.
8552 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8553 the macro when native cpu detection is available.
8555 2018-02-21 Martin Liska <mliska@suse.cz>
8558 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8559 Add "native" as a possible value.
8560 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8561 when native cpu detection is available.
8563 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8564 Martin Sebor <msebor@redhat.com>
8566 PR tree-optimization/84478
8567 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8569 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8570 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8571 support which is conservatively correct, for 2 only stay conservative
8572 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8573 argument to the 2 argument get_range_strlen, adjust 6 arg
8574 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8576 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8577 (gimple_fold_builtin_strlen): Pass true as last argument to
8580 2018-02-20 Martin Sebor <msebor@redhat.com>
8583 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8584 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8585 (builtin_memref::builtin_memref): Factor out parts into
8586 set_base_and_offset and call it.
8588 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8591 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8592 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8593 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8594 search at the associated MODE_INT.
8596 2018-02-20 Jeff Law <law@redhat.com>
8599 PR tree-optimization/81592
8601 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8602 for range data rather than using global data.
8603 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8604 range data rather than using global data.
8605 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8606 pass it to children as needed.
8607 (struct directive::fmtresult): Similarly.
8608 (struct directive::set_width): Similarly.
8609 (struct directive::set_precision): Similarly.
8610 (format_integer, format_directive, parse_directive): Similarly.
8611 (format_none): Accept unnamed vr_values parameter.
8612 (format_percent, format_floating, format_character): Similarly.
8613 (format_string, format_plain): Similarly.
8614 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8615 the EVRP range analyzer for range data rather than using global data.
8616 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8617 gimple-ssa-evrp-analyze.h
8618 (class sprintf_dom_walker): Add after_dom_children member function.
8619 Add evrp_range_analyzer member.
8620 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8621 range analyzer as needed.
8622 (sprintf_dom_walker::after_dom_children): New member function.
8623 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8625 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8626 (evrp_range_analyzer::pop_to_marker): Likewise.
8628 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8630 PR tree-optimization/84419
8631 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8632 with the required type if its current type is compatible but
8635 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8638 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8639 after vectorization.
8641 2018-02-20 Martin Liska <mliska@suse.cz>
8644 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8645 possible values if we don't have a hint.
8647 2018-02-20 Martin Liska <mliska@suse.cz>
8651 * final.c (shorten_branches): Build align_tab array with one
8653 * opts.c (finish_options): Add alignment option limit check.
8654 (MAX_CODE_ALIGN): Likewise.
8655 (MAX_CODE_ALIGN_VALUE): Likewise.
8656 * doc/invoke.texi: Document maximum allowed option value for
8657 all -falign-* options.
8659 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8662 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8663 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8664 * var-tracking.c (emit_note_insn_var_location): Remove all references
8665 to NOTE_INSN_CALL_ARG_LOCATION.
8666 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8667 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8668 Use copy_rtx_if_shared.
8669 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8670 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8671 (dwarf2out_var_location): Remove handling of
8672 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8674 * final.c (final_scan_insn): Remove all references to
8675 NOTE_INSN_CALL_ARG_LOCATION.
8676 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8677 before dumping final insns.
8678 * except.c (emit_note_eh_region_end): Remove all references to
8679 NOTE_INSN_CALL_ARG_LOCATION.
8680 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8681 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8682 * config/arc/arc.c (hwloop_optimize): Likewise.
8683 * config/arm/arm.c (create_fix_barrier): Likewise.
8684 * config/s390/s390.c (s390_chunkify_start): Likewise.
8685 * config/sh/sh.c (find_barrier): Likewise.
8686 * config/i386/i386.c (rest_of_insert_endbranch,
8687 ix86_seh_fixup_eh_fallthru): Likewise.
8688 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8689 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8690 * config/frv/frv.c (frv_function_prologue): Likewise.
8691 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8693 (note_outside_basic_block_p): Remove all references to
8694 NOTE_INSN_CALL_ARG_LOCATION.
8695 * gengtype.c (adjust_field_rtx_def): Likewise.
8696 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8698 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8699 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8702 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8705 PR tree-optimization/84452
8706 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8707 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8710 2018-02-19 Martin Liska <mliska@suse.cz>
8713 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8715 2018-02-19 Martin Liska <mliska@suse.cz>
8716 Richard Sandiford <richard.sandiford@linaro.org>
8718 PR tree-optimization/82491
8719 * gimple-fold.c (get_base_constructor): Make earlier bail out
8722 2018-02-19 Carl Love <cel@us.ibm.com>
8724 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8725 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8727 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8728 P8V_BUILTIN_VEC_NEG.
8730 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8732 * config/rl78/rl78.md (movdf): New define expand.
8734 2018-02-19 Martin Liska <mliska@suse.cz>
8737 * doc/invoke.texi: Fix typo.
8738 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8740 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8742 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8743 handle rs6000_single_float and rs6000_double_float specially for
8746 2018-02-16 Jeff Law <law@redhat.com>
8748 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8751 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8754 * ipa-inline.c (inline_small_functions): Fix a typo.
8756 2018-02-16 Nathan Sidwell <nathan@acm.org>
8758 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8760 2018-02-16 Carl Love <cel@us.ibm.com>
8762 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8763 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8764 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8765 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8766 expansion to P8V_BUILTIN_VEC_FLOAT2.
8768 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8770 PR rtl-optimization/70023
8771 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8772 src_regno into account.
8774 2018-02-16 Carl Love <cel@us.ibm.com>
8776 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8777 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8778 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8779 * config/rs6000/rs6000.c: Remove case statements for
8780 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8781 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8782 and P9V_BUILTIN_VEC_VINSERT4B.
8783 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8784 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8785 * config/rs6000/vsx.md:
8786 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8789 2018-02-16 Carl Love <cel@us.ibm.com>
8791 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8793 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8795 * config/rs6000/rs6000-c.c: Add the definitions for
8796 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8797 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8798 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8799 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8800 definition for insert4b and define insn *insert3b_internal.
8801 * doc/extend.texi: Add documentation for vec_extract4b.
8803 2018-02-16 Nathan Sidwell <nathan@acm.org>
8805 * doc/extend.texi (Backwards Compatibility): Mention friend
8806 injection. Note for-scope is deprecated.
8807 * doc/invoke.texi (-ffriend-injection): Deprecate.
8809 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8811 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8812 that moved to I2, also allow destinations that are a paradoxical
8813 subreg (instead of a normal reg).
8815 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8818 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8821 2018-02-16 Richard Biener <rguenther@suse.de>
8823 PR tree-optimization/84037
8824 PR tree-optimization/84016
8826 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8827 Adjust vec_construct for the fact we need additional higher latency
8828 128bit inserts for AVX256 and AVX512 vector builds.
8829 (ix86_add_stmt_cost): Scale vector construction cost for
8832 2018-02-16 Richard Biener <rguenther@suse.de>
8834 PR tree-optimization/84417
8835 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8836 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8837 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8839 2018-02-16 Martin Liska <mliska@suse.cz>
8842 * internal-fn.def (ASAN_CHECK): Set proper flags.
8843 (ASAN_MARK): Likewise.
8845 2018-02-16 Julia Koval <julia.koval@intel.com>
8847 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8848 from PTA_CANNONLAKE.
8850 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8853 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8854 Use ++iter rather than iter++ for std::list iterators.
8855 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8856 defer deleting them until all nodes in the forest are processed. Do
8857 free even leaf nodes. Change to_process into auto_vec.
8860 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8861 * vec.h (vec_default_construct): Use memset instead of placement new
8862 if BROKEN_VALUE_INITIALIZATION is defined.
8863 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8864 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8867 PR rtl-optimization/83723
8868 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8869 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8870 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8872 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8874 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8876 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8878 PR rtl-optimization/81443
8879 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8880 from inner REGs to paradoxical SUBREGs.
8882 2018-02-16 Richard Biener <rguenther@suse.de>
8884 PR tree-optimization/84399
8885 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8886 For operands we can analyze at their definition make sure we can
8887 analyze them at each use as well.
8889 2018-02-16 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/84190
8892 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8893 volatile accesses if the decl isn't volatile.
8895 2018-02-15 Jason Merrill <jason@redhat.com>
8897 PR c++/84314 - ICE with templates and fastcall attribute.
8898 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8899 TYPE_CANONICAL on an existing type.
8901 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8903 PR tree-optimization/84383
8904 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8905 dstoff nor call operand_equal_p if dstbase is NULL.
8907 PR tree-optimization/84334
8908 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8909 also a CONSTANT_CLASS_P, punt.
8911 2018-02-14 Jim Wilson <jimw@sifive.com>
8913 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8914 first SMALL_OPERAND check. New local min_second_step. Move assert
8915 to where locals are set. Add TARGET_RVC support.
8916 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8918 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8920 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8922 2018-02-14 Martin Sebor <msebor@redhat.com>
8924 PR tree-optimization/83698
8925 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8926 arrays constrain the offset range to their bounds.
8927 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8928 (builtin_access::overlap): Avoid setting the size of overlap if it's
8930 (maybe_diag_overlap): Also consider arrays when deciding what values
8931 of offsets to include in diagnostics.
8933 2018-02-14 Martin Sebor <msebor@redhat.com>
8936 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8937 that correspond to the kind of a declaration.
8939 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8942 * config/pa/pa.md: Load address of PIC label using the linkage table
8943 if the label is nonlocal.
8945 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8947 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8948 warning message if user requests -maltivec=be.
8949 * doc/invoke.texi: Document deprecation of -maltivec=be.
8951 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8954 * config/rs6000/rs6000-c.c: Update definitions for
8955 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8956 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8958 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8961 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8962 add _get_ssp intrinsics. Remove argument from
8963 __builtin_ia32_rdssp[d|q].
8964 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8965 * config/i386/i386-builtin.def: Remove argument from
8966 __builtin_ia32_rdssp[d|q].
8967 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8968 ix86_expand_special_args_builtin for _rdssp[d|q].
8969 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8970 Clear register before usage.
8971 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8972 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8974 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8976 PR tree-optimization/84357
8977 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8978 operand 1 of an ARRAY_REF too.
8980 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8983 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8984 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8986 (set_of_reg): New struct.
8987 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8988 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8989 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8991 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8992 Split into bitclr, bitset, bitinvert patterns if appropriate.
8993 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8994 use rx_fuse_in_memory_bitop.
8995 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8996 to named insn, correct maximum insn length.
8998 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9001 * machmode.def: Define a complex mode for PARTIAL_INT.
9002 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9004 * doc/rtl.texi: Document CSPImode.
9005 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9007 (msp430_hard_regno_nregs_with_padding): Likewise.
9009 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9012 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9014 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9016 PR rtl-optimization/84169
9017 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9018 we generated a parallel as new i3 and we split that to new i2 and i3
9019 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9020 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9021 those to i2, not i1. Partially rewrite this scan code.
9023 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9026 * stor-layout.c (place_field): For variable length fields, adjust
9027 offset_align afterwards not just based on the field's alignment,
9028 but also on the size.
9031 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9032 of exps and logs in the use_exp2 case.
9034 2018-02-13 Jeff Law <law@redhat.com>
9036 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9039 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9042 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9046 * common.opt (gas-loc-support, gas-locview-support): New.
9047 (ginline-points, ginternal-reset-location-views): New.
9048 * doc/invoke.texi: Document them. Use @itemx where intended.
9049 (gvariable-location-views): Adjust.
9050 * target.def (reset_location_view): New.
9051 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9052 (TARGET_RESET_LOCATION_VIEW): New.
9053 * doc/tm.texi: Rebuilt.
9054 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9055 (dwarf2out_default_as_locview_support): New.
9056 (output_asm_line_debug_info): Use option variables.
9057 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9058 (output_loc_list): Likewise.
9059 (add_high_low_attributes): Check option variables.
9060 Don't output entry view attribute in strict mode.
9061 (gen_inlined_subroutine_die): Check option variables.
9062 (dwarf2out_inline_entry): Likewise.
9063 (init_sections_and_labels): Likewise.
9064 (dwarf2out_early_finish): Likewise.
9065 (maybe_reset_location_view): New, from...
9066 (dwarf2out_var_location): ... here. Call it.
9067 * debug.h (dwarf2out_default_as_loc_support): Declare.
9068 (dwarf2out_default_as_locview_support): Declare.
9069 * hooks.c (hook_int_rtx_insn_0): New.
9070 * hooks.h (hook_int_rtx_insn_0): Declare.
9071 * toplev.c (process_options): Take -gas-loc-support and
9072 -gas-locview-support from dwarf2out. Enable
9073 -gvariable-location-views by default only with locview
9074 assembler support. Enable -ginternal-reset-location-views by
9075 default only if the target defines the corresponding hook.
9076 Enable -ginline-points by default if location views are
9077 enabled; force it disabled if statement frontiers are
9079 * tree-inline.c (expand_call_inline): Check option variables.
9080 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9082 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9084 PR tree-optimization/84321
9085 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9086 handling. Also check whether the anti-range contains any values
9087 that satisfy the mask; switch to a VR_RANGE if not.
9089 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9092 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9094 2018-02-13 Martin Jambor <mjambor@suse.cz>
9097 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9098 of call statements, also set location of a load to a temporary.
9100 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9102 * config/rl78/rl78.c (add_vector_labels): New function.
9103 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9104 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9105 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9106 which checks that no arguments are passed.
9107 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9108 * doc/extend.texi: Documentation for the new attribute.
9110 2018-02-13 Andreas Schwab <schwab@suse.de>
9112 * config/riscv/linux.h (CPP_SPEC): Define.
9114 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9117 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9118 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9119 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9120 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9121 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9122 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9123 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9124 temporarily for AES and PCLMUL builtins.
9126 PR tree-optimization/84339
9127 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9128 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9132 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9133 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9134 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9136 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9138 * omp-simd-clone.h: New file.
9139 * omp-simd-clone.c: Include omp-simd-clone.h.
9140 (expand_simd_clones): No longer static.
9141 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9142 cgraph.h and omp-simd-clone.h.
9143 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9144 (vect_recog_widen_shift_pattern): Formatting fix.
9145 (vect_pattern_recog_1): Don't check optab for calls.
9148 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9149 operands[2] into a REG before using gen_lowpart on it.
9151 2018-02-12 Jeff Law <law@redhat.com>
9154 * config/sh/sh.c (find_barrier): Consider a sibling call
9157 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9158 successfully back substituting a reg.
9160 2018-02-12 Richard Biener <rguenther@suse.de>
9162 PR tree-optimization/84037
9163 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9164 parameter, move visited init to caller.
9165 (vect_slp_analyze_operations): Separate cost from validity
9166 check, initialize visited once for all instances.
9167 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9169 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9171 (vect_model_promotion_demotion_cost): Likewise.
9172 (vectorizable_bswap): Guard cost modeling with !slp_node
9173 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9175 (vectorizable_call): Likewise.
9176 (vectorizable_conversion): Likewise.
9177 (vectorizable_assignment): Likewise.
9178 (vectorizable_shift): Likewise.
9179 (vectorizable_operation): Likewise.
9180 (vectorizable_store): Likewise.
9181 (vectorizable_load): Likewise.
9182 (vectorizable_condition): Likewise.
9183 (vectorizable_comparison): Likewise.
9185 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9188 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9189 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9190 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9192 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9195 * params.def (inline-min-speedup): Increase from 8 to 15.
9196 (max-inline-insns-auto): Decrease from 40 to 30.
9197 * ipa-split.c (consider_split): Add some buffer for function to
9198 be considered inlining candidate.
9199 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9202 2018-02-12 Richard Biener <rguenther@suse.de>
9204 PR tree-optimization/84037
9205 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9206 matched stmts if we cannot swap the non-matched ones.
9208 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9210 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9211 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9212 _mm_maskz_scalef_round_ss): New intrinsics.
9213 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9214 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9215 __builtin_ia32_scalefss_round): Remove.
9216 (__builtin_ia32_scalefsd_mask_round,
9217 __builtin_ia32_scalefss_mask_round): New intrinsics.
9218 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9219 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9220 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9221 "<round_constraint>")): Changed to ...
9222 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9223 "<round_scalar_constraint>")): ... this.
9224 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9225 %0, %1, %2<round_op3>}"): Changed to ...
9226 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9227 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9228 %2<round_scalar_mask_op3>}"): ... this.
9229 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9231 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9233 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9234 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9235 (_mm_maskz_sqrt_round_ss): New intrinsics.
9236 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9237 (__builtin_ia32_sqrtsd_mask_round)
9238 (__builtin_ia32_sqrtss_mask_round): New builtins.
9239 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9240 (__builtin_ia32_sqrtss_round): Remove.
9241 (__builtin_ia32_sqrtsd_mask_round)
9242 (__builtin_ia32_sqrtss_mask_round): New builtins.
9243 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9244 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9245 ((match_operand:VF_128 1 "vector_operand"
9246 "xBm,<round_constraint>")): Changed to ...
9247 ((match_operand:VF_128 1 "vector_operand"
9248 "xBm,<round_scalar_constraint>")): ... this.
9249 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9250 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9251 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9252 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9253 %<iptr>1<round_scalar_mask_op3>}): ... this.
9254 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9255 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9257 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9260 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9261 Cast vec_cmpeq result to correct type.
9262 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9263 Cast vec_cmpgt result to correct type.
9265 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9267 * final.c (final_scan_insn_1): Renamed from...
9268 (final_scan_insn): ... this. New wrapper, to recover
9269 seen from the outermost call in recursive ones.
9270 * config/sparc/sparc.c (output_return): Drop seen from call.
9271 (output_sibcall): Likewise.
9272 * config/visium/visium.c (output_branch): Likewise.
9274 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9276 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9279 2018-02-10 Alan Modra <amodra@gmail.com>
9282 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9283 Specify LR as an input.
9285 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9288 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9289 remove_member_access_dummy_vars): New functions.
9290 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9291 lower_omp_1, execute_lower_omp): Use them.
9293 PR rtl-optimization/84308
9294 * shrink-wrap.c (spread_components): Release todo vector.
9296 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9298 PR rtl-optimization/57193
9299 * ira-color.c (struct allocno_color_data): Add member
9300 conflict_allocno_hard_prefs.
9301 (update_conflict_allocno_hard_prefs): New.
9302 (bucket_allocno_compare_func): Add a preference based on
9303 conflict_allocno_hard_prefs.
9304 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9305 (color_allocnos): Remove a dead code. Initiate
9306 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9308 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9311 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9312 constraint from =wa to wa. Avoid a subreg on the output operand,
9313 instead use a pseudo and subreg it in a move.
9314 (p9_xxbrd_<mode>): Changed to ...
9315 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9316 (p9_xxbrd_v2df): New expander.
9317 (p9_xxbrw_<mode>): Changed to ...
9318 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9319 (p9_xxbrw_v4sf): New expander.
9321 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9323 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9325 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9328 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9329 multiply in 32-bit mode.
9330 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9331 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9334 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9336 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9337 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9338 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9339 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9341 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9344 * dwarf2out.c (is_trivial_indirect_ref): New function.
9345 (dwarf2out_late_global_decl): Do not generate a location
9346 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9347 and that are not defined in the current unit.
9349 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9351 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9352 instead of a libcall for UNORDERED.
9354 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9357 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9358 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9360 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9363 * config/s390/s390.c (s390_set_current_function): Invoke
9364 s390_indirect_branch_settings also if fndecl didn't change.
9366 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9368 * config/rs6000/rs6000.md (blockage): Set length to zero.
9370 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9372 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9374 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9377 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9378 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9382 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9383 PARALLEL incoming that failed vt_get_decl_and_offset check.
9386 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9387 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9389 (get_variable_section): For decls in named .bss* sections pass true as
9390 second argument to bss_initializer_p.
9392 2018-02-09 Marek Polacek <polacek@redhat.com>
9393 Jakub Jelinek <jakub@redhat.com>
9396 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9397 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9398 Sync some changes from cxx_fold_indirect_ref.
9400 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9402 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9404 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9405 (BLOCK_INLINE_ENTRY_LABEL): New.
9406 (dwarf2out_var_location): Disregard inline entry markers.
9407 (inline_entry_data): New struct.
9408 (inline_entry_data_hasher): New hashtable type.
9409 (inline_entry_data_hasher::hash): New.
9410 (inline_entry_data_hasher::equal): New.
9411 (inline_entry_data_table): New variable.
9412 (add_high_low_attributes): Add DW_AT_entry_pc and
9413 DW_AT_GNU_entry_view attributes if a pending entry is found
9414 in inline_entry_data_table. Add old entry_pc attribute only
9415 if debug nonbinding markers are disabled.
9416 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9417 markers are enabled.
9418 (block_within_block_p, dwarf2out_inline_entry): New.
9419 (dwarf2out_finish): Check that no entries remained in
9420 inline_entry_data_table.
9421 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9422 (final_scan_insn, notice_source_line): Likewise.
9423 (rest_of_clean_state): Skip inline entry markers.
9424 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9426 * gimple.c (gimple_build_debug_inline_entry): New.
9427 * gimple.h (enum gimple_debug_subcode): Add
9428 GIMPLE_DEBUG_INLINE_ENTRY.
9429 (gimple_build_debug_inline_entry): Declare.
9430 (gimple_debug_inline_entry_p): New.
9431 (gimple_debug_nonbind_marker_p): Adjust.
9432 * insn-notes.def (INLINE_ENTRY): New.
9433 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9434 inline entry marker notes.
9435 (print_insn): Likewise.
9436 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9437 (INSN_DEBUG_MARKER_KIND): Likewise.
9438 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9439 * tree-inline.c (expand_call_inline): Build and insert
9440 debug_inline_entry stmt.
9441 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9442 inline entry blocks early, if nonbind markers are enabled.
9443 (dump_scope_block): Dump fragment info.
9444 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9445 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9446 (gimple_build_debug_inline_entry): New.
9447 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9448 Enable/disable inline entry points too.
9449 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9450 (DEBUG_INSN): Describe inline entry markers.
9452 * common.opt (gvariable-location-views): New.
9453 (gvariable-location-views=incompat5): New.
9454 * config.in: Rebuilt.
9455 * configure: Rebuilt.
9456 * configure.ac: Test assembler for view support.
9457 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9458 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9459 * dwarf2out.c (var_loc_view): New typedef.
9460 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9461 (dwarf2out_locviews_in_attribute): New.
9462 (dwarf2out_locviews_in_loclist): New.
9463 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9464 (enum dw_line_info_opcode): Add LI_adv_address.
9465 (struct dw_line_info_table): Add view.
9466 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9467 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9468 (zero_view_p): New variable.
9469 (ZERO_VIEW_P): New macro.
9470 (output_asm_line_debug_info): New.
9471 (struct var_loc_node): Add view.
9472 (add_AT_view_list, AT_loc_list): New.
9473 (add_var_loc_to_decl): Add view param. Test it against last.
9474 (new_loc_list): Add view params. Record them.
9475 (AT_loc_list_ptr): Handle loc and view lists.
9476 (view_list_to_loc_list_val_node): New.
9477 (print_dw_val): Handle dw_val_class_view_list.
9478 (size_of_die): Likewise.
9479 (value_format): Likewise.
9480 (loc_list_has_views): New.
9481 (gen_llsym): Set vl_symbol too.
9482 (maybe_gen_llsym, skip_loc_list_entry): New.
9483 (dwarf2out_maybe_output_loclist_view_pair): New.
9484 (output_loc_list): Output view list or entries too.
9485 (output_view_list_offset): New.
9486 (output_die): Handle dw_val_class_view_list.
9487 (output_dwarf_version): New.
9488 (output_compilation_unit_header): Use it.
9489 (output_skeleton_debug_sections): Likewise.
9490 (output_rnglists, output_line_info): Likewise.
9491 (output_pubnames, output_aranges): Update version comments.
9492 (output_one_line_info_table): Output view numbers in asm comments.
9493 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9494 Call maybe_gen_llsym.
9495 (loc_list_from_tree_1): Adjust.
9496 (add_AT_location_description): Create view list attribute if
9497 needed, check it's absent otherwise.
9498 (convert_cfa_to_fb_loc_list): Adjust.
9499 (maybe_emit_file): Call output_asm_line_debug_info for test.
9500 (dwarf2out_var_location): Reset views as needed. Precompute
9501 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9502 attribute. Set view.
9503 (new_line_info_table): Reset next view.
9504 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9505 (dwarf2out_source_line): Likewise. Output view resets and labels to
9506 the assembler, or select appropriate line info opcodes.
9507 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9508 (optimize_string_length): Catch it. Adjust.
9509 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9510 dw_val_class_view_list, and remove it if no longer needed.
9511 (hash_loc_list): Hash view numbers.
9512 (loc_list_hasher::equal): Compare them.
9513 (optimize_location_lists): Check whether a view list symbol is
9514 needed, and whether the locview attribute is present, and
9515 whether they match. Remove the locview attribute if no longer
9517 (index_location_lists): Call skip_loc_list_entry for test.
9518 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9519 Use output_dwarf_version.
9520 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9521 (struct dw_val_node): Add val_view_list.
9522 * final.c (SEEN_NEXT_VIEW): New.
9523 (set_next_view_needed): New.
9524 (clear_next_view_needed): New.
9525 (maybe_output_next_view): New.
9526 (final_start_function): Rename to...
9527 (final_start_function_1): ... this. Take pointer to FIRST,
9528 add SEEN parameter. Emit param bindings in the initial view.
9529 (final_start_function): Reintroduce SEEN-less interface.
9530 (final): Rename to...
9531 (final_1): ... this. Take SEEN parameter. Output final pending
9532 next view at the end.
9533 (final): Reintroduce seen-less interface.
9534 (final_scan_insn): Output pending next view before switching
9535 sections or ending a block. Mark the next view as needed when
9536 outputting variable locations. Notify debug backend of section
9537 changes, and of location view changes.
9538 (rest_of_handle_final): Adjust.
9539 * toplev.c (process_options): Autodetect value for debug variable
9540 location views option. Warn on incompat5 without -gdwarf-5.
9541 * doc/invoke.texi (gvariable-location-views): New.
9542 (gvariable-location-views=incompat5): New.
9543 (gno-variable-location-views): New.
9545 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9547 PR tree-optimization/84136
9548 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9549 that the result of find_edge is non-NULL.
9551 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9554 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9555 storing integer register in SImode. Fix cost of 256 and 512
9556 byte aligned SSE register store.
9558 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9560 * config/i386/i386.c (ix86_multiplication_cost): Fix
9561 multiplication cost for TARGET_AVX512DQ.
9563 2018-02-08 Marek Polacek <polacek@redhat.com>
9565 PR tree-optimization/84238
9566 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9569 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9571 PR tree-optimization/84265
9572 * tree-vect-stmts.c (vectorizable_store): Don't treat
9573 VMAT_CONTIGUOUS accesses as grouped.
9574 (vectorizable_load): Likewise.
9576 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9578 PR tree-optimization/81635
9579 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9580 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9581 (test_round_for_mask): New functions.
9582 (wide_int_cc_tests): Call test_round_for_mask.
9583 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9584 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9585 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9586 range returned by get_range_info.
9588 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9591 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9592 * symtab.c: Include builtins.h
9593 (symtab_node::output_to_lto_symbol_table_p): Move here
9594 from lto-streamer-out.c:output_symbol_p.
9595 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9596 (output_symbol_p): Move all logic to symtab.c
9597 (produce_symtab): Update.
9599 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9601 * config/s390/s390-opts.h (enum indirect_branch): Define.
9602 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9603 (s390_indirect_branch_via_thunk)
9604 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9605 (enum s390_indirect_branch_type): Define.
9606 * config/s390/s390.c (struct s390_frame_layout, struct
9607 machine_function): Remove.
9608 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9609 (indirect_branch_table_label_no, indirect_branch_table_name):
9611 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9612 (enum s390_indirect_branch_option): Define.
9613 (s390_return_addr_from_memory): New function.
9614 (s390_handle_string_attribute): New function.
9615 (s390_attribute_table): Add new attribute handler.
9616 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9617 (s390_indirect_branch_via_thunk): New function.
9618 (s390_indirect_branch_via_inline_thunk): New function.
9619 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9620 sibling call optimization for non z10 compiles.
9621 (s390_emit_call): Force indirect branch target to be a single
9622 register. Add r1 clobber for non-z10 compiles.
9623 (s390_emit_epilogue): Emit return jump via return_use expander.
9624 (s390_reorg): Handle JUMP_INSNs as execute targets.
9625 (s390_option_override_internal): Perform validity checks for the
9626 new command line options.
9627 (s390_indirect_branch_attrvalue): New function.
9628 (s390_indirect_branch_settings): New function.
9629 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9630 (s390_output_indirect_thunk_function): New function.
9631 (s390_code_end): Implement target hook.
9632 (s390_case_values_threshold): Implement target hook.
9633 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9635 * config/s390/s390.h (struct s390_frame_layout)
9636 (struct machine_function): Move here from s390.c.
9637 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9638 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9639 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9640 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9641 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9642 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9643 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9644 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9645 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9646 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9647 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9648 (mnemonic attribute): Add values which aren't recognized
9650 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9651 pattern for branch conversion. Fix mnemonic attribute.
9652 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9653 indirect branch via thunk if requested.
9654 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9655 ("*indirect_jump"): Disable for branch conversion using out of
9657 ("indirect_jump_via_thunk<mode>_z10")
9658 ("indirect_jump_via_thunk<mode>")
9659 ("indirect_jump_via_inlinethunk<mode>_z10")
9660 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9661 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9662 ("casesi_jump_via_inlinethunk<mode>_z10")
9663 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9664 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9665 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9666 ("*indirect2_jump"): Disable for branch conversion.
9667 ("casesi_jump"): Turn into expander and expand patterns for branch
9669 ("return_use"): New expander.
9670 ("*return"): Emit return via thunk and rename it to ...
9671 ("*return<mode>"): ... this one.
9672 * config/s390/s390.opt: Add new options and and enum for the
9675 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9677 * lra-constraints.c (match_reload): Unconditionally use
9678 gen_lowpart_SUBREG, rather than selecting between that
9679 and equivalent gen_rtx_SUBREG code.
9681 2018-02-08 Richard Biener <rguenther@suse.de>
9683 PR tree-optimization/84233
9684 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9685 changed flag instead of boguously re-using phi_inserted.
9687 2018-02-08 Martin Jambor <mjambor@suse.cz>
9689 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9690 static local variables.
9692 2018-02-08 Richard Biener <rguenther@suse.de>
9694 PR tree-optimization/84278
9695 * tree-vect-stmts.c (vectorizable_store): When looking for
9696 smaller vector types to perform grouped strided loads/stores
9697 make sure the mode is supported by the target.
9698 (vectorizable_load): Likewise.
9700 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9702 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9703 Increase LDP/STP opportunities by adding adjacent callee-saves.
9705 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9707 PR rtl-optimization/84068
9708 PR rtl-optimization/83459
9709 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9711 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9713 PR tree-optimization/84224
9714 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9715 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9718 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9721 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9722 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9723 count, remove one USE.
9725 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9727 * doc/install.texi (Configuration): Document the
9728 --with-long-double-format={ibm,ieee} PowerPC configuration
9732 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9733 Convert from define_expand to be define_insn_and_split. Rework
9734 float/double/_Float128 conversions to QI/HI/SImode to work with
9735 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9736 conversions to QI/HImode types did a store and then a load to
9737 truncate the value. For conversions to VSX registers, don't split
9738 the insn, instead emit the code directly. Use the code iterator
9739 any_fix to combine signed and unsigned conversions.
9740 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9741 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9742 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9743 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9744 (fix_<mode>di2_hw): Likewise.
9745 (fixuns_<mode>di2_hw): Likewise.
9746 (fix_<mode>si2_hw): Likewise.
9747 (fixuns_<mode>si2_hw): Likewise.
9748 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9749 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9750 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9751 fix<uns>_trunc<SFDF:mode>si2_p8.
9752 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9753 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9754 (fix<uns>_<mode>_mem): Likewise.
9755 (fctiw<u>z_<mode>_mem): Likewise.
9756 (fix<uns>_<mode>_mem): Likewise.
9757 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9758 the register allocator from doing a direct move to the GPRs to do
9759 a store, and instead use the ISA 3.0 store byte/half-word from
9760 vector register instruction. For IEEE 128-bit floating point,
9761 also optimize stores of 32-bit ints.
9762 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9764 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9766 * genextract.c (push_pathstr_operand): New function to support
9768 (walk_rtx): Call push_pathstr_operand.
9769 (print_path): Support [a-zA-Z].
9771 2018-02-07 Richard Biener <rguenther@suse.de>
9773 PR tree-optimization/84037
9774 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9775 (cse_and_gimplify_to_preheader): Declare.
9776 (vect_get_place_in_interleaving_chain): Likewise.
9777 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9779 (_loop_vec_info::~_loop_vec_info): Delete it.
9780 (cse_and_gimplify_to_preheader): New function.
9781 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9782 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9783 (vectorizable_load): Likewise. For grouped stores always base
9784 the IV on the first element.
9785 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9786 condition before gimplifying.
9788 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9790 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9791 *DIV_EXPR and *MOD_EXPR.
9793 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9796 * config/i386/i386.c (ix86_option_override_internal): Mask out
9797 the CF_SET bit when checking -fcf-protection.
9799 2018-02-07 Tom de Vries <tom@codesourcery.com>
9802 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9805 2018-02-07 Richard Biener <rguenther@suse.de>
9807 PR tree-optimization/84204
9808 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9811 PR tree-optimization/84205
9812 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9813 special-case isl_ast_op_zdiv_r.
9815 PR tree-optimization/84223
9816 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9817 Only add conditions from within the region.
9818 (gather_bbs::after_dom_children): Adjust.
9820 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9823 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9824 * config/avr/avr.md: Only post-reload split REG-REG moves if
9825 either register is GENERAL_REG_P.
9827 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9829 PR tree-optimization/84235
9830 * tree-ssa-scopedtables.c
9831 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9832 if the subtraction is performed in floating point type where NaNs are
9833 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9834 build 1. Formatting fix.
9836 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9839 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9840 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9841 and skip it regardless of bb boundaries. Use CALL_P macro,
9842 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9843 unnecessarily, formatting fix.
9845 2018-02-06 Michael Collison <michael.collison@arm.com>
9847 * config/arm/thumb2.md:
9848 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9849 (*thumb_mov_notscc): Ditto.
9851 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9854 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9855 unsigned_fix, not "s".
9857 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9859 * configure.ac (gcc_fn_eh_frame_ro): New function.
9860 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9861 correct .eh_frame permissions.
9862 * configure: Regenerate.
9864 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9866 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9869 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9871 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9872 Display warning message for -mno-speculate-indirect-jumps.
9874 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9876 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9878 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9880 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9882 PR tree-optimization/84225
9883 * tree-eh.c (find_trapping_overflow): Only call
9884 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9886 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9889 * config/i386/i386.c: Reimplement the check of possible options
9890 -mibt/-mshstk conbination. Change error messages.
9891 * doc/invoke.texi: Fix a typo: remove extra '='.
9893 2018-02-06 Marek Polacek <polacek@redhat.com>
9895 PR tree-optimization/84228
9896 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9898 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9901 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9902 emitted arch directives.
9903 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9904 __ARM_FEATURE_COPROC before changing architectures.
9906 2018-02-06 Richard Biener <rguenther@suse.de>
9908 * config/i386/i386.c (print_reg): Fix typo.
9909 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9911 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9913 * configure: Regenerate.
9915 2018-02-05 Martin Sebor <msebor@redhat.com>
9917 PR tree-optimization/83369
9918 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9921 2018-02-05 Martin Liska <mliska@suse.cz>
9923 * doc/invoke.texi: Cherry-pick upstream r323995.
9925 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9927 * ira.c (ira_init_register_move_cost): Adjust comment.
9929 2018-02-05 Martin Liska <mliska@suse.cz>
9931 PR gcov-profile/84137
9932 * doc/gcov.texi: Fix typo in documentation.
9934 2018-02-05 Martin Liska <mliska@suse.cz>
9936 PR gcov-profile/83879
9937 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9938 using dlopen functionality.
9940 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9942 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9943 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9944 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9945 _mm_maskz_range_round_ss): New intrinsics.
9946 (__builtin_ia32_rangesd128_round)
9947 (__builtin_ia32_rangess128_round): Remove.
9948 (__builtin_ia32_rangesd128_mask_round,
9949 __builtin_ia32_rangess128_mask_round): New builtins.
9950 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9951 __builtin_ia32_rangess128_round): Remove.
9952 (__builtin_ia32_rangesd128_mask_round,
9953 __builtin_ia32_rangess128_mask_round): New builtins.
9954 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9955 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9956 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9957 "<round_saeonly_constraint>")): Changed to ...
9958 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9959 "<round_saeonly_scalar_constraint>")): ... this.
9960 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9961 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9962 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9963 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9964 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9966 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9968 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9970 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9971 Remove all values except native, 8540 and 8548.
9973 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9975 * config/i386/i386.c (ix86_output_function_return): Pass
9976 INVALID_REGNUM, instead of -1, as invalid register number to
9977 indirect_thunk_name and output_indirect_thunk.
9979 2018-02-02 Julia Koval <julia.koval@intel.com>
9981 * config.gcc: Add -march=icelake.
9982 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9983 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9984 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9985 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9986 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9987 (processor_target_table): Add icelake.
9988 (ix86_option_override_internal): Handle new PTAs.
9989 (get_builtin_code_for_version): Handle icelake.
9990 (M_INTEL_COREI7_ICELAKE): New.
9991 (fold_builtin_cpu): Handle icelake.
9992 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9993 * doc/invoke.texi: Add -march=icelake.
9995 2018-02-02 Julia Koval <julia.koval@intel.com>
9997 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9998 to wide_int_bitmask.
9999 * wide-int-bitmask.h: New.
10001 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10004 * config/i386/i386.md: Replace Pmode with word_mode in
10005 builtin_setjmp_setup and builtin_longjmp to support x32.
10007 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10011 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10013 (rs6000_supported_cpu_names): New static variable.
10014 (linux_cpu_translation_table): Likewise.
10015 (elf_platform) <cpu>: Define new static variable and use it.
10016 Translate kernel AT_PLATFORM name to canonical name if needed.
10017 Error if platform name is unknown.
10019 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10022 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10024 2018-02-01 Jeff Law <law@redhat.com>
10027 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10028 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10029 the scratch if RELEASE_VIA_POP is false.
10030 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10031 If we have to save a temporary register, decrement SIZE appropriately.
10032 Pass new arguments to release_scratch_register_on_entry.
10033 (ix86_adjust_stack_and_probe): Likewise.
10034 (ix86_emit_probe_stack_range): Pass new arguments to
10035 release_scratch_register_on_entry.
10037 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10039 PR rtl-optimization/84157
10040 * combine.c (change_zero_ext): Use REG_P predicate in
10041 front of HARD_REGISTER_P predicate.
10043 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10045 * config/avr/avr.c (avr_option_override): Move disabling of
10046 -fdelete-null-pointer-checks to...
10047 * common/config/avr/avr-common.c (avr_option_optimization_table):
10050 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10052 PR tree-optimization/81635
10053 * tree-data-ref.c (split_constant_offset_1): For types that
10054 wrap on overflow, try to use range info to prove that wrapping
10057 2018-02-01 Renlin Li <renlin.li@arm.com>
10060 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10061 TAILCALL_ADDR_REGS.
10062 (aarch64_register_move_cost): Likewise.
10063 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10064 TAILCALL_ADDR_REGS.
10065 (REG_CLASS_NAMES): Likewise.
10066 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10067 TAILCALL_ADDR_REGS. Remove IP registers.
10068 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10070 2018-02-01 Richard Biener <rguenther@suse.de>
10072 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10073 for specifying RPO order and allow NULL for that.
10074 * domwalk.c (dom_walker::dom_walker): Likewise.
10075 (dom_walker::walk): Handle NULL RPO order.
10076 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10078 (rewrite_update_dom_walker): Likewise.
10079 (mark_def_dom_walker): Likewise.
10081 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10083 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10084 (aarch64_maybe_expand_sve_subreg_move): Declare.
10085 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10086 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10088 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10089 that are semantically a reverse operation.
10090 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10091 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10092 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10094 (aarch64_can_change_mode_class): For big-endian, forbid changes
10095 between two SVE modes if they have different element sizes.
10097 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10099 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10100 the TImode handling for big-endian targets.
10102 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10104 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10105 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10107 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10108 Remove BSWAP handing for big-endian targets and use the form of
10109 LD1RQ appropariate for the mode.
10111 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10113 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10114 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10115 duplicated element.
10117 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10120 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10121 check for operands that need to go through aarch64_sve_reload_be.
10123 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10125 PR tree-optimization/81661
10126 PR tree-optimization/84117
10127 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10128 * tree-eh.c: Include gimplify.h.
10129 (find_trapping_overflow, replace_trapping_overflow,
10130 rewrite_to_non_trapping_overflow): New functions.
10131 * tree-vect-loop.c: Include tree-eh.h.
10132 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10133 * tree-data-ref.c: Include tree-eh.h.
10134 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10136 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10138 PR rtl-optimization/84123
10139 * combine.c (change_zero_ext): Check if hard register satisfies
10140 can_change_dest_mode before calling gen_lowpart_SUBREG.
10142 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10145 * ira.c (ira_init_register_move_cost): Remove assert.
10147 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10149 PR rtl-optimization/84071
10150 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10151 * doc/tm.texi: Regenerate.
10153 2018-01-31 Richard Biener <rguenther@suse.de>
10155 PR tree-optimization/84132
10156 * tree-data-ref.c (analyze_miv_subscript): Properly
10157 check whether evolution_function_is_affine_multivariate_p
10158 before calling gcd_of_steps_may_divide_p.
10160 2018-01-31 Julia Koval <julia.koval@intel.com>
10163 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10164 * config/i386/i386.md (rdpid_rex64) New.
10165 (rdpid): Make 32bit only.
10167 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10170 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10171 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10173 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10176 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10178 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10180 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10182 PR rtl-optimization/84071
10183 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10184 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10186 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10188 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10189 (arc_attribute_table): Add 'aux' attribute.
10190 (arc_in_small_data_p): Consider aux like variables.
10191 (arc_is_aux_reg_p): New function.
10192 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10193 (arc_get_aux_arg): New function.
10194 (prepare_move_operands): Handle aux-register access.
10195 (arc_handle_aux_attribute): New function.
10196 * doc/extend.texi (ARC Variable attributes): Add subsection.
10198 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10200 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10201 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10202 (arc_attribute_table): Add 'uncached' attribute.
10203 (arc_print_operand): Print '.di' flag for uncached memory
10205 (arc_in_small_data_p): Do not consider for small data the uncached
10207 (arc_is_uncached_mem_p): New function.
10208 * config/arc/predicates.md (compact_store_memory_operand): Check
10209 for uncached memory accesses.
10210 (nonvol_nonimm_operand): Likewise.
10211 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10213 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10216 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10217 falign-loops=): Add Optimization flag.
10219 2018-01-30 Jeff Law <law@redhat.com>
10222 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10223 INT_REGISTERS_SAVED. Check it prior to calling
10224 get_scratch_register_on_entry.
10225 (ix86_adjust_stack_and_probe): Similarly.
10226 (ix86_emit_probe_stack_range): Similarly.
10227 (ix86_expand_prologue): Corresponding changes.
10229 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10232 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10233 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10235 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10238 * lra-constraints.c (curr_insn_transform): Process AND in the
10241 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10243 PR rtl-optimization/83986
10244 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10245 dependence against last_pending_memory_flush in addition to
10246 pending_jump_insns.
10248 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10250 PR tree-optimization/81611
10251 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10254 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10257 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10260 2018-01-30 Richard Biener <rguenther@suse.de>
10261 Jakub Jelinek <jakub@redhat.com>
10263 PR tree-optimization/84111
10264 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10265 inner loops added during recursion, as they don't have up-to-date
10268 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10271 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10272 (can_inline_edge_by_limits_p): ... here.
10273 (can_early_inline_edge_p, check_callers,
10274 update_caller_keys, update_callee_keys, recursive_inlining,
10275 add_new_edges_to_heap, speculation_useful_p,
10276 inline_small_functions,
10277 inline_small_functions, flatten_function,
10278 inline_to_all_callers_1): Update.
10280 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10282 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10285 2018-01-30 Richard Biener <rguenther@suse.de>
10287 PR tree-optimization/83008
10288 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10289 invariant and constant vector uses in stmts when they need
10290 more than one stmt.
10292 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10295 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10296 * configure: Regenerate.
10298 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10300 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10302 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10303 Use gen_rtx_REG rather than gen_lowpart.
10305 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10307 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10308 rather than 0 when creating partial subregs.
10310 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10312 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10315 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10318 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10319 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10320 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10321 flags. This restores the settings used before the 2017-07-24.
10322 Turning off pre increment/decrement/modify allows IVOPTS to
10323 optimize DF/SF loops where the index is an int.
10325 2018-01-29 Richard Biener <rguenther@suse.de>
10326 Kelvin Nilsen <kelvin@gcc.gnu.org>
10329 * tree-vect-stmts.c (vectorizable_call): Don't call
10330 targetm.vectorize_builtin_md_vectorized_function if callee is
10333 2018-01-22 Carl Love <cel@us.ibm.com>
10335 * doc/extend.tex: Fix typo in second arg in
10336 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10338 2018-01-29 Richard Biener <rguenther@suse.de>
10340 PR tree-optimization/84086
10341 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10342 (flush_ssaname_freelist): When SSA names were released reset
10343 the SCEV hash table.
10345 2018-01-29 Richard Biener <rguenther@suse.de>
10347 PR tree-optimization/84057
10348 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10349 removed paths when removing edges.
10351 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10353 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10354 -mfunction-return=@var{choice}.
10356 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10358 PR diagnostic/84034
10359 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10360 Handle CR like TAB.
10361 (layout::print_source_line): Likewise.
10362 (test_get_line_width_without_trailing_whitespace): Add test cases.
10364 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10366 PR middle-end/84040
10367 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10370 2018-01-26 Jim Wilson <jimw@sifive.com>
10372 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10374 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10377 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10379 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10380 and CMP + SUB-immediate -> SUBS.
10382 2018-01-26 Martin Sebor <msebor@redhat.com>
10384 PR tree-optimization/83896
10385 * tree-ssa-strlen.c (get_string_len): Rename...
10386 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10387 Avoid assuming length is constant.
10388 (handle_char_store): Use HOST_WIDE_INT for string length.
10390 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10393 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10394 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10396 2018-01-26 Richard Biener <rguenther@suse.de>
10398 PR rtl-optimization/84003
10399 * dse.c (record_store): Only record redundant stores when
10400 the earlier store aliases at least all accesses the later one does.
10402 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10404 PR rtl-optimization/83985
10405 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10406 REG_CFA_RESTORE insns.
10407 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10408 REG_CFA_RESTORE insns here.
10411 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10412 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10414 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10416 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10417 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10418 (arc_init): Likewise.
10419 (arc_override_options): Likewise.
10420 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10422 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10424 * config/arc/arc.h (TARGET_DBNZ): Define.
10425 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10426 properly set the tune attribute.
10427 (dbnz): Use TARGET_DBNZ guard.
10428 * config/arc/arc.opt (mtune): Add core3 option.
10430 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10432 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10433 recognize new pic like addresses.
10434 (arc_delegitimize_address): Clean up.
10436 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10438 * config/arc/arc-arches.def: Option mrf16 valid for all
10440 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10441 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10442 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10443 * config/arc/arc-tables.opt: Regenerate.
10444 * config/arc/arc.c (arc_conditional_register_usage): Handle
10445 reduced register file case.
10446 (arc_file_start): Set must have build attributes.
10447 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10448 mrf16 option value.
10449 * config/arc/arc.opt (mrf16): Add new option.
10450 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10451 * config/arc/genmultilib.awk: Handle new mrf16 option.
10452 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10453 * config/arc/t-multilib: Regenerate.
10454 * doc/invoke.texi (ARC Options): Document mrf16 option.
10456 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10458 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10459 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10460 (arc_attribute_table): Add 'secure_call' attribute.
10461 (arc_print_operand): Print secure call operand.
10462 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10464 (arc_is_secure_call_p): New function. * config/arc/arc.md
10465 (call_i): Add support for sjli instruction.
10466 (call_value_i): Likewise.
10467 * config/arc/constraints.md (Csc): New constraint.
10469 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10470 John Eric Martin <John.Martin@emmicro-us.com>
10472 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10473 * config/arc/arc.c (_arc_jli_section): New struct.
10474 (arc_jli_section): New type.
10475 (rc_jli_sections): New static variable.
10476 (arc_handle_jli_attribute): New function.
10477 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10478 (arc_file_end): New function.
10479 (TARGET_ASM_FILE_END): Define.
10480 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10481 (arc_add_jli_section): New function.
10482 (jli_call_scan): Likewise.
10483 (arc_reorg): Call jli_call_scan.
10484 (arc_output_addsi): Remove 'S' from printing asm operand.
10485 (arc_is_jli_call_p): New function.
10486 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10488 (movhi_insn): Likewise.
10489 (movsi_insn): Likewise.
10490 (movsi_set_cc_insn): Likewise.
10491 (loadqi_update): Likewise.
10492 (load_zeroextendqisi_update): Likewise.
10493 (load_signextendqisi_update): Likewise.
10494 (loadhi_update): Likewise.
10495 (load_zeroextendhisi_update): Likewise.
10496 (load_signextendhisi_update): Likewise.
10497 (loadsi_update): Likewise.
10498 (loadsf_update): Likewise.
10499 (movsicc_insn): Likewise.
10500 (bset_insn): Likewise.
10501 (bxor_insn): Likewise.
10502 (bclr_insn): Likewise.
10503 (bmsk_insn): Likewise.
10504 (bicsi3_insn): Likewise.
10505 (cmpsi_cc_c_insn): Likewise.
10506 (movsi_ne): Likewise.
10507 (movsi_cond_exec): Likewise.
10508 (clrsbsi2): Likewise.
10509 (norm_f): Likewise.
10519 (call_i): Remove 'S' asm letter, add jli instruction.
10520 (call_value_i): Likewise.
10521 * config/arc/arc.op (mjli-always): New option.
10522 * config/arc/constraints.md (Cji): New constraint.
10523 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10525 (subsf3_fpx): Likewise.
10526 (mulsf3_fpx): Likewise.
10527 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10529 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10530 function attrbutes.
10531 * doc/invoke.texi (ARC): Document mjli-always option.
10533 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10535 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10536 avoid addition with 0 and use incw and decw where possible.
10538 2018-01-26 Richard Biener <rguenther@suse.de>
10540 PR tree-optimization/81082
10541 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10542 association if it requires casting to unsigned.
10543 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10544 from fold_plusminus_mult_expr to catch important cases late when
10545 range info is available.
10547 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10549 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10550 * configure.ac (hidden_linkonce): New test.
10551 * configure: Regenerate.
10552 * config.in: Regenerate.
10554 2018-01-26 Julia Koval <julia.koval@intel.com>
10556 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10557 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10558 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10559 _mm_mask_bitshuffle_epi64_mask): Fix type.
10560 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10561 USI_FTYPE_V4DI_V4DI_USI): Remove.
10562 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10563 __builtin_ia32_vpshufbitqmb256_mask,
10564 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10565 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10566 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10568 2018-01-26 Alan Modra <amodra@gmail.com>
10571 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10572 UNSPEC_VBPERMQ. Sort other unspecs.
10574 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10576 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10578 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10580 PR middle-end/83055
10581 * predict.c (drop_profile): Do not push/pop cfun; update also
10583 (handle_missing_profiles): Fix logic looking for zero profiles.
10585 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10587 PR middle-end/83977
10588 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10589 on functions with #pragma omp declare simd or functions with simd
10591 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10592 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10593 Remove trailing \n from warning_at calls.
10595 2018-01-25 Tom de Vries <tom@codesourcery.com>
10598 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10599 for neutered workers.
10601 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10604 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10605 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10607 2018-01-24 Jeff Law <law@redhat.com>
10610 * i386.c (get_probe_interval): Move to earlier point.
10611 (ix86_compute_frame_layout): If -fstack-clash-protection and
10612 the frame is larger than the probe interval, then use pushes
10613 to save registers rather than reg->mem moves.
10614 (ix86_expand_prologue): Remove conditional for int_registers_saved
10617 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10620 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10621 min/max for never referenced object.
10623 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10625 PR middle-end/83977
10626 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10628 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10629 attributes from DECL_ATTRIBUTES (decl) without affecting
10630 DECL_ATTRIBUTES (current_function_decl).
10631 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10632 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10634 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10636 PR tree-optimization/83979
10637 * fold-const.c (fold_comparison): Use constant_boolean_node
10638 instead of boolean_{true,false}_node.
10640 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10642 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10645 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10647 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10648 Simplify the clause that sets the length attribute.
10649 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10650 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10651 clause that sets the length attribute.
10652 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10654 2018-01-24 Tom de Vries <tom@codesourcery.com>
10657 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10658 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10659 Add strict parameter.
10660 (prevent_branch_around_nothing): Insert dummy insn between branch to
10661 label and label with no ptx insn inbetween.
10662 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10664 2018-01-24 Tom de Vries <tom@codesourcery.com>
10667 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10668 for neutered threads in warp.
10669 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10671 2018-01-24 Richard Biener <rguenther@suse.de>
10673 PR tree-optimization/83176
10674 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10677 2018-01-24 Richard Biener <rguenther@suse.de>
10679 PR tree-optimization/82819
10680 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10681 code generating pluses that are no-ops in the target precision.
10683 2018-01-24 Richard Biener <rguenther@suse.de>
10685 PR middle-end/84000
10686 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10688 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10690 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10691 to merge probabilities.
10692 * predict.c (probably_never_executed): Also mark as cold functions
10693 with global 0 profile and guessed local profile.
10694 * profile-count.c (profile_probability::combine_with_count): New
10696 * profile-count.h (profile_probability::operator*,
10697 profile_probability::operator*=, profile_probability::operator/,
10698 profile_probability::operator/=): Reduce precision to adjusted
10699 and set value to guessed on contradictory divisions.
10700 (profile_probability::combine_with_freq): Remove.
10701 (profile_probability::combine_wiht_count): Declare.
10702 (profile_count::force_nonzero):: Set to adjusted.
10703 (profile_count::probability_in):: Set quality to adjusted.
10704 * tree-ssa-tail-merge.c (replace_block_by): Use
10705 combine_with_count.
10707 2018-01-23 Andrew Waterman <andrew@sifive.com>
10708 Jim Wilson <jimw@sifive.com>
10710 * config/riscv/riscv.c (riscv_stack_boundary): New.
10711 (riscv_option_override): Set riscv_stack_boundary. Handle
10712 riscv_preferred_stack_boundary_arg.
10713 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10714 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10715 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10716 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10717 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10718 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10720 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10723 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10724 of struct ix86_frame.
10725 (ix86_expand_epilogue): Likewise. Add a local variable for
10726 the reg_save_offset field in struct ix86_frame.
10728 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10730 PR tree-optimization/82604
10731 * tree-loop-distribution.c (enum partition_kind): New enum item
10732 PKIND_PARTIAL_MEMSET.
10733 (partition_builtin_p): Support above new enum item.
10734 (generate_code_for_partition): Ditto.
10735 (compute_access_range): Differentiate cases that equality can be
10736 proven at all loops, the innermost loops or no loops.
10737 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10738 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10739 (finalize_partitions, distribute_loop): Don't fuse partition of
10740 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10741 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10742 parloop is enabled.
10744 2018-01-23 Martin Liska <mliska@suse.cz>
10746 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10747 order to ignore the predictor.
10748 (PRED_POLYMORPHIC_CALL): Likewise.
10749 (PRED_RECURSIVE_CALL): Likewise.
10751 2018-01-23 Martin Liska <mliska@suse.cz>
10753 * tree-profile.c (tree_profiling): Print function header to
10754 aware reader which function we are working on.
10755 * value-prof.c (gimple_find_values_to_profile): Do not print
10756 not interesting value histograms.
10758 2018-01-23 Martin Liska <mliska@suse.cz>
10760 * profile-count.h (enum profile_quality): Add
10761 profile_uninitialized as the first value. Do not number values
10762 as they are zero based.
10763 (profile_count::verify): Update sanity check.
10764 (profile_probability::verify): Likewise.
10766 2018-01-23 Nathan Sidwell <nathan@acm.org>
10768 * doc/invoke.texi (ffor-scope): Deprecate.
10770 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10772 PR tree-optimization/83510
10773 * domwalk.c (set_all_edges_as_executable): New function.
10774 (dom_walker::dom_walker): Convert bool param
10775 "skip_unreachable_blocks" to enum reachability. Move setup of
10776 edge flags to set_all_edges_as_executable and only do it when
10777 reachability is REACHABLE_BLOCKS.
10778 * domwalk.h (enum dom_walker::reachability): New enum.
10779 (dom_walker::dom_walker): Convert bool param
10780 "skip_unreachable_blocks" to enum reachability.
10781 (set_all_edges_as_executable): New decl.
10782 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10783 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10785 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10786 but converting true to REACHABLE_BLOCKS.
10787 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10789 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10790 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10791 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10793 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10794 if check_all_array_refs will be called.
10796 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10798 * tree.c (selftest::test_location_wrappers): Add more test
10801 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10803 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10804 (selftest::test_bit_in_range): Likewise.
10806 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10809 * doc/sourcebuild.texi (vect_float): Say that the selector
10810 only describes the situation when -funsafe-math-optimizations is on.
10811 (vect_float_strict): Document.
10813 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10815 PR tree-optimization/83965
10816 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10817 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10818 instead of checking only for a reduction.
10819 (vect_recog_widen_sum_pattern): Likewise.
10821 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10823 * predict.c (probably_never_executed): Only use precise profile info.
10824 (compute_function_frequency): Skip after inlining hack since we now
10825 have quality checking.
10827 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10829 * profile-count.h (profile_probability::very_unlikely,
10830 profile_probability::unlikely, profile_probability::even): Set
10831 precision to guessed.
10833 2018-01-23 Richard Biener <rguenther@suse.de>
10835 PR tree-optimization/83963
10836 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10837 Properly terminate dominator walk when crossing the exit edge not
10838 when visiting its source block.
10840 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10843 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10844 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10846 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10848 PR tree-optimization/83957
10849 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10850 semicolon after for body surrounded by braces.
10852 PR tree-optimization/83081
10853 * profile-count.h (profile_probability::split): New method.
10854 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10855 Use profile_probability::split.
10856 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10857 when splitting a single conditional jump into 2.
10859 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10861 PR tree-optimization/69452
10862 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10865 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10867 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10868 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10869 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10871 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10873 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10875 * config/rl78/rl78.md (movdi): New define_expand.
10876 * config/rl78/rl78.c (rl78_split_movdi): New function.
10878 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10881 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10883 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10884 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10885 128-bit to produce an UNSPEC move to get the double word with the
10886 signbit and then a shift directly to do signbit.
10887 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10888 implementation with a new version that just does either a direct
10889 move or a regular move. Move memory interface to separate insns.
10890 Move insns so they are next to the expander.
10891 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10892 with signbit move. Split big and little endian case.
10893 (signbit<mode>2_dm_mem_le): Likewise.
10894 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10895 (signbit<mode>2_dm2): Likewise.
10897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10899 * config/rl78/rl78.md (anddi3): New define_expand.
10901 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10903 * config/rl78/rl78.md (umindi3): New define_expand.
10905 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10907 * config/rl78/rl78.md (smindi3): New define_expand.
10909 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10911 * config/rl78/rl78.md (smaxdi3): New define_expand.
10913 2018-01-22 Carl Love <cel@us.ibm.com>
10915 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10916 LVX_V1TI): Add macro expansion.
10917 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10918 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10919 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10920 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10921 Change check to determine if the instruction is a byte reversing
10922 entry. Fix typo in comment.
10923 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10924 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10925 Add def_builtin calls for new builtins.
10926 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10927 Add define_insn expansion.
10929 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10931 * config/rl78/rl78.md (umaxdi3): New define_expand.
10933 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10935 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10936 for non-QImode registers.
10938 2018-01-22 Richard Biener <rguenther@suse.de>
10940 PR tree-optimization/83963
10941 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10942 including the loop exit block.
10943 (scop_detection::merge_sese): Likewise.
10944 (scop_detection::add_scop): Do it here instead.
10946 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10948 * doc/sourcebuild.texi (arm_softfloat): Document.
10950 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10953 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10954 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10955 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10957 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10958 David Edelsohn <dje.gcc@gmail.com>
10961 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10962 Change "crset eq" to "crset 2".
10963 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10964 (*call_indirect_aix<mode>_nospec): Likewise.
10965 (*call_value_indirect_aix<mode>_nospec): Likewise.
10966 (*call_indirect_elfv2<mode>_nospec): Likewise.
10967 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10968 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10969 change assembly output from . to $.
10970 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10971 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10972 (*tablejump<mode>_internal1_nospec): Likewise.
10974 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10977 * config/sh/sh_optimize_sett_clrt.cc:
10978 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10980 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10982 PR tree-optimization/83940
10983 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10984 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10985 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10986 use it to pass back the definition type.
10987 (vect_check_store_rhs): Likewise rhs_dt_out.
10988 (vect_build_gather_load_calls): Add a mask_dt argument and use
10989 it instead of a call to vect_is_simple_use.
10990 (vectorizable_store): Update calls to vect_check_load_store_mask
10991 and vect_check_store_rhs. Use the dt returned by the latter instead
10992 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10993 instead of calls to vect_is_simple_use. Pass the scalar rather
10994 than the vector operand to vect_is_simple_use when handling
10995 second and subsequent copies of an rhs value.
10996 (vectorizable_load): Update calls to vect_check_load_store_mask
10997 and vect_build_gather_load_calls. Use the cached mask_dt and
10998 gs_info.offset_dt instead of calls to vect_is_simple_use.
11000 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11002 PR middle-end/83945
11003 * tree-emutls.c: Include gimplify.h.
11004 (lower_emutls_2): New function.
11005 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11006 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11007 it before further processing.
11010 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11011 UINTVAL (trueop1) instead of INTVAL (op1).
11013 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11017 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11018 INCOMING_FRAME_SP_OFFSET if not defined.
11019 (scan_trace): Add ENTRY argument. If true and
11020 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11021 emit a note to adjust the CFA offset.
11022 (create_cfi_notes): Adjust scan_trace callers.
11023 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11024 INCOMING_FRAME_SP_OFFSET in the CIE.
11025 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11026 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11028 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11029 * doc/tm.texi: Regenerated.
11031 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11033 PR rtl-optimization/83147
11034 * lra-constraints.c (remove_inheritance_pseudos): Use
11035 lra_substitute_pseudo_within_insn.
11037 2018-01-19 Tom de Vries <tom@codesourcery.com>
11038 Cesar Philippidis <cesar@codesourcery.com>
11041 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11043 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11046 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11047 spaces for function labels.
11049 2018-01-19 Martin Liska <mliska@suse.cz>
11051 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11052 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11053 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11054 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11055 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11056 (PRED_CONST_RETURN): Change from 69 to 65.
11057 (PRED_NULL_RETURN): Change from 91 to 71.
11058 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11059 (PRED_LOOP_GUARD): Change from 66 to 73.
11061 2018-01-19 Martin Liska <mliska@suse.cz>
11063 * predict.c (predict_insn_def): Add new assert.
11064 (struct branch_predictor): Change type to signed integer.
11065 (test_prediction_value_range): Amend test to cover
11066 PROB_UNINITIALIZED.
11067 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11068 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11069 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11070 (PRED_LOOP_IV_COMPARE): Likewise.
11071 * predict.h (PROB_UNINITIALIZED): Define new constant.
11073 2018-01-19 Martin Liska <mliska@suse.cz>
11075 * predict.c (dump_prediction): Add new format for
11076 analyze_brprob.py script which is enabled with -details
11078 * profile-count.h (precise_p): New function.
11080 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11082 PR tree-optimization/83922
11083 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11084 there are no statements that need masking.
11085 (vect_active_double_reduction_p): New function.
11086 (vect_analyze_loop_operations): Use it when handling phis that
11087 are not in the loop header.
11089 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11091 PR tree-optimization/83914
11092 * tree-vect-loop.c (vectorizable_induction): Don't convert
11093 init_expr or apply the peeling adjustment for inductions
11094 that are nested within the vectorized loop.
11096 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11098 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11101 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11105 * function.h (gimplify_parameters): Add gimple_seq * argument.
11106 * function.c: Include gimple.h and options.h.
11107 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11108 for the added local temporaries if needed.
11109 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11110 if there are any parameter cleanups, wrap whole body into a
11111 try/finally with the cleanups.
11113 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11116 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11117 Use GET_MODE_CLASS for scalar floating point.
11119 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11123 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11124 Fix call of call_cgraph_insertion_hooks.
11126 2018-01-18 Martin Sebor <msebor@redhat.com>
11128 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11130 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11133 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11136 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11139 * common.opt: (-ffile-prefix-map): New option.
11140 * opts.c (common_handle_option): Defer it.
11141 * opts-global.c (handle_common_deferred_options): Handle it.
11142 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11143 * file-prefix-map.h: New file.
11144 (remap_debug_filename, add_debug_prefix_map): ...here.
11145 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11146 * final.c (debug_prefix_map, add_debug_prefix_map
11147 remap_debug_filename): Move to...
11148 * file-prefix-map.c: New file.
11149 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11150 generalize, get rid of alloca(), use strrchr() instead of strchr().
11151 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11152 Implement in terms of add_prefix_map().
11153 (remap_macro_filename, remap_debug_filename): Implement in term of
11155 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11156 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11157 * dbxout.c: Include file-prefix-map.h.
11158 * varasm.c: Likewise.
11159 * vmsdbgout.c: Likewise.
11160 * xcoffout.c: Likewise.
11161 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11162 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11163 * doc/invoke.texi (-ffile-prefix-map): Document.
11164 (-fdebug-prefix-map): Update description.
11166 2018-01-18 Martin Liska <mliska@suse.cz>
11168 * config/i386/i386.c (indirect_thunk_name): Document that also
11170 (output_indirect_thunk): Document why both instructions
11171 (pause and lfence) are generated.
11173 2018-01-18 Richard Biener <rguenther@suse.de>
11175 PR tree-optimization/83887
11176 * graphite-scop-detection.c
11177 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11178 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11179 (scop_detection::merge_sese): Re-implement with a flood-fill
11180 algorithm that properly finds a SESE region if it exists.
11182 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11185 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11186 pointer_diff optimizations use view_convert instead of convert.
11188 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11190 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11191 Generate different code for -mno-speculate-indirect-jumps.
11192 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11193 (*call_indirect_aix<mode>): Disable for
11194 -mno-speculate-indirect-jumps.
11195 (*call_indirect_aix<mode>_nospec): New define_insn.
11196 (*call_value_indirect_aix<mode>): Disable for
11197 -mno-speculate-indirect-jumps.
11198 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11199 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11200 -mno-speculate-indirect-jumps.
11201 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11203 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11205 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11206 long double type, set the flags for noting the default long double
11207 type, even if we don't pass or return a long double type.
11209 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11212 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11215 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11217 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11218 support for merge[hl].
11219 (fold_mergehl_helper): New helper function.
11220 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11221 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11222 (altivec_vmrglw_direct): Add xxmrglw insn.
11224 2018-01-17 Andrew Waterman <andrew@sifive.com>
11226 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11227 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11229 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11232 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11233 call the lto_location_cache before reading the
11234 DECL_SOURCE_LOCATION of the types.
11236 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11237 Richard Sandiford <richard.sandiford@linaro.org>
11239 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11240 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11241 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11242 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11243 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11245 * config/aarch64/constraints.md (aarch64_movti_operand):
11247 * config/aarch64/predicates.md (Uti): Add new constraint.
11249 2018-01-17 Carl Love <cel@us.ibm.com>
11251 * config/rs6000/vsx.md (define_expand xl_len_r,
11252 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11253 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11255 (define_expand, define_insn): Move the shift left from the
11256 define_insn to the define_expand for lxvl and stxvl instructions.
11257 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11258 and XL_LEN_R definitions to PURE.
11260 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11262 * config/i386/i386.c (indirect_thunk_name): Declare regno
11263 as unsigned int. Compare regno with INVALID_REGNUM.
11264 (output_indirect_thunk): Ditto.
11265 (output_indirect_thunk_function): Ditto.
11266 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11267 in the call to output_indirect_thunk_function.
11269 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11271 PR middle-end/83884
11272 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11273 rather than the size of inner_type to determine the stack slot size
11274 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11276 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11279 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11282 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11284 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11285 endian Linux systems to optionally enable multilibs for selecting
11286 the long double type if the user configured an explicit type.
11287 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11288 have no long double multilibs if not defined.
11289 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11290 warn if the user used -mabi={ieee,ibm}longdouble and we built
11291 multilibs for long double.
11292 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11293 appropriate multilib option.
11294 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11296 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11297 for building long double multilibs.
11298 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11300 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11302 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11305 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11307 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11310 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11313 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11316 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11318 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11319 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11321 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11323 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11324 different rtl trees depending on TARGET_64BIT.
11325 (rs6000_gen_lvx): Likewise.
11327 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11329 * config/visium/visium.md (nop): Tweak comment.
11330 (hazard_nop): Likewise.
11332 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11334 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11335 -mspeculate-indirect-jumps.
11336 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11337 for -mno-speculate-indirect-jumps.
11338 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11339 (*call_value_indirect_elfv2<mode>): Disable for
11340 -mno-speculate-indirect-jumps.
11341 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11342 (indirect_jump): Emit different RTL for
11343 -mno-speculate-indirect-jumps.
11344 (*indirect_jump<mode>): Disable for
11345 -mno-speculate-indirect-jumps.
11346 (*indirect_jump<mode>_nospec): New define_insn.
11347 (tablejump): Emit different RTL for
11348 -mno-speculate-indirect-jumps.
11349 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11350 (tablejumpsi_nospec): New define_expand.
11351 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11352 (tablejumpdi_nospec): New define_expand.
11353 (*tablejump<mode>_internal1): Disable for
11354 -mno-speculate-indirect-jumps.
11355 (*tablejump<mode>_internal1_nospec): New define_insn.
11356 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11359 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11361 * caller-save.c (insert_save): Drop unnecessary parameter. All
11364 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11365 Richard Biener <rguenth@suse.de>
11368 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11369 return early, inline manually is_gimple_sizepos. Make sure if we
11370 call gimplify_expr we don't end up with a gimple constant.
11371 * tree.c (variably_modified_type_p): Don't return true for
11372 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11373 * gimplify.h (is_gimple_sizepos): Remove.
11375 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11377 PR tree-optimization/83857
11378 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11379 vectorizable_live_operation for pure SLP statements.
11380 (vectorizable_live_operation): Handle PHIs.
11382 2018-01-16 Richard Biener <rguenther@suse.de>
11384 PR tree-optimization/83867
11385 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11386 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11388 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11391 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11392 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11393 If off is not INTEGER_CST, issue a may not be aligned warning
11394 rather than isn't aligned. Use isn%'t rather than isn't.
11395 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11397 <case MULT_EXPR>: Improve the case when bottom and one of the
11398 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11399 operand, in that case check if the other operand is multiple of
11400 bottom divided by the INTEGER_CST operand.
11402 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11405 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11406 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11407 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11408 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11409 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11410 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11411 (pa_function_arg_size): New function.
11413 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11415 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11416 in a separate statement.
11418 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11420 PR tree-optimization/83847
11421 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11422 group gathers and scatters.
11424 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11426 PR rtl-optimization/86620
11427 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11429 PR rtl-optimization/83213
11430 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11431 to last if both are JUMP_INSNs.
11433 PR tree-optimization/83843
11434 * gimple-ssa-store-merging.c
11435 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11436 store_immediate_info for bswap/nop orig_stores.
11438 2018-01-15 Andrew Waterman <andrew@sifive.com>
11440 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11442 <UDIV>: Increase cost if !TARGET_DIV.
11444 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11446 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11447 (define_attr "cr_logical_3op"): New.
11448 (cceq_ior_compare): Adjust.
11449 (cceq_ior_compare_complement): Adjust.
11450 (*cceq_rev_compare): Adjust.
11451 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11452 (is_cracked_insn): Adjust.
11453 (insn_must_be_first_in_group): Adjust.
11454 * config/rs6000/40x.md: Adjust.
11455 * config/rs6000/440.md: Adjust.
11456 * config/rs6000/476.md: Adjust.
11457 * config/rs6000/601.md: Adjust.
11458 * config/rs6000/603.md: Adjust.
11459 * config/rs6000/6xx.md: Adjust.
11460 * config/rs6000/7450.md: Adjust.
11461 * config/rs6000/7xx.md: Adjust.
11462 * config/rs6000/8540.md: Adjust.
11463 * config/rs6000/cell.md: Adjust.
11464 * config/rs6000/e300c2c3.md: Adjust.
11465 * config/rs6000/e500mc.md: Adjust.
11466 * config/rs6000/e500mc64.md: Adjust.
11467 * config/rs6000/e5500.md: Adjust.
11468 * config/rs6000/e6500.md: Adjust.
11469 * config/rs6000/mpc.md: Adjust.
11470 * config/rs6000/power4.md: Adjust.
11471 * config/rs6000/power5.md: Adjust.
11472 * config/rs6000/power6.md: Adjust.
11473 * config/rs6000/power7.md: Adjust.
11474 * config/rs6000/power8.md: Adjust.
11475 * config/rs6000/power9.md: Adjust.
11476 * config/rs6000/rs64.md: Adjust.
11477 * config/rs6000/titan.md: Adjust.
11479 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11481 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11482 ix86_indirect_branch_register logic.
11484 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11486 * config/i386/constraints.md (Bs): Update
11487 ix86_indirect_branch_register check. Don't check
11488 ix86_indirect_branch_register with GOT_memory_operand.
11490 * config/i386/predicates.md (GOT_memory_operand): Don't check
11491 ix86_indirect_branch_register here.
11492 (GOT32_symbol_operand): Likewise.
11494 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11496 * config/i386/predicates.md (constant_call_address_operand):
11497 Rewrite ix86_indirect_branch_register logic.
11498 (sibcall_insn_operand): Likewise.
11500 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11502 * config/i386/constraints.md (Bs): Replace
11503 ix86_indirect_branch_thunk_register with
11504 ix86_indirect_branch_register.
11506 * config/i386/i386.md (indirect_jump): Likewise.
11507 (tablejump): Likewise.
11508 (*sibcall_memory): Likewise.
11509 (*sibcall_value_memory): Likewise.
11510 Peepholes of indirect call and jump via memory: Likewise.
11511 * config/i386/i386.opt: Likewise.
11512 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11513 (GOT_memory_operand): Likewise.
11514 (call_insn_operand): Likewise.
11515 (sibcall_insn_operand): Likewise.
11516 (GOT32_symbol_operand): Likewise.
11518 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11520 PR middle-end/83837
11521 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11522 type rather than type addr's type points to.
11523 (expand_omp_atomic_mutex): Likewise.
11524 (expand_omp_atomic): Likewise.
11526 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11529 * config/i386/i386.c (output_indirect_thunk_function): Use
11530 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11531 for __x86_return_thunk.
11533 2018-01-15 Richard Biener <rguenther@suse.de>
11535 PR middle-end/83850
11536 * expmed.c (extract_bit_field_1): Fix typo.
11538 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11541 * config/arm/iterators.md (VF): New mode iterator.
11542 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11543 Remove integer-related logic from pattern.
11544 (neon_vabd<mode>_3): Likewise.
11546 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11548 PR middle-end/82694
11549 * common.opt (fstrict-overflow): No longer an alias.
11550 (fwrapv-pointer): New option.
11551 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11552 also for pointer types based on flag_wrapv_pointer.
11553 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11554 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11555 opts->x_flag_wrapv got set.
11556 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11557 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11558 POINTER_TYPE_OVERFLOW_UNDEFINED.
11559 * match.pd: Likewise in address comparison pattern.
11560 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11562 2018-01-15 Richard Biener <rguenther@suse.de>
11565 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11566 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11567 Reset type names to their identifier if their TYPE_DECL doesn't
11568 have linkage (and thus is used for ODR and devirt).
11569 (save_debug_info_for_decl): Remove.
11570 (save_debug_info_for_type): Likewise.
11571 (add_tree_to_fld_list): Adjust.
11572 * tree-pretty-print.c (dump_generic_node): Make dumping of
11573 type names more robust.
11575 2018-01-15 Richard Biener <rguenther@suse.de>
11577 * BASE-VER: Bump to 8.0.1.
11579 2018-01-14 Martin Sebor <msebor@redhat.com>
11582 * builtins.c (check_access): Avoid warning when the no-warning bit
11585 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11587 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11588 * ira-color (allocno_hard_regs_compare): Likewise.
11590 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11593 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11594 Use .pushsection/.popsection.
11596 2018-01-14 Martin Sebor <msebor@redhat.com>
11599 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11601 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11603 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11604 entry from extra_headers.
11605 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11606 extra_headers, make the list bitwise identical to the i?86-*-* one.
11608 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11610 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11611 -mcmodel=large with -mindirect-branch=thunk,
11612 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11613 -mfunction-return=thunk-extern.
11614 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11615 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11616 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11618 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11620 * config/i386/i386.c (print_reg): Print the name of the full
11621 integer register without '%'.
11622 (ix86_print_operand): Handle 'V'.
11623 * doc/extend.texi: Document 'V' modifier.
11625 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11627 * config/i386/constraints.md (Bs): Disallow memory operand for
11628 -mindirect-branch-register.
11630 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11631 (GOT_memory_operand): Likewise.
11632 (call_insn_operand): Likewise.
11633 (sibcall_insn_operand): Likewise.
11634 (GOT32_symbol_operand): Likewise.
11635 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11636 for -mindirect-branch-register.
11637 (tablejump): Likewise.
11638 (*sibcall_memory): Likewise.
11639 (*sibcall_value_memory): Likewise.
11640 Disallow peepholes of indirect call and jump via memory for
11641 -mindirect-branch-register.
11642 (*call_pop): Replace m with Bw.
11643 (*call_value_pop): Likewise.
11644 (*sibcall_pop_memory): Replace m with Bs.
11645 * config/i386/i386.opt (mindirect-branch-register): New option.
11646 * doc/invoke.texi: Document -mindirect-branch-register option.
11648 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11650 * config/i386/i386-protos.h (ix86_output_function_return): New.
11651 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11652 set function_return_type.
11653 (indirect_thunk_name): Add ret_p to indicate thunk for function
11655 (output_indirect_thunk_function): Pass false to
11656 indirect_thunk_name.
11657 (ix86_output_indirect_branch_via_reg): Likewise.
11658 (ix86_output_indirect_branch_via_push): Likewise.
11659 (output_indirect_thunk_function): Create alias for function
11660 return thunk if regno < 0.
11661 (ix86_output_function_return): New function.
11662 (ix86_handle_fndecl_attribute): Handle function_return.
11663 (ix86_attribute_table): Add function_return.
11664 * config/i386/i386.h (machine_function): Add
11665 function_return_type.
11666 * config/i386/i386.md (simple_return_internal): Use
11667 ix86_output_function_return.
11668 (simple_return_internal_long): Likewise.
11669 * config/i386/i386.opt (mfunction-return=): New option.
11670 (indirect_branch): Mention -mfunction-return=.
11671 * doc/extend.texi: Document function_return function attribute.
11672 * doc/invoke.texi: Document -mfunction-return= option.
11674 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11676 * config/i386/i386-opts.h (indirect_branch): New.
11677 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11678 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11679 with local indirect jump when converting indirect call and jump.
11680 (ix86_set_indirect_branch_type): New.
11681 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11682 (indirectlabelno): New.
11683 (indirect_thunk_needed): Likewise.
11684 (indirect_thunk_bnd_needed): Likewise.
11685 (indirect_thunks_used): Likewise.
11686 (indirect_thunks_bnd_used): Likewise.
11687 (INDIRECT_LABEL): Likewise.
11688 (indirect_thunk_name): Likewise.
11689 (output_indirect_thunk): Likewise.
11690 (output_indirect_thunk_function): Likewise.
11691 (ix86_output_indirect_branch_via_reg): Likewise.
11692 (ix86_output_indirect_branch_via_push): Likewise.
11693 (ix86_output_indirect_branch): Likewise.
11694 (ix86_output_indirect_jmp): Likewise.
11695 (ix86_code_end): Call output_indirect_thunk_function if needed.
11696 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11698 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11699 (ix86_attribute_table): Add indirect_branch.
11700 * config/i386/i386.h (machine_function): Add indirect_branch_type
11701 and has_local_indirect_jump.
11702 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11704 (tablejump): Likewise.
11705 (*indirect_jump): Use ix86_output_indirect_jmp.
11706 (*tablejump_1): Likewise.
11707 (simple_return_indirect_internal): Likewise.
11708 * config/i386/i386.opt (mindirect-branch=): New option.
11709 (indirect_branch): New.
11712 (thunk-inline): Likewise.
11713 (thunk-extern): Likewise.
11714 * doc/extend.texi: Document indirect_branch function attribute.
11715 * doc/invoke.texi: Document -mindirect-branch= option.
11717 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11720 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11722 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11724 * ipa-inline.c (want_inline_small_function_p): Return false if
11725 inlining has already failed with CIF_FINAL_ERROR.
11726 (update_caller_keys): Call want_inline_small_function_p before
11728 (update_callee_keys): Likewise.
11730 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11732 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11734 (rs6000_quadword_masked_address_p): Likewise.
11735 (quad_aligned_load_p): Likewise.
11736 (quad_aligned_store_p): Likewise.
11737 (const_load_sequence_p): Add comment to describe the outer-most loop.
11738 (mimic_memory_attributes_and_flags): New function.
11739 (rs6000_gen_stvx): Likewise.
11740 (replace_swapped_aligned_store): Likewise.
11741 (rs6000_gen_lvx): Likewise.
11742 (replace_swapped_aligned_load): Likewise.
11743 (replace_swapped_load_constant): Capitalize argument name in
11744 comment describing this function.
11745 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11746 and stores that access quad-word aligned addresses and replace
11747 with stvx or lvx instructions when appropriate.
11748 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11749 New function prototype.
11750 (rs6000_quadword_masked_address_p): Likewise.
11751 (rs6000_gen_lvx): Likewise.
11752 (rs6000_gen_stvx): Likewise.
11753 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11754 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11755 when memory address is aligned.
11756 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11757 this split to select lvx instruction when memory address is aligned.
11758 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11759 instruction when memory address is aligned.
11760 (*vsx_le_perm_load_v16qi): Likewise.
11761 (four unnamed splitters): Modify to select the stvx instruction
11762 when memory is aligned.
11764 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11766 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11767 which appears in the queue multiple times.
11769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11770 Alan Hayward <alan.hayward@arm.com>
11771 David Sherwood <david.sherwood@arm.com>
11773 * tree-vectorizer.h (vec_lower_bound): New structure.
11774 (_loop_vec_info): Add check_nonzero and lower_bounds.
11775 (LOOP_VINFO_CHECK_NONZERO): New macro.
11776 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11777 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11778 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11779 fields. Make seg_len the distance travelled, not including the
11781 (dr_direction_indicator): Declare.
11782 (dr_zero_step_indicator): Likewise.
11783 (dr_known_forward_stride_p): Likewise.
11784 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11786 (runtime_alias_check_p): Allow runtime alias checks with
11788 (operator ==): Compare access_size and align.
11789 (prune_runtime_alias_test_list): Rework for new distinction between
11790 the access_size and seg_len.
11791 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11793 (get_segment_min_max): New function.
11794 (create_intersect_range_checks): Use it.
11795 (dr_step_indicator): New function.
11796 (dr_direction_indicator): Likewise.
11797 (dr_zero_step_indicator): Likewise.
11798 (dr_known_forward_stride_p): Likewise.
11799 * tree-loop-distribution.c (data_ref_segment_size): Return
11800 DR_STEP * (niters - 1).
11801 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11803 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11804 (vect_preserves_scalar_order_p): New function, split out from...
11805 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11806 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11807 (vect_vfa_access_size): New function.
11808 (vect_vfa_align): Likewise.
11809 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11810 (dump_lower_bound): New function.
11811 (vect_check_lower_bound): Likewise.
11812 (vect_small_gap_p): Likewise.
11813 (vectorizable_with_step_bound_p): Likewise.
11814 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11815 depencies if the vectorization factor is 1. Convert the checks
11816 for nonzero steps into checks on the bounds of DR_STEP. Try using
11817 a bunds check for variable steps if the minimum required step is
11818 relatively small. Update calls to the dr_with_seg_len
11819 constructor and to vect_compile_time_alias.
11820 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11822 (vect_loop_versioning): Call it.
11823 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11825 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11828 Alan Hayward <alan.hayward@arm.com>
11829 David Sherwood <david.sherwood@arm.com>
11831 * doc/sourcebuild.texi (vect_scatter_store): Document.
11832 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11834 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11836 * genopinit.c (main): Add supports_vec_scatter_store and
11837 supports_vec_scatter_store_cached to target_optabs.
11838 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11839 IFN_MASK_SCATTER_STORE.
11840 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11842 * internal-fn.h (internal_store_fn_p): Declare.
11843 (internal_fn_stored_value_index): Likewise.
11844 * internal-fn.c (scatter_store_direct): New macro.
11845 (expand_scatter_store_optab_fn): New function.
11846 (direct_scatter_store_optab_supported_p): New macro.
11847 (internal_store_fn_p): New function.
11848 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11849 IFN_MASK_SCATTER_STORE.
11850 (internal_fn_mask_index): Likewise.
11851 (internal_fn_stored_value_index): New function.
11852 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11853 for scatter stores.
11854 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11855 * optabs-query.c (supports_vec_scatter_store_p): New function.
11856 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11857 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11858 true for scatter stores.
11859 (vect_gather_scatter_fn_p): Handle scatter stores too.
11860 (vect_check_gather_scatter): Consider using scatter stores if
11861 supports_vec_scatter_store_p.
11862 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11863 scatter stores too.
11864 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11865 internal_fn_stored_value_index.
11866 (check_load_store_masking): Handle scatter stores too.
11867 (vect_get_store_rhs): Make public.
11868 (vectorizable_call): Use internal_store_fn_p.
11869 (vectorizable_store): Handle scatter store internal functions.
11870 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11871 when deciding whether the end of the group has been reached.
11872 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11873 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11874 (mask_scatter_store<mode>): New insns.
11876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11877 Alan Hayward <alan.hayward@arm.com>
11878 David Sherwood <david.sherwood@arm.com>
11880 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11881 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11882 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11884 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11885 Use vect_truncate_gather_scatter_offset if we can't treat the
11886 operation as a normal gather load or scatter store.
11887 (get_group_load_store_type): Take the gather_scatter_info
11888 as argument. Try using a gather load or scatter store for
11889 single-element groups.
11890 (get_load_store_type): Update calls to get_group_load_store_type
11891 and vect_use_strided_gather_scatters_p.
11893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11894 Alan Hayward <alan.hayward@arm.com>
11895 David Sherwood <david.sherwood@arm.com>
11897 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11898 optional tree argument.
11899 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11901 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11902 but continue to use the current value as a fallback.
11903 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11904 to compare the updates.
11905 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11906 (get_load_store_type): Use it when handling a strided access.
11907 (vect_get_strided_load_store_ops): New function.
11908 (vect_get_data_ptr_increment): Likewise.
11909 (vectorizable_load): Handle strided gather loads. Always pass
11910 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11913 Alan Hayward <alan.hayward@arm.com>
11914 David Sherwood <david.sherwood@arm.com>
11916 * doc/md.texi (gather_load@var{m}): Document.
11917 (mask_gather_load@var{m}): Likewise.
11918 * genopinit.c (main): Add supports_vec_gather_load and
11919 supports_vec_gather_load_cached to target_optabs.
11920 * optabs-tree.c (init_tree_optimization_optabs): Use
11921 ggc_cleared_alloc to allocate target_optabs.
11922 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11923 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11925 * internal-fn.h (internal_load_fn_p): Declare.
11926 (internal_gather_scatter_fn_p): Likewise.
11927 (internal_fn_mask_index): Likewise.
11928 (internal_gather_scatter_fn_supported_p): Likewise.
11929 * internal-fn.c (gather_load_direct): New macro.
11930 (expand_gather_load_optab_fn): New function.
11931 (direct_gather_load_optab_supported_p): New macro.
11932 (direct_internal_fn_optab): New function.
11933 (internal_load_fn_p): Likewise.
11934 (internal_gather_scatter_fn_p): Likewise.
11935 (internal_fn_mask_index): Likewise.
11936 (internal_gather_scatter_fn_supported_p): Likewise.
11937 * optabs-query.c (supports_at_least_one_mode_p): New function.
11938 (supports_vec_gather_load_p): Likewise.
11939 * optabs-query.h (supports_vec_gather_load_p): Declare.
11940 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11941 and memory_type field.
11942 (NUM_PATTERNS): Bump to 15.
11943 * tree-vect-data-refs.c: Include internal-fn.h.
11944 (vect_gather_scatter_fn_p): New function.
11945 (vect_describe_gather_scatter_call): Likewise.
11946 (vect_check_gather_scatter): Try using internal functions for
11947 gather loads. Recognize existing calls to a gather load function.
11948 (vect_analyze_data_refs): Consider using gather loads if
11949 supports_vec_gather_load_p.
11950 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11951 (vect_get_gather_scatter_offset_type): Likewise.
11952 (vect_convert_mask_for_vectype): Likewise.
11953 (vect_add_conversion_to_patterm): Likewise.
11954 (vect_try_gather_scatter_pattern): Likewise.
11955 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11956 (vect_vect_recog_func_ptrs): Add it.
11957 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11958 internal_fn_mask_index and internal_gather_scatter_fn_p.
11959 (check_load_store_masking): Take the gather_scatter_info as an
11960 argument and handle gather loads.
11961 (vect_get_gather_scatter_ops): New function.
11962 (vectorizable_call): Check internal_load_fn_p.
11963 (vectorizable_load): Likewise. Handle gather load internal
11965 (vectorizable_store): Update call to check_load_store_masking.
11966 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11967 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11968 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11969 (aarch64_gather_scale_operand_d): New predicates.
11970 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11971 (mask_gather_load<mode>): New insns.
11973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11974 Alan Hayward <alan.hayward@arm.com>
11975 David Sherwood <david.sherwood@arm.com>
11977 * optabs.def (fold_left_plus_optab): New optab.
11978 * doc/md.texi (fold_left_plus_@var{m}): Document.
11979 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11980 * internal-fn.c (fold_left_direct): Define.
11981 (expand_fold_left_optab_fn): Likewise.
11982 (direct_fold_left_optab_supported_p): Likewise.
11983 * fold-const-call.c (fold_const_fold_left): New function.
11984 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11985 * tree-parloops.c (valid_reduction_p): New function.
11986 (gather_scalar_reductions): Use it.
11987 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11988 (vect_finish_replace_stmt): Declare.
11989 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11990 (needs_fold_left_reduction_p): New function, split out from...
11991 (vect_is_simple_reduction): ...here. Accept reductions that
11992 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11993 (vect_force_simple_reduction): Also store the reduction type in
11994 the assignment's STMT_VINFO_REDUC_TYPE.
11995 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11996 (merge_with_identity): New function.
11997 (vect_expand_fold_left): Likewise.
11998 (vectorize_fold_left_reduction): Likewise.
11999 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12000 scalar phi in place for it. Check for target support and reject
12001 cases that would reassociate the operation. Defer the transform
12002 phase to vectorize_fold_left_reduction.
12003 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12004 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12005 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12009 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12010 call to ifc_temp_var.
12012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12013 Alan Hayward <alan.hayward@arm.com>
12014 David Sherwood <david.sherwood@arm.com>
12016 * target.def (legitimize_address_displacement): Take the original
12017 offset as a poly_int.
12018 * targhooks.h (default_legitimize_address_displacement): Update
12020 * targhooks.c (default_legitimize_address_displacement): Likewise.
12021 * doc/tm.texi: Regenerate.
12022 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12023 as an argument, moving assert of ad->disp == ad->disp_term to...
12024 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12025 Try calling targetm.legitimize_address_displacement before expanding
12026 the address rather than afterwards, and adjust for the new interface.
12027 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12028 Match the new hook interface. Handle SVE addresses.
12029 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12030 new hook interface.
12032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12034 * Makefile.in (OBJS): Add early-remat.o.
12035 * target.def (select_early_remat_modes): New hook.
12036 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12037 * doc/tm.texi: Regenerate.
12038 * targhooks.h (default_select_early_remat_modes): Declare.
12039 * targhooks.c (default_select_early_remat_modes): New function.
12040 * timevar.def (TV_EARLY_REMAT): New timevar.
12041 * passes.def (pass_early_remat): New pass.
12042 * tree-pass.h (make_pass_early_remat): Declare.
12043 * early-remat.c: New file.
12044 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12046 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12049 Alan Hayward <alan.hayward@arm.com>
12050 David Sherwood <david.sherwood@arm.com>
12052 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12053 vfm1 with a bound_epilog parameter.
12054 (vect_do_peeling): Update calls accordingly, and move the prologue
12055 call earlier in the function. Treat the base bound_epilog as 0 for
12056 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12057 this base when peeling for gaps.
12058 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12059 with fully-masked loops.
12060 (vect_estimate_min_profitable_iters): Handle the single peeled
12061 iteration in that case.
12063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12064 Alan Hayward <alan.hayward@arm.com>
12065 David Sherwood <david.sherwood@arm.com>
12067 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12068 single-element interleaving even if the size is not a power of 2.
12069 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12070 accesses for single-element interleaving if the group size is
12073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12074 Alan Hayward <alan.hayward@arm.com>
12075 David Sherwood <david.sherwood@arm.com>
12077 * doc/md.texi (fold_extract_last_@var{m}): Document.
12078 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12079 * optabs.def (fold_extract_last_optab): New optab.
12080 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12081 * internal-fn.c (fold_extract_direct): New macro.
12082 (expand_fold_extract_optab_fn): Likewise.
12083 (direct_fold_extract_optab_supported_p): Likewise.
12084 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12085 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12086 EXTRACT_LAST_REDUCTION.
12087 (get_initial_def_for_reduction): Do not create an initial vector
12088 for EXTRACT_LAST_REDUCTION reductions.
12089 (vectorizable_reduction): Leave the scalar phi in place for
12090 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12091 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12092 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12093 transform phase to vectorizable_condition.
12094 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12096 (vect_finish_stmt_generation): ...here.
12097 (vect_finish_replace_stmt): New function.
12098 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12099 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12101 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12104 Alan Hayward <alan.hayward@arm.com>
12105 David Sherwood <david.sherwood@arm.com>
12107 * doc/md.texi (extract_last_@var{m}): Document.
12108 * optabs.def (extract_last_optab): New optab.
12109 * internal-fn.def (EXTRACT_LAST): New internal function.
12110 * internal-fn.c (cond_unary_direct): New macro.
12111 (expand_cond_unary_optab_fn): Likewise.
12112 (direct_cond_unary_optab_supported_p): Likewise.
12113 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12114 loops using EXTRACT_LAST.
12115 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12116 (extract_last_<mode>): ...this optab.
12117 (vec_extract<mode><Vel>): Update accordingly.
12119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12120 Alan Hayward <alan.hayward@arm.com>
12121 David Sherwood <david.sherwood@arm.com>
12123 * target.def (empty_mask_is_expensive): New hook.
12124 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12125 * doc/tm.texi: Regenerate.
12126 * targhooks.h (default_empty_mask_is_expensive): Declare.
12127 * targhooks.c (default_empty_mask_is_expensive): New function.
12128 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12129 if the target says that empty masks are expensive.
12130 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12132 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12135 Alan Hayward <alan.hayward@arm.com>
12136 David Sherwood <david.sherwood@arm.com>
12138 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12139 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12140 (vect_use_loop_mask_for_alignment_p): New function.
12141 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12142 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12143 niters_skip argument. Make sure that the first niters_skip elements
12144 of the first iteration are inactive.
12145 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12146 Update call to vect_set_loop_masks_directly.
12147 (get_misalign_in_elems): New function, split out from...
12148 (vect_gen_prolog_loop_niters): ...here.
12149 (vect_update_init_of_dr): Take a code argument that specifies whether
12150 the adjustment should be added or subtracted.
12151 (vect_update_init_of_drs): Likewise.
12152 (vect_prepare_for_masked_peels): New function.
12153 (vect_do_peeling): Skip prologue peeling if we're using a mask
12154 instead. Update call to vect_update_inits_of_drs.
12155 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12157 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12158 alignment. Do not include the number of peeled iterations in
12159 the minimum threshold in that case.
12160 (vectorizable_induction): Adjust the start value down by
12161 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12162 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12163 Take the number of skipped iterations into account when calculating
12165 * tree-vect-stmts.c (vect_gen_while_not): New function.
12167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12168 Alan Hayward <alan.hayward@arm.com>
12169 David Sherwood <david.sherwood@arm.com>
12171 * doc/sourcebuild.texi (vect_fully_masked): Document.
12172 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12173 default value to 0.
12174 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12176 (vect_analyze_loop_2): ...here. Don't check the vectorization
12177 factor against the number of loop iterations if the loop is
12180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12181 Alan Hayward <alan.hayward@arm.com>
12182 David Sherwood <david.sherwood@arm.com>
12184 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12185 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12186 (dump_groups): Update accordingly.
12187 (iv_use::mem_type): New member variable.
12188 (address_p): New function.
12189 (record_use): Add a mem_type argument and initialize the new
12191 (record_group_use): Add a mem_type argument. Use address_p.
12192 Remove obsolete null checks of base_object. Update call to record_use.
12193 (find_interesting_uses_op): Update call to record_group_use.
12194 (find_interesting_uses_cond): Likewise.
12195 (find_interesting_uses_address): Likewise.
12196 (get_mem_type_for_internal_fn): New function.
12197 (find_address_like_use): Likewise.
12198 (find_interesting_uses_stmt): Try find_address_like_use before
12199 calling find_interesting_uses_op.
12200 (addr_offset_valid_p): Use the iv mem_type field as the type
12201 of the addressed memory.
12202 (add_autoinc_candidates): Likewise.
12203 (get_address_cost): Likewise.
12204 (split_small_address_groups_p): Use address_p.
12205 (split_address_groups): Likewise.
12206 (add_iv_candidate_for_use): Likewise.
12207 (autoinc_possible_for_pair): Likewise.
12208 (rewrite_groups): Likewise.
12209 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12210 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12211 (get_alias_ptr_type_for_ptr_address): New function.
12212 (rewrite_use_address): Rewrite address uses in calls that were
12213 identified by find_address_like_use.
12215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12216 Alan Hayward <alan.hayward@arm.com>
12217 David Sherwood <david.sherwood@arm.com>
12219 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12221 * gimple-expr.h (is_gimple_addressable: Likewise.
12222 * gimple-expr.c (is_gimple_address): Likewise.
12223 * internal-fn.c (expand_call_mem_ref): New function.
12224 (expand_mask_load_optab_fn): Use it.
12225 (expand_mask_store_optab_fn): Likewise.
12227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12228 Alan Hayward <alan.hayward@arm.com>
12229 David Sherwood <david.sherwood@arm.com>
12231 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12232 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12233 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12234 (cond_umax@var{mode}): Document.
12235 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12236 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12237 (cond_umin_optab, cond_umax_optab): New optabs.
12238 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12239 (COND_IOR, COND_XOR): New internal functions.
12240 * internal-fn.h (get_conditional_internal_fn): Declare.
12241 * internal-fn.c (cond_binary_direct): New macro.
12242 (expand_cond_binary_optab_fn): Likewise.
12243 (direct_cond_binary_optab_supported_p): Likewise.
12244 (get_conditional_internal_fn): New function.
12245 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12246 Cope with reduction statements that are vectorized as calls rather
12248 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12249 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12250 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12251 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12252 (UNSPEC_COND_EOR): New unspecs.
12253 (optab): Add mappings for them.
12254 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12255 (sve_int_op, sve_fp_op): New int attributes.
12257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12258 Alan Hayward <alan.hayward@arm.com>
12259 David Sherwood <david.sherwood@arm.com>
12261 * optabs.def (while_ult_optab): New optab.
12262 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12263 * internal-fn.def (WHILE_ULT): New internal function.
12264 * internal-fn.h (direct_internal_fn_supported_p): New override
12265 that takes two types as argument.
12266 * internal-fn.c (while_direct): New macro.
12267 (expand_while_optab_fn): New function.
12268 (convert_optab_supported_p): Likewise.
12269 (direct_while_optab_supported_p): New macro.
12270 * wide-int.h (wi::udiv_ceil): New function.
12271 * tree-vectorizer.h (rgroup_masks): New structure.
12272 (vec_loop_masks): New typedef.
12273 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12274 and fully_masked_p.
12275 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12276 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12277 (vect_max_vf): New function.
12278 (slpeel_make_loop_iterate_ntimes): Delete.
12279 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12280 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12281 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12282 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12283 internal-fn.h, stor-layout.h and optabs-query.h.
12284 (vect_set_loop_mask): New function.
12285 (add_preheader_seq): Likewise.
12286 (add_header_seq): Likewise.
12287 (interleave_supported_p): Likewise.
12288 (vect_maybe_permute_loop_masks): Likewise.
12289 (vect_set_loop_masks_directly): Likewise.
12290 (vect_set_loop_condition_masked): Likewise.
12291 (vect_set_loop_condition_unmasked): New function, split out from
12292 slpeel_make_loop_iterate_ntimes.
12293 (slpeel_make_loop_iterate_ntimes): Rename to..
12294 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12295 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12296 (vect_do_peeling): Update call accordingly.
12297 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12299 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12300 mask_compare_type, can_fully_mask_p and fully_masked_p.
12301 (release_vec_loop_masks): New function.
12302 (_loop_vec_info): Use it to free the loop masks.
12303 (can_produce_all_loop_masks_p): New function.
12304 (vect_get_max_nscalars_per_iter): Likewise.
12305 (vect_verify_full_masking): Likewise.
12306 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12307 retries, and free the mask rgroups before retrying. Check loop-wide
12308 reasons for disallowing fully-masked loops. Make the final decision
12309 about whether use a fully-masked loop or not.
12310 (vect_estimate_min_profitable_iters): Do not assume that peeling
12311 for the number of iterations will be needed for fully-masked loops.
12312 (vectorizable_reduction): Disable fully-masked loops.
12313 (vectorizable_live_operation): Likewise.
12314 (vect_halve_mask_nunits): New function.
12315 (vect_double_mask_nunits): Likewise.
12316 (vect_record_loop_mask): Likewise.
12317 (vect_get_loop_mask): Likewise.
12318 (vect_transform_loop): Handle the case in which the final loop
12319 iteration might handle a partial vector. Call vect_set_loop_condition
12320 instead of slpeel_make_loop_iterate_ntimes.
12321 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12322 (check_load_store_masking): New function.
12323 (prepare_load_store_mask): Likewise.
12324 (vectorizable_store): Handle fully-masked loops.
12325 (vectorizable_load): Likewise.
12326 (supportable_widening_operation): Use vect_halve_mask_nunits for
12328 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12329 (vect_gen_while): New function.
12330 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12331 (aarch64_uqdec<mode>): New insn.
12333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12334 Alan Hayward <alan.hayward@arm.com>
12335 David Sherwood <david.sherwood@arm.com>
12337 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12338 (reduc_xor_scal_optab): New optabs.
12339 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12340 (reduc_xor_scal_@var{m}): Document.
12341 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12342 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12343 internal functions.
12344 * fold-const-call.c (fold_const_call): Handle them.
12345 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12346 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12347 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12348 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12349 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12350 (UNSPEC_XORV): New unspecs.
12351 (optab): Add entries for them.
12352 (BITWISEV): New int iterator.
12353 (bit_reduc_op): New int attributes.
12355 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12356 Alan Hayward <alan.hayward@arm.com>
12357 David Sherwood <david.sherwood@arm.com>
12359 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12360 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12361 * optabs.def (vec_shl_insert_optab): New optab.
12362 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12363 (duplicate_and_interleave): Likewise.
12364 * tree-vect-loop.c: Include internal-fn.h.
12365 (neutral_op_for_slp_reduction): New function, split out from
12366 get_initial_defs_for_reduction.
12367 (get_initial_def_for_reduction): Handle option 2 for variable-length
12368 vectors by loading the neutral value into a vector and then shifting
12369 the initial value into element 0.
12370 (get_initial_defs_for_reduction): Replace the code argument with
12371 the neutral value calculated by neutral_op_for_slp_reduction.
12372 Use gimple_build_vector for constant-length vectors.
12373 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12374 but the first group_size elements have a neutral value.
12375 Use duplicate_and_interleave otherwise.
12376 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12377 Update call to get_initial_defs_for_reduction. Handle SLP
12378 reductions for variable-length vectors by creating one vector
12379 result for each scalar result, with the elements associated
12380 with other scalar results stubbed out with the neutral value.
12381 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12382 Require IFN_VEC_SHL_INSERT for double reductions on
12383 variable-length vectors, or SLP reductions that have
12384 a neutral value. Require can_duplicate_and_interleave_p
12385 support for variable-length unchained SLP reductions if there
12386 is no neutral value, such as for MIN/MAX reductions. Also require
12387 the number of vector elements to be a multiple of the number of
12388 SLP statements when doing variable-length unchained SLP reductions.
12389 Update call to vect_create_epilog_for_reduction.
12390 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12391 and remove initial values.
12392 (duplicate_and_interleave): Make public.
12393 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12394 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12397 Alan Hayward <alan.hayward@arm.com>
12398 David Sherwood <david.sherwood@arm.com>
12400 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12401 (can_duplicate_and_interleave_p): New function.
12402 (vect_get_and_check_slp_defs): Take the vector of statements
12403 rather than just the current one. Remove excess parentheses.
12404 Restriction rejectinon of vect_constant_def and vect_external_def
12405 for variable-length vectors to boolean types, or types for which
12406 can_duplicate_and_interleave_p is false.
12407 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12408 (duplicate_and_interleave): New function.
12409 (vect_get_constant_vectors): Use gimple_build_vector for
12410 constant-length vectors and suitable variable-length constant
12411 vectors. Use duplicate_and_interleave for other variable-length
12412 vectors. Don't defer the update when inserting new statements.
12414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12415 Alan Hayward <alan.hayward@arm.com>
12416 David Sherwood <david.sherwood@arm.com>
12418 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12419 min_profitable_iters doesn't go negative.
12421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12422 Alan Hayward <alan.hayward@arm.com>
12423 David Sherwood <david.sherwood@arm.com>
12425 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12426 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12427 * optabs.def (vec_mask_load_lanes_optab): New optab.
12428 (vec_mask_store_lanes_optab): Likewise.
12429 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12430 (MASK_STORE_LANES): Likewise.
12431 * internal-fn.c (mask_load_lanes_direct): New macro.
12432 (mask_store_lanes_direct): Likewise.
12433 (expand_mask_load_optab_fn): Handle masked operations.
12434 (expand_mask_load_lanes_optab_fn): New macro.
12435 (expand_mask_store_optab_fn): Handle masked operations.
12436 (expand_mask_store_lanes_optab_fn): New macro.
12437 (direct_mask_load_lanes_optab_supported_p): Likewise.
12438 (direct_mask_store_lanes_optab_supported_p): Likewise.
12439 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12441 (vect_load_lanes_supported): Likewise.
12442 * tree-vect-data-refs.c (strip_conversion): New function.
12443 (can_group_stmts_p): Likewise.
12444 (vect_analyze_data_ref_accesses): Use it instead of checking
12445 for a pair of assignments.
12446 (vect_store_lanes_supported): Take a masked_p parameter.
12447 (vect_load_lanes_supported): Likewise.
12448 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12449 vect_store_lanes_supported and vect_load_lanes_supported.
12450 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12451 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12452 parameter. Don't allow gaps for masked accesses.
12453 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12454 and vect_load_lanes_supported.
12455 (get_load_store_type): Take a masked_p parameter and update
12456 call to get_group_load_store_type.
12457 (vectorizable_store): Update call to get_load_store_type.
12458 Handle IFN_MASK_STORE_LANES.
12459 (vectorizable_load): Update call to get_load_store_type.
12460 Handle IFN_MASK_LOAD_LANES.
12462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12463 Alan Hayward <alan.hayward@arm.com>
12464 David Sherwood <david.sherwood@arm.com>
12466 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12468 * config/aarch64/aarch64-protos.h
12469 (aarch64_sve_struct_memory_operand_p): Declare.
12470 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12471 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12472 (VPRED, vpred): Handle SVE structure modes.
12473 * config/aarch64/constraints.md (Utx): New constraint.
12474 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12475 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12476 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12477 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12478 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12479 structure modes. Split into pieces after RA.
12480 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12481 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12483 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12484 SVE structure modes.
12485 (aarch64_classify_address): Likewise.
12486 (sizetochar): Move earlier in file.
12487 (aarch64_print_operand): Handle SVE register lists.
12488 (aarch64_array_mode): New function.
12489 (aarch64_sve_struct_memory_operand_p): Likewise.
12490 (TARGET_ARRAY_MODE): Redefine.
12492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12493 Alan Hayward <alan.hayward@arm.com>
12494 David Sherwood <david.sherwood@arm.com>
12496 * target.def (array_mode): New target hook.
12497 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12498 * doc/tm.texi: Regenerate.
12499 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12500 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12501 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12502 targetm.array_mode.
12503 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12507 Alan Hayward <alan.hayward@arm.com>
12508 David Sherwood <david.sherwood@arm.com>
12510 * fold-const.c (fold_binary_loc): Check the argument types
12511 rather than the result type when testing for a vector operation.
12513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12515 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12516 * doc/tm.texi: Regenerate.
12518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12519 Alan Hayward <alan.hayward@arm.com>
12520 David Sherwood <david.sherwood@arm.com>
12522 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12523 (sve): Document new AArch64 extension.
12524 * doc/md.texi (w): Extend the description of the AArch64
12525 constraint to include SVE vectors.
12526 (Upl, Upa): Document new AArch64 predicate constraints.
12527 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12529 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12530 (msve-vector-bits=): New option.
12531 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12532 SVE when these are disabled.
12533 (sve): New extension.
12534 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12535 modes. Adjust their number of units based on aarch64_sve_vg.
12536 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12537 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12538 aarch64_addr_query_type.
12539 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12540 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12541 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12542 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12543 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12544 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12545 (aarch64_simd_imm_zero_p): Delete.
12546 (aarch64_check_zero_based_sve_index_immediate): Declare.
12547 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12548 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12549 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12550 (aarch64_sve_float_mul_immediate_p): Likewise.
12551 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12552 rather than an rtx.
12553 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12554 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12555 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12556 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12557 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12558 (aarch64_regmode_natural_size): Likewise.
12559 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12560 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12562 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12563 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12564 for VG and the SVE predicate registers.
12565 (V_ALIASES): Add a "z"-prefixed alias.
12566 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12567 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12568 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12569 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12570 (REG_CLASS_NAMES): Add entries for them.
12571 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12572 and the predicate registers.
12573 (aarch64_sve_vg): Declare.
12574 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12575 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12576 (REGMODE_NATURAL_SIZE): Define.
12577 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12579 * config/aarch64/aarch64.c: Include cfgrtl.h.
12580 (simd_immediate_info): Add a constructor for series vectors,
12581 and an associated step field.
12582 (aarch64_sve_vg): New variable.
12583 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12584 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12585 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12586 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12587 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12588 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12589 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12590 (aarch64_get_mask_mode): New functions.
12591 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12592 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12593 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12594 predicate modes and predicate registers. Explicitly restrict
12595 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12596 to store a vector mode if it is recognized by
12597 aarch64_classify_vector_mode.
12598 (aarch64_regmode_natural_size): New function.
12599 (aarch64_hard_regno_caller_save_mode): Return the original mode
12601 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12602 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12603 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12604 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12606 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12607 does not overlap dest if the function is frame-related. Handle
12609 (aarch64_split_add_offset): New function.
12610 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12611 them aarch64_add_offset.
12612 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12613 and update call to aarch64_sub_sp.
12614 (aarch64_add_cfa_expression): New function.
12615 (aarch64_expand_prologue): Pass extra temporary registers to the
12616 functions above. Handle the case in which we need to emit new
12617 DW_CFA_expressions for registers that were originally saved
12618 relative to the stack pointer, but now have to be expressed
12619 relative to the frame pointer.
12620 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12622 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12623 IP0 and IP1 values for SVE frames.
12624 (aarch64_expand_vec_series): New function.
12625 (aarch64_expand_sve_widened_duplicate): Likewise.
12626 (aarch64_expand_sve_const_vector): Likewise.
12627 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12628 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12629 into the register, rather than emitting a SET directly.
12630 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12631 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12632 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12633 (offset_9bit_signed_scaled_p): New functions.
12634 (aarch64_replicate_bitmask_imm): New function.
12635 (aarch64_bitmask_imm): Use it.
12636 (aarch64_cannot_force_const_mem): Reject expressions involving
12637 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12638 (aarch64_classify_index): Handle SVE indices, by requiring
12639 a plain register index with a scale that matches the element size.
12640 (aarch64_classify_address): Handle SVE addresses. Assert that
12641 the mode of the address is VOIDmode or an integer mode.
12642 Update call to aarch64_classify_symbol.
12643 (aarch64_classify_symbolic_expression): Update call to
12644 aarch64_classify_symbol.
12645 (aarch64_const_vec_all_in_range_p): New function.
12646 (aarch64_print_vector_float_operand): Likewise.
12647 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12648 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12649 and the FP immediates 1.0 and 0.5.
12650 (aarch64_print_address_internal): Handle SVE addresses.
12651 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12652 (aarch64_regno_regclass): Handle predicate registers.
12653 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12655 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12656 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12657 (aarch64_convert_sve_vector_bits): New function.
12658 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12659 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12660 rather than an rtx.
12661 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12662 Handle SVE vector and predicate modes. Accept VL-based constants
12663 that need only one temporary register, and VL offsets that require
12664 no temporary registers.
12665 (aarch64_conditional_register_usage): Mark the predicate registers
12666 as fixed if SVE isn't available.
12667 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12668 Return true for SVE vector and predicate modes.
12669 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12670 rather than an unsigned int. Handle SVE modes.
12671 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12673 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12675 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12676 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12677 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12678 (aarch64_sve_float_mul_immediate_p): New functions.
12679 (aarch64_sve_valid_immediate): New function.
12680 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12681 Explicitly reject structure modes. Check for INDEX constants.
12682 Handle PTRUE and PFALSE constants.
12683 (aarch64_check_zero_based_sve_index_immediate): New function.
12684 (aarch64_simd_imm_zero_p): Delete.
12685 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12686 vector modes. Accept constants in the range of CNT[BHWD].
12687 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12688 ask for an Advanced SIMD mode.
12689 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12690 (aarch64_simd_vector_alignment): Handle SVE predicates.
12691 (aarch64_vectorize_preferred_vector_alignment): New function.
12692 (aarch64_simd_vector_alignment_reachable): Use it instead of
12694 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12695 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12697 (MAX_VECT_LEN): Delete.
12698 (expand_vec_perm_d): Add a vec_flags field.
12699 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12700 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12701 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12703 (aarch64_evpc_rev): Rename to...
12704 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12705 (aarch64_evpc_rev_global): New function.
12706 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12707 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12709 (aarch64_evpc_sve_tbl): New function.
12710 (aarch64_expand_vec_perm_const_1): Update after rename of
12711 aarch64_evpc_rev. Handle SVE permutes too, trying
12712 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12713 than aarch64_evpc_tbl.
12714 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12715 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12716 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12717 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12718 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12719 (aarch64_expand_sve_vcond): New functions.
12720 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12721 of aarch64_vector_mode_p.
12722 (aarch64_dwarf_poly_indeterminate_value): New function.
12723 (aarch64_compute_pressure_classes): Likewise.
12724 (aarch64_can_change_mode_class): Likewise.
12725 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12726 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12727 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12728 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12729 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12730 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12731 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12732 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12734 (Dn, Dl, Dr): Accept const as well as const_vector.
12735 (Dz): Likewise. Compare against CONST0_RTX.
12736 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12737 of "vector" where appropriate.
12738 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12739 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12740 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12741 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12742 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12743 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12744 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12745 (v_int_equiv): Extend to SVE modes.
12746 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12748 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12749 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12750 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12751 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12752 (SVE_COND_FP_CMP): New int iterators.
12753 (perm_hilo): Handle the new unpack unspecs.
12754 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12756 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12757 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12758 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12759 (aarch64_equality_operator, aarch64_constant_vector_operand)
12760 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12761 (aarch64_sve_nonimmediate_operand): Likewise.
12762 (aarch64_sve_general_operand): Likewise.
12763 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12764 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12765 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12766 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12767 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12768 (aarch64_sve_float_arith_immediate): Likewise.
12769 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12770 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12771 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12772 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12773 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12774 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12775 (aarch64_sve_float_arith_operand): Likewise.
12776 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12777 (aarch64_sve_float_mul_operand): Likewise.
12778 (aarch64_sve_vec_perm_operand): Likewise.
12779 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12780 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12781 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12782 as well as const_vector.
12783 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12784 in file. Use CONST0_RTX and CONSTM1_RTX.
12785 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12786 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12787 Use aarch64_simd_imm_zero.
12788 * config/aarch64/aarch64-sve.md: New file.
12789 * config/aarch64/aarch64.md: Include it.
12790 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12791 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12792 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12793 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12794 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12795 (sve): New attribute.
12796 (enabled): Disable instructions with the sve attribute unless
12798 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12799 aarch64_expand_mov_immediate.
12800 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12801 CNT[BHSD] immediates.
12802 (movti): Split CONST_POLY_INT moves into two halves.
12803 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12804 Split additions that need a temporary here if the destination
12805 is the stack pointer.
12806 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12807 (*add<mode>3_poly_1): New instruction.
12808 (set_clobber_cc): New expander.
12810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12812 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12813 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12814 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12815 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12816 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12817 Change innermode from fixed_mode_size to machine_mode.
12818 (simplify_subreg): Update call accordingly. Handle a constant-sized
12819 subreg of a variable-length CONST_VECTOR.
12821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12826 (add_offset_to_base): New function, split out from...
12827 (create_mem_ref): ...here. When handling a scale other than 1,
12828 check first whether the address is valid without the offset.
12829 Add it into the base if so, leaving the index and scale as-is.
12831 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12834 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12835 fold_for_warn before checking if arg2 is INTEGER_CST.
12837 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12839 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12840 (store_multiple_operation): Delete.
12841 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12842 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12843 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12844 guarded by TARGET_STRING.
12845 (rs6000_output_load_multiple): Delete.
12846 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12847 OPTION_MASK_STRING / TARGET_STRING handling.
12848 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12849 (const rs6000_opt_masks) <"string">: Change mask to 0.
12850 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12851 (MASK_STRING): Delete.
12852 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12854 (load_multiple): Delete.
12861 (store_multiple): Delete.
12868 (movmemsi_8reg): Delete.
12869 (corresponding unnamed define_insn): Delete.
12870 (movmemsi_6reg): Delete.
12871 (corresponding unnamed define_insn): Delete.
12872 (movmemsi_4reg): Delete.
12873 (corresponding unnamed define_insn): Delete.
12874 (movmemsi_2reg): Delete.
12875 (corresponding unnamed define_insn): Delete.
12876 (movmemsi_1reg): Delete.
12877 (corresponding unnamed define_insn): Delete.
12878 * config/rs6000/rs6000.opt (mno-string): New.
12879 (mstring): Replace by deprecation warning stub.
12880 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12882 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12884 * regrename.c (regrename_do_replace): If replacing the same
12885 reg multiple times, try to reuse last created gen_raw_REG.
12888 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12889 main to workaround a bug in GDB.
12891 2018-01-12 Tom de Vries <tom@codesourcery.com>
12894 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12896 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12898 PR rtl-optimization/80481
12899 * ira-color.c (get_cap_member): New function.
12900 (allocnos_conflict_by_live_ranges_p): Use it.
12901 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12902 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12904 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12907 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12908 (*saddl_se_1): Ditto.
12909 (*ssubsi_1): Ditto.
12910 (*ssubl_se_1): Ditto.
12912 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12914 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12915 rather than wi::to_widest for DR_INITs.
12916 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12917 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12918 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12920 (vect_analyze_group_access_1): Note that here.
12922 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12924 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12925 polynomial type sizes.
12927 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12929 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12930 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12931 (gimple_add_tmp_var): Likewise.
12933 2018-01-12 Martin Liska <mliska@suse.cz>
12935 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12936 (gimple_alloc_sizes): Likewise.
12937 (dump_gimple_statistics): Use PRIu64 in printf format.
12938 * gimple.h: Change uint64_t to int.
12940 2018-01-12 Martin Liska <mliska@suse.cz>
12942 * tree-core.h: Use uint64_t instead of int.
12943 * tree.c (tree_node_counts): Likewise.
12944 (tree_node_sizes): Likewise.
12945 (dump_tree_statistics): Use PRIu64 in printf format.
12947 2018-01-12 Martin Liska <mliska@suse.cz>
12949 * Makefile.in: As qsort_chk is implemented in vec.c, add
12950 vec.o to linkage of gencfn-macros.
12951 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12952 passing the info to record_node_allocation_statistics.
12953 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12955 * ggc-common.c (struct ggc_usage): Add operator== and use
12956 it in operator< and compare function.
12957 * mem-stats.h (struct mem_usage): Likewise.
12958 * vec.c (struct vec_usage): Remove operator< and compare
12959 function. Can be simply inherited.
12961 2018-01-12 Martin Jambor <mjambor@suse.cz>
12964 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12965 * tree-ssa-math-opts.c: Include domwalk.h.
12966 (convert_mult_to_fma_1): New function.
12967 (fma_transformation_info): New type.
12968 (fma_deferring_state): Likewise.
12969 (cancel_fma_deferring): New function.
12970 (result_of_phi): Likewise.
12971 (last_fma_candidate_feeds_initial_phi): Likewise.
12972 (convert_mult_to_fma): Added deferring logic, split actual
12973 transformation to convert_mult_to_fma_1.
12974 (math_opts_dom_walker): New type.
12975 (math_opts_dom_walker::after_dom_children): New method, body moved
12976 here from pass_optimize_widening_mul::execute, added deferring logic
12978 (pass_optimize_widening_mul::execute): Moved most of code to
12979 math_opts_dom_walker::after_dom_children.
12980 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12981 * config/i386/i386.c (ix86_option_override_internal): Added
12982 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12984 2018-01-12 Richard Biener <rguenther@suse.de>
12987 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12988 inline instance vars.
12990 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12993 * config/rx/rx.c (rx_is_restricted_memory_address):
12994 Handle SUBREG case.
12996 2018-01-12 Richard Biener <rguenther@suse.de>
12998 PR tree-optimization/80846
12999 * target.def (split_reduction): New target hook.
13000 * targhooks.c (default_split_reduction): New function.
13001 * targhooks.h (default_split_reduction): Declare.
13002 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13003 target requests first reduce vectors by combining low and high
13005 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13006 (get_vectype_for_scalar_type_and_size): Export.
13007 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13008 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13009 * doc/tm.texi: Regenerate.
13010 * config/i386/i386.c (ix86_split_reduction): Implement
13011 TARGET_VECTORIZE_SPLIT_REDUCTION.
13013 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13016 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13017 in PIC mode except for TARGET_VXWORKS_RTP.
13018 * config/sparc/sparc.c: Include cfgrtl.h.
13019 (TARGET_INIT_PIC_REG): Define.
13020 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13021 (sparc_pic_register_p): New predicate.
13022 (sparc_legitimate_address_p): Use it.
13023 (sparc_legitimize_pic_address): Likewise.
13024 (sparc_delegitimize_address): Likewise.
13025 (sparc_mode_dependent_address_p): Likewise.
13026 (gen_load_pcrel_sym): Remove 4th parameter.
13027 (load_got_register): Adjust call to above. Remove obsolete stuff.
13028 (sparc_expand_prologue): Do not call load_got_register here.
13029 (sparc_flat_expand_prologue): Likewise.
13030 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13031 (sparc_use_pseudo_pic_reg): New function.
13032 (sparc_init_pic_reg): Likewise.
13033 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13034 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13036 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13038 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13039 Add item for branch_cost.
13041 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13043 PR rtl-optimization/83565
13044 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13045 not extend the result to a larger mode for rotate operations.
13046 (num_sign_bit_copies1): Likewise.
13048 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13051 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13053 Use values-Xc.o for -pedantic.
13054 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13056 2018-01-12 Martin Liska <mliska@suse.cz>
13059 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13061 (possible_polymorphic_call_targets): Use it.
13062 (ipa_devirt): Likewise.
13064 2018-01-12 Martin Liska <mliska@suse.cz>
13066 * profile-count.h (enum profile_quality): Use 0 as invalid
13067 enum value of profile_quality.
13069 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13071 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13072 -mext-string options.
13074 2018-01-12 Richard Biener <rguenther@suse.de>
13076 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13077 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13078 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13080 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13082 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13084 * configure.ac (--with-long-double-format): Add support for the
13085 configuration option to change the default long double format on
13087 * config.gcc (powerpc*-linux*-*): Likewise.
13088 * configure: Regenerate.
13089 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13090 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13091 used without modification.
13093 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13095 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13096 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13097 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13098 MISC_BUILTIN_SPEC_BARRIER.
13099 (rs6000_init_builtins): Likewise.
13100 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13102 (speculation_barrier): New define_insn.
13103 * doc/extend.texi: Document __builtin_speculation_barrier.
13105 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13108 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13109 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13110 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13112 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13113 integral modes instead of "ss" and "sd".
13114 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13115 vectors with 32-bit and 64-bit elements.
13116 (vecdupssescalarmodesuffix): New mode attribute.
13117 (vec_dup<mode>): Use it.
13119 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13122 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13123 frame if argument is passed on stack.
13125 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13128 * ree.c (combine_reaching_defs): Optimize also
13129 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13130 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13132 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13134 PR middle-end/83189
13135 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13137 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13139 PR middle-end/83718
13140 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13141 after they are computed.
13143 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13145 PR tree-optimization/83695
13146 * gimple-loop-linterchange.cc
13147 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13148 reset cached scev information after interchange.
13149 (pass_linterchange::execute): Remove call to scev_reset_htab.
13151 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13153 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13154 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13155 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13156 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13157 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13158 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13159 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13160 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13161 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13162 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13163 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13164 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13165 (V_lane_reg): Likewise.
13166 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13168 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13169 (vfmal_lane_low<mode>_intrinsic,
13170 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13171 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13172 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13173 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13174 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13175 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13177 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13179 * config/arm/arm-cpus.in (fp16fml): New feature.
13180 (ALL_SIMD): Add fp16fml.
13181 (armv8.2-a): Add fp16fml as an option.
13182 (armv8.3-a): Likewise.
13183 (armv8.4-a): Add fp16fml as part of fp16.
13184 * config/arm/arm.h (TARGET_FP16FML): Define.
13185 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13187 * config/arm/arm-modes.def (V2HF): Define.
13188 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13189 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13190 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13191 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13192 vfmsl_low, vfmsl_high): New set of builtins.
13193 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13194 (vfml_op): New code attribute.
13195 (VFMLHALVES): New int iterator.
13196 (VFML, VFMLSEL): New mode attributes.
13197 (V_reg): Define mapping for V2HF.
13198 (V_hi, V_lo): New mode attributes.
13199 (VF_constraint): Likewise.
13200 (vfml_half, vfml_half_selector): New int attributes.
13201 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13203 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13204 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13206 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13207 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13208 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13209 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13211 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13212 Document new effective target and option set.
13214 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13216 * config/arm/arm-cpus.in (armv8_4): New feature.
13217 (ARMv8_4a): New fgroup.
13218 (armv8.4-a): New arch.
13219 * config/arm/arm-tables.opt: Regenerate.
13220 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13221 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13222 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13223 Add matching rules for -march=armv8.4-a and extensions.
13224 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13226 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13229 * config/rx/rx.md (BW): New mode attribute.
13230 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13232 2018-01-11 Richard Biener <rguenther@suse.de>
13234 PR tree-optimization/83435
13235 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13236 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13237 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13239 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13240 Alan Hayward <alan.hayward@arm.com>
13241 David Sherwood <david.sherwood@arm.com>
13243 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13245 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13246 (aarch64_print_address_internal): Use it to check for a zero offset.
13248 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13249 Alan Hayward <alan.hayward@arm.com>
13250 David Sherwood <david.sherwood@arm.com>
13252 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13253 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13254 Return a poly_int64 rather than a HOST_WIDE_INT.
13255 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13256 rather than a HOST_WIDE_INT.
13257 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13258 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13259 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13260 final_offset from HOST_WIDE_INT to poly_int64.
13261 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13262 to_constant when getting the number of units in an Advanced SIMD
13264 (aarch64_builtin_vectorized_function): Check for a constant number
13266 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13268 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13269 attribute instead of GET_MODE_NUNITS.
13270 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13271 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13272 GET_MODE_SIZE for fixed-size registers.
13273 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13274 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13275 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13276 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13277 (aarch64_print_operand, aarch64_print_address_internal)
13278 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13279 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13280 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13281 Handle polynomial GET_MODE_SIZE.
13282 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13283 wider than SImode without modification.
13284 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13285 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13286 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13287 passing and returning SVE modes.
13288 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13289 rather than GEN_INT.
13290 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13291 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13292 (aarch64_allocate_and_probe_stack_space): Likewise.
13293 (aarch64_layout_frame): Cope with polynomial offsets.
13294 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13295 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13296 polynomial offsets.
13297 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13298 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13299 poly_int64 rather than a HOST_WIDE_INT.
13300 (aarch64_get_separate_components, aarch64_process_components)
13301 (aarch64_expand_prologue, aarch64_expand_epilogue)
13302 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13303 (aarch64_anchor_offset): New function, split out from...
13304 (aarch64_legitimize_address): ...here.
13305 (aarch64_builtin_vectorization_cost): Handle polynomial
13306 TYPE_VECTOR_SUBPARTS.
13307 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13309 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13310 number of elements from the PARALLEL rather than the mode.
13311 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13312 rather than GET_MODE_BITSIZE.
13313 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13314 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13315 (aarch64_expand_vec_perm_const_1): Handle polynomial
13316 d->perm.length () and d->perm elements.
13317 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13318 Apply to_constant to d->perm elements.
13319 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13320 polynomial CONST_VECTOR_NUNITS.
13321 (aarch64_move_pointer): Take amount as a poly_int64 rather
13323 (aarch64_progress_pointer): Avoid temporary variable.
13324 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13325 the mode attribute instead of GET_MODE.
13327 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13328 Alan Hayward <alan.hayward@arm.com>
13329 David Sherwood <david.sherwood@arm.com>
13331 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13332 x exists before using it.
13333 (aarch64_add_constant_internal): Rename to...
13334 (aarch64_add_offset_1): ...this. Replace regnum with separate
13335 src and dest rtxes. Handle the case in which they're different,
13336 including when the offset is zero. Replace scratchreg with an rtx.
13337 Use 2 additions if there is no spare register into which we can
13338 move a 16-bit constant.
13339 (aarch64_add_constant): Delete.
13340 (aarch64_add_offset): Replace reg with separate src and dest
13341 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13342 Use aarch64_add_offset_1.
13343 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13344 an rtx rather than an int. Take the delta as a poly_int64
13345 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13346 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13347 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13348 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13349 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13350 and aarch64_add_sp.
13351 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13352 aarch64_add_constant.
13354 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13356 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13357 Use scalar_float_mode.
13359 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13361 * config/aarch64/aarch64-simd.md
13362 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13363 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13364 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13365 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13366 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13367 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13368 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13369 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13370 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13371 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13373 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13376 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13377 targ_options->x_arm_arch_string is non NULL.
13379 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13381 * config/aarch64/aarch64.h
13382 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13384 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13387 * expmed.c (emit_store_flag_force): Swap if const op0
13388 and change VOIDmode to mode of op0.
13390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13392 PR rtl-optimization/83761
13393 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13394 than bytes to mode_for_size.
13396 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13398 PR middle-end/83189
13399 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13400 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13403 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13405 PR middle-end/83575
13406 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13407 when in layout mode.
13408 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13409 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13412 2018-01-10 Michael Collison <michael.collison@arm.com>
13414 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13415 * config/aarch64/aarch64-option-extension.def: Add
13416 AARCH64_OPT_EXTENSION of 'fp16fml'.
13417 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13418 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13419 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13420 * config/aarch64/constraints.md (Ui7): New constraint.
13421 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13422 (VFMLA_SEL_W): Ditto.
13425 (VFMLA16_LOW): New int iterator.
13426 (VFMLA16_HIGH): Ditto.
13427 (UNSPEC_FMLAL): New unspec.
13428 (UNSPEC_FMLSL): Ditto.
13429 (UNSPEC_FMLAL2): Ditto.
13430 (UNSPEC_FMLSL2): Ditto.
13431 (f16mac): New code attribute.
13432 * config/aarch64/aarch64-simd-builtins.def
13433 (aarch64_fmlal_lowv2sf): Ditto.
13434 (aarch64_fmlsl_lowv2sf): Ditto.
13435 (aarch64_fmlalq_lowv4sf): Ditto.
13436 (aarch64_fmlslq_lowv4sf): Ditto.
13437 (aarch64_fmlal_highv2sf): Ditto.
13438 (aarch64_fmlsl_highv2sf): Ditto.
13439 (aarch64_fmlalq_highv4sf): Ditto.
13440 (aarch64_fmlslq_highv4sf): Ditto.
13441 (aarch64_fmlal_lane_lowv2sf): Ditto.
13442 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13443 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13444 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13445 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13446 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13447 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13448 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13449 (aarch64_fmlal_lane_highv2sf): Ditto.
13450 (aarch64_fmlsl_lane_highv2sf): Ditto.
13451 (aarch64_fmlal_laneq_highv2sf): Ditto.
13452 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13453 (aarch64_fmlalq_lane_highv4sf): Ditto.
13454 (aarch64_fmlsl_lane_highv4sf): Ditto.
13455 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13456 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13457 * config/aarch64/aarch64-simd.md:
13458 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13459 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13460 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13461 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13462 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13463 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13464 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13465 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13466 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13467 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13468 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13469 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13470 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13471 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13472 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13473 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13474 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13475 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13476 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13477 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13478 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13479 (vfmlsl_low_u32): Ditto.
13480 (vfmlalq_low_u32): Ditto.
13481 (vfmlslq_low_u32): Ditto.
13482 (vfmlal_high_u32): Ditto.
13483 (vfmlsl_high_u32): Ditto.
13484 (vfmlalq_high_u32): Ditto.
13485 (vfmlslq_high_u32): Ditto.
13486 (vfmlal_lane_low_u32): Ditto.
13487 (vfmlsl_lane_low_u32): Ditto.
13488 (vfmlal_laneq_low_u32): Ditto.
13489 (vfmlsl_laneq_low_u32): Ditto.
13490 (vfmlalq_lane_low_u32): Ditto.
13491 (vfmlslq_lane_low_u32): Ditto.
13492 (vfmlalq_laneq_low_u32): Ditto.
13493 (vfmlslq_laneq_low_u32): Ditto.
13494 (vfmlal_lane_high_u32): Ditto.
13495 (vfmlsl_lane_high_u32): Ditto.
13496 (vfmlal_laneq_high_u32): Ditto.
13497 (vfmlsl_laneq_high_u32): Ditto.
13498 (vfmlalq_lane_high_u32): Ditto.
13499 (vfmlslq_lane_high_u32): Ditto.
13500 (vfmlalq_laneq_high_u32): Ditto.
13501 (vfmlslq_laneq_high_u32): Ditto.
13502 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13503 (AARCH64_FL_FOR_ARCH8_4): New.
13504 (AARCH64_ISA_F16FML): New ISA flag.
13505 (TARGET_F16FML): New feature flag for fp16fml.
13506 (doc/invoke.texi): Document new fp16fml option.
13508 2018-01-10 Michael Collison <michael.collison@arm.com>
13510 * config/aarch64/aarch64-builtins.c:
13511 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13512 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13513 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13514 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13515 (AARCH64_ISA_SHA3): New ISA flag.
13516 (TARGET_SHA3): New feature flag for sha3.
13517 * config/aarch64/iterators.md (sha512_op): New int attribute.
13518 (CRYPTO_SHA512): New int iterator.
13519 (UNSPEC_SHA512H): New unspec.
13520 (UNSPEC_SHA512H2): Ditto.
13521 (UNSPEC_SHA512SU0): Ditto.
13522 (UNSPEC_SHA512SU1): Ditto.
13523 * config/aarch64/aarch64-simd-builtins.def
13524 (aarch64_crypto_sha512hqv2di): New builtin.
13525 (aarch64_crypto_sha512h2qv2di): Ditto.
13526 (aarch64_crypto_sha512su0qv2di): Ditto.
13527 (aarch64_crypto_sha512su1qv2di): Ditto.
13528 (aarch64_eor3qv8hi): Ditto.
13529 (aarch64_rax1qv2di): Ditto.
13530 (aarch64_xarqv2di): Ditto.
13531 (aarch64_bcaxqv8hi): Ditto.
13532 * config/aarch64/aarch64-simd.md:
13533 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13534 (aarch64_crypto_sha512su0qv2di): Ditto.
13535 (aarch64_crypto_sha512su1qv2di): Ditto.
13536 (aarch64_eor3qv8hi): Ditto.
13537 (aarch64_rax1qv2di): Ditto.
13538 (aarch64_xarqv2di): Ditto.
13539 (aarch64_bcaxqv8hi): Ditto.
13540 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13541 (vsha512h2q_u64): Ditto.
13542 (vsha512su0q_u64): Ditto.
13543 (vsha512su1q_u64): Ditto.
13544 (veor3q_u16): Ditto.
13545 (vrax1q_u64): Ditto.
13546 (vxarq_u64): Ditto.
13547 (vbcaxq_u16): Ditto.
13548 * config/arm/types.md (crypto_sha512): New type attribute.
13549 (crypto_sha3): Ditto.
13550 (doc/invoke.texi): Document new sha3 option.
13552 2018-01-10 Michael Collison <michael.collison@arm.com>
13554 * config/aarch64/aarch64-builtins.c:
13555 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13556 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13557 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13558 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13559 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13560 (AARCH64_ISA_SM4): New ISA flag.
13561 (TARGET_SM4): New feature flag for sm4.
13562 * config/aarch64/aarch64-simd-builtins.def
13563 (aarch64_sm3ss1qv4si): Ditto.
13564 (aarch64_sm3tt1aq4si): Ditto.
13565 (aarch64_sm3tt1bq4si): Ditto.
13566 (aarch64_sm3tt2aq4si): Ditto.
13567 (aarch64_sm3tt2bq4si): Ditto.
13568 (aarch64_sm3partw1qv4si): Ditto.
13569 (aarch64_sm3partw2qv4si): Ditto.
13570 (aarch64_sm4eqv4si): Ditto.
13571 (aarch64_sm4ekeyqv4si): Ditto.
13572 * config/aarch64/aarch64-simd.md:
13573 (aarch64_sm3ss1qv4si): Ditto.
13574 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13575 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13576 (aarch64_sm4eqv4si): Ditto.
13577 (aarch64_sm4ekeyqv4si): Ditto.
13578 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13579 (sm3part_op): Ditto.
13580 (CRYPTO_SM3TT): Ditto.
13581 (CRYPTO_SM3PART): Ditto.
13582 (UNSPEC_SM3SS1): New unspec.
13583 (UNSPEC_SM3TT1A): Ditto.
13584 (UNSPEC_SM3TT1B): Ditto.
13585 (UNSPEC_SM3TT2A): Ditto.
13586 (UNSPEC_SM3TT2B): Ditto.
13587 (UNSPEC_SM3PARTW1): Ditto.
13588 (UNSPEC_SM3PARTW2): Ditto.
13589 (UNSPEC_SM4E): Ditto.
13590 (UNSPEC_SM4EKEY): Ditto.
13591 * config/aarch64/constraints.md (Ui2): New constraint.
13592 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13593 * config/arm/types.md (crypto_sm3): New type attribute.
13594 (crypto_sm4): Ditto.
13595 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13596 (vsm3tt1aq_u32): Ditto.
13597 (vsm3tt1bq_u32): Ditto.
13598 (vsm3tt2aq_u32): Ditto.
13599 (vsm3tt2bq_u32): Ditto.
13600 (vsm3partw1q_u32): Ditto.
13601 (vsm3partw2q_u32): Ditto.
13602 (vsm4eq_u32): Ditto.
13603 (vsm4ekeyq_u32): Ditto.
13604 (doc/invoke.texi): Document new sm4 option.
13606 2018-01-10 Michael Collison <michael.collison@arm.com>
13608 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13609 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13610 (AARCH64_FL_FOR_ARCH8_4): New.
13611 (AARCH64_FL_V8_4): New flag.
13612 (doc/invoke.texi): Document new armv8.4-a option.
13614 2018-01-10 Michael Collison <michael.collison@arm.com>
13616 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13617 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13618 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13619 * config/aarch64/aarch64-option-extension.def: Add
13620 AARCH64_OPT_EXTENSION of 'sha2'.
13621 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13622 (crypto): Disable sha2 and aes if crypto disabled.
13623 (crypto): Enable aes and sha2 if enabled.
13624 (simd): Disable sha2 and aes if simd disabled.
13625 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13627 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13628 (TARGET_SHA2): New feature flag for sha2.
13629 (TARGET_AES): New feature flag for aes.
13630 * config/aarch64/aarch64-simd.md:
13631 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13632 conditional on TARGET_AES.
13633 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13634 (aarch64_crypto_sha1hsi): Make pattern conditional
13636 (aarch64_crypto_sha1hv4si): Ditto.
13637 (aarch64_be_crypto_sha1hv4si): Ditto.
13638 (aarch64_crypto_sha1su1v4si): Ditto.
13639 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13640 (aarch64_crypto_sha1su0v4si): Ditto.
13641 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13642 (aarch64_crypto_sha256su0v4si): Ditto.
13643 (aarch64_crypto_sha256su1v4si): Ditto.
13644 (doc/invoke.texi): Document new aes and sha2 options.
13646 2018-01-10 Martin Sebor <msebor@redhat.com>
13648 PR tree-optimization/83781
13649 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13652 2018-01-11 Martin Sebor <msebor@gmail.com>
13653 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13655 PR tree-optimization/83501
13656 PR tree-optimization/81703
13658 * tree-ssa-strlen.c (get_string_cst): Rename...
13659 (get_string_len): ...to this. Handle global constants.
13660 (handle_char_store): Adjust.
13662 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13663 Jim Wilson <jimw@sifive.com>
13665 * config/riscv/riscv-protos.h (riscv_output_return): New.
13666 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13667 (riscv_attribute_table, riscv_output_return),
13668 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13669 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13670 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13672 (riscv_expand_prologue): Add early return for naked function.
13673 (riscv_expand_epilogue): Likewise.
13674 (riscv_function_ok_for_sibcall): Return false for naked function.
13675 (riscv_set_current_function): New.
13676 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13677 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13678 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13679 * doc/extend.texi (RISC-V Function Attributes): New.
13681 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13683 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13684 check for 128-bit long double before checking TCmode.
13685 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13686 128-bit long doubles before checking TFmode or TCmode.
13687 (FLOAT128_IBM_P): Likewise.
13689 2018-01-10 Martin Sebor <msebor@redhat.com>
13691 PR tree-optimization/83671
13692 * builtins.c (c_strlen): Unconditionally return zero for the empty
13694 Use -Warray-bounds for warnings.
13695 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13696 for non-constant array indices with COMPONENT_REF, arrays of
13697 arrays, and pointers to arrays.
13698 (gimple_fold_builtin_strlen): Determine and set length range for
13699 non-constant character arrays.
13701 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13703 PR middle-end/81897
13704 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13707 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13709 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13711 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13714 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13715 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13716 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13717 indexed_or_indirect_operand predicate.
13718 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13719 (*vsx_le_perm_load_v8hi): Likewise.
13720 (*vsx_le_perm_load_v16qi): Likewise.
13721 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13722 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13723 (*vsx_le_perm_store_v8hi): Likewise.
13724 (*vsx_le_perm_store_v16qi): Likewise.
13725 (eight unnamed splitters): Likewise.
13727 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13729 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13730 * config/rs6000/emmintrin.h: Likewise.
13731 * config/rs6000/mmintrin.h: Likewise.
13732 * config/rs6000/xmmintrin.h: Likewise.
13734 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13737 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13739 * tree.c (tree_nop_conversion): Return true for location wrapper
13741 (maybe_wrap_with_location): New function.
13742 (selftest::check_strip_nops): New function.
13743 (selftest::test_location_wrappers): New function.
13744 (selftest::tree_c_tests): Call it.
13745 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13746 (maybe_wrap_with_location): New decl.
13747 (EXPR_LOCATION_WRAPPER_P): New macro.
13748 (location_wrapper_p): New inline function.
13749 (tree_strip_any_location_wrapper): New inline function.
13751 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13754 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13755 stack_realign_offset for the largest alignment of stack slot
13757 (ix86_find_max_used_stack_alignment): New function.
13758 (ix86_finalize_stack_frame_flags): Use it. Set
13759 max_used_stack_alignment if we don't realign stack.
13760 * config/i386/i386.h (machine_function): Add
13761 max_used_stack_alignment.
13763 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13765 * config/arm/arm.opt (-mbranch-cost): New option.
13766 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13769 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13772 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13773 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13775 2018-01-10 Richard Biener <rguenther@suse.de>
13778 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13779 early out so it also covers the case where we have a non-NULL
13782 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13784 PR tree-optimization/83753
13785 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13786 for non-strided grouped accesses if the number of elements is 1.
13788 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13791 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13792 * i386.h (TARGET_USE_GATHER): Define.
13793 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13795 2018-01-10 Martin Liska <mliska@suse.cz>
13798 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13799 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13801 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13802 CLEANUP_NO_PARTITIONING is not set.
13804 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13806 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13807 for vectors, as a partial revert of r254296.
13808 * rtl.h (const_vec_p): Delete.
13809 (const_vec_duplicate_p): Don't test for vector CONSTs.
13810 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13811 * expmed.c (make_tree): Likewise.
13814 * common.md (E, F): Use CONSTANT_P instead of checking for
13816 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13817 checking for CONST_VECTOR.
13819 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13821 PR middle-end/83575
13822 * predict.c (force_edge_cold): Handle in more sane way edges
13823 with no prediction.
13825 2018-01-09 Carl Love <cel@us.ibm.com>
13827 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13829 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13830 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13831 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13832 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13833 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13834 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13835 * config/rs6000/rs6000-protos.h: Add extern defition for
13836 rs6000_generate_float2_double_code.
13837 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13839 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13840 (float2_v2df): Add define_expand.
13842 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13845 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13846 op_mode in the force_to_mode call.
13848 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13850 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13851 instead of checking each element individually.
13852 (aarch64_evpc_uzp): Likewise.
13853 (aarch64_evpc_zip): Likewise.
13854 (aarch64_evpc_ext): Likewise.
13855 (aarch64_evpc_rev): Likewise.
13856 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13857 instead of checking each element individually. Return true without
13859 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13860 whether all selected elements come from the same input, instead of
13861 checking each element individually. Remove calls to gen_rtx_REG,
13862 start_sequence and end_sequence and instead assert that no rtl is
13865 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13867 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13868 order of HIGH and CONST checks.
13870 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13872 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13873 if the destination isn't an SSA_NAME.
13875 2018-01-09 Richard Biener <rguenther@suse.de>
13877 PR tree-optimization/83668
13878 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13880 (canonicalize_loop_form): ... here, renamed from ...
13881 (canonicalize_loop_closed_ssa_form): ... this and amended to
13882 swap successor edges for loop exit blocks to make us use
13883 the RPO order we need for initial schedule generation.
13885 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13887 PR tree-optimization/64811
13888 * match.pd: When optimizing comparisons with Inf, avoid
13889 introducing or losing exceptions from comparisons with NaN.
13891 2018-01-09 Martin Liska <mliska@suse.cz>
13894 * asan.c (shadow_mem_size): Add gcc_assert.
13896 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13898 Don't save registers in main().
13901 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13902 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13903 * config/avr/avr.c (avr_set_current_function): Don't error if
13904 naked, OS_task or OS_main are specified at the same time.
13905 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13907 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13909 * common/config/avr/avr-common.c (avr_option_optimization_table):
13910 Switch on -mmain-is-OS_task for optimizing compilations.
13912 2018-01-09 Richard Biener <rguenther@suse.de>
13914 PR tree-optimization/83572
13915 * graphite.c: Include cfganal.h.
13916 (graphite_transform_loops): Connect infinite loops to exit
13917 and remove fake edges at the end.
13919 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13921 * ipa-inline.c (edge_badness): Revert accidental checkin.
13923 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13926 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13927 symbols; not inline clones.
13929 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13932 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13933 hard registers. Formatting fixes.
13935 PR preprocessor/83722
13936 * gcc.c (try_generate_repro): Pass
13937 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13938 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13941 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13942 Kito Cheng <kito.cheng@gmail.com>
13944 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13945 (riscv_leaf_function_p): Delete.
13946 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13948 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13950 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13952 (do_ifelse): New function.
13953 (do_isel): New function.
13954 (do_sub3): New function.
13955 (do_add3): New function.
13956 (do_load_mask_compare): New function.
13957 (do_overlap_load_compare): New function.
13958 (expand_compare_loop): New function.
13959 (expand_block_compare): Call expand_compare_loop() when appropriate.
13960 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13961 option description.
13962 (-mblock-compare-inline-loop-limit): New option.
13964 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13967 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13968 Reverse order of second and third operands in first alternative.
13969 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13970 of first and second elements in UNSPEC_VPERMR vector.
13971 (altivec_expand_vec_perm_le): Likewise.
13973 2018-01-08 Jeff Law <law@redhat.com>
13975 PR rtl-optimizatin/81308
13976 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13977 (process_switch): If group_case_labels makes a change, then set
13979 (pass_convert_switch::execute): If a switch is converted, then
13980 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13982 PR rtl-optimization/81308
13983 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13986 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13988 PR target/83663 - Revert r255946
13989 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13990 generation for cases where splatting a value is not useful.
13991 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13992 across a vec_duplicate and a paradoxical subreg forming a vector
13993 mode to a vec_concat.
13995 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13997 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13998 -march=armv8.3-a variants.
13999 * config/arm/t-multilib: Likewise.
14000 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14002 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14004 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14006 (cceq_ior_compare_complement): Give it a name so I can use it, and
14007 change boolean_or_operator predicate to boolean_operator so it can
14008 be used to generate a crand.
14009 (eqne): New code iterator.
14010 (bd/bd_neg): New code_attrs.
14011 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14012 a single define_insn.
14013 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14014 decrement (bdnzt/bdnzf/bdzt/bdzf).
14015 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14016 with the new names of the branch decrement patterns, and added the
14017 names of the branch decrement conditional patterns.
14019 2018-01-08 Richard Biener <rguenther@suse.de>
14021 PR tree-optimization/83563
14022 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14025 2018-01-08 Richard Biener <rguenther@suse.de>
14027 PR middle-end/83713
14028 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14030 2018-01-08 Richard Biener <rguenther@suse.de>
14032 PR tree-optimization/83685
14033 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14034 references to abnormals.
14036 2018-01-08 Richard Biener <rguenther@suse.de>
14039 * dwarf2out.c (output_indirect_strings): Handle empty
14040 skeleton_debug_str_hash.
14041 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14043 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14045 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14046 (emit_store_direct): Likewise.
14047 (arc_trampoline_adjust_address): Likewise.
14048 (arc_asm_trampoline_template): New function.
14049 (arc_initialize_trampoline): Use asm_trampoline_template.
14050 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14051 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14052 * config/arc/arc.md (flush_icache): Delete pattern.
14054 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14056 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14057 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14060 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14063 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14064 by not USED_FOR_TARGET.
14065 (make_pass_resolve_sw_modes): Likewise.
14067 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14069 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14072 2018-01-08 Richard Biener <rguenther@suse.de>
14074 PR middle-end/83580
14075 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14077 2018-01-08 Richard Biener <rguenther@suse.de>
14079 PR middle-end/83517
14080 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14082 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14084 PR middle-end/81897
14085 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14086 basic blocks with a small number of successors.
14087 (convert_control_dep_chain_into_preds): Improve handling of
14089 (dump_predicates): Split apart into...
14090 (dump_pred_chain): ...here...
14091 (dump_pred_info): ...and here.
14092 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14093 (can_chain_union_be_invalidated_p): Improve check for invalidation
14095 (uninit_uses_cannot_happen): Avoid unnecessary if
14096 convert_control_dep_chain_into_preds yielded nothing.
14098 2018-01-06 Martin Sebor <msebor@redhat.com>
14100 PR tree-optimization/83640
14101 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14102 subtracting negative offset from size.
14103 (builtin_access::overlap): Adjust offset bounds of the access to fall
14104 within the size of the object if possible.
14106 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14108 PR rtl-optimization/83699
14109 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14110 extract_bit_field_as_subreg to cases in which the extracted
14111 value is also a vector.
14113 * lra-constraints.c (process_alt_operands): Test for the equivalence
14114 substitutions when detecting a possible reload cycle.
14116 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14119 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14120 by default if flag_selective_schedling{,2}. Formatting fixes.
14122 PR rtl-optimization/83682
14123 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14124 if it has non-VECTOR_MODE element mode.
14125 (vec_duplicate_p): Likewise.
14127 PR middle-end/83694
14128 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14129 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14131 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14134 * config/i386/i386-builtin.def
14135 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14136 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14137 Require also OPTION_MASK_ISA_AVX512F in addition to
14138 OPTION_MASK_ISA_GFNI.
14139 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14140 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14141 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14142 to OPTION_MASK_ISA_GFNI.
14143 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14144 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14145 OPTION_MASK_ISA_AVX512BW.
14146 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14147 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14148 addition to OPTION_MASK_ISA_GFNI.
14149 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14150 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14151 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14152 to OPTION_MASK_ISA_GFNI.
14153 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14154 a requirement for all ISAs rather than any of them with a few
14156 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14158 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14159 bitmasks to be enabled with 3 exceptions, instead of requiring any
14160 enabled ISA with lots of exceptions.
14161 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14162 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14163 Change avx512bw in isa attribute to avx512f.
14164 * config/i386/sgxintrin.h: Add license boilerplate.
14165 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14166 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14167 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14168 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14170 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14171 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14172 temporarily sse2 rather than sse if not enabled already.
14175 * config/i386/sse.md (VI248_VLBW): Rename to ...
14176 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14177 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14178 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14179 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14180 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14181 mode iterator instead of VI248_VLBW.
14183 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14185 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14186 (record_modified): Skip clobbers; add debug output.
14187 (param_change_prob): Use sreal frequencies.
14189 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14191 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14192 punt for user-aligned variables.
14194 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14196 * tree-chrec.c (chrec_contains_symbols): Return true for
14199 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14202 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14203 of (x|y) == x for BICS pattern.
14205 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14207 PR tree-optimization/83605
14208 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14209 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14212 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14214 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14215 * config/epiphany/rtems.h: New file.
14217 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14218 Uros Bizjak <ubizjak@gmail.com>
14221 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14222 QIreg_operand instead of register_operand predicate.
14223 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14224 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14225 comments instead of -fmitigate[-_]rop.
14227 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14230 * cgraphunit.c (symbol_table::compile): Switch to text_section
14231 before calling assembly_start debug hook.
14232 * run-rtl-passes.c (run_rtl_passes): Likewise.
14235 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14237 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14238 range_int_cst_p rather than !symbolic_range_p before calling
14239 extract_range_from_multiplicative_op_1.
14241 2018-01-04 Jeff Law <law@redhat.com>
14243 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14244 redundant test in assertion.
14246 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14248 * doc/rtl.texi: Document machine_mode wrapper classes.
14250 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14252 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14253 using tree_to_uhwi.
14255 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14257 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14258 the VEC_PERM_EXPR fold to fail.
14260 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14263 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14264 to switched_sections.
14266 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14269 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14270 test for d.testing.
14272 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14275 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14276 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14278 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14281 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14282 is BLKmode and bitpos not zero or mode change is needed.
14284 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14287 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14290 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14293 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14294 instead of MULT rtx. Update all corresponding splitters.
14295 (*saddl_se): Ditto.
14296 (*ssub<modesuffix>): Ditto.
14297 (*ssubl_se): Ditto.
14298 (*cmp_sadd_di): Update split patterns.
14299 (*cmp_sadd_si): Ditto.
14300 (*cmp_sadd_sidi): Ditto.
14301 (*cmp_ssub_di): Ditto.
14302 (*cmp_ssub_si): Ditto.
14303 (*cmp_ssub_sidi): Ditto.
14304 * config/alpha/predicates.md (const23_operand): New predicate.
14305 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14306 Look for ASHIFT, not MULT inner operand.
14307 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14309 2018-01-04 Martin Liska <mliska@suse.cz>
14311 PR gcov-profile/83669
14312 * gcov.c (output_intermediate_file): Add version to intermediate
14314 * doc/gcov.texi: Document new field 'version' in intermediate
14315 file format. Fix location of '-k' option of gcov command.
14317 2018-01-04 Martin Liska <mliska@suse.cz>
14320 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14322 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14324 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14326 2018-01-03 Martin Sebor <msebor@redhat.com>
14328 PR tree-optimization/83655
14329 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14330 checking calls with invalid arguments.
14332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14334 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14335 (vectorizable_mask_load_store): Delete.
14336 (vectorizable_call): Return false for masked loads and stores.
14337 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14338 instead of gimple_assign_rhs1.
14339 (vectorizable_load): Handle IFN_MASK_LOAD.
14340 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14344 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14346 (vectorizable_mask_load_store): ...here.
14347 (vectorizable_load): ...and here.
14349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14351 * tree-vect-stmts.c (vect_build_all_ones_mask)
14352 (vect_build_zero_merge_argument): New functions, split out from...
14353 (vectorizable_load): ...here.
14355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14357 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14359 (vectorizable_mask_load_store): ...here.
14360 (vectorizable_store): ...and here.
14362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14364 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14366 (vectorizable_mask_load_store): ...here.
14368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14370 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14371 (vect_model_store_cost): Take a vec_load_store_type instead of a
14373 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14374 (vect_model_store_cost): Take a vec_load_store_type instead of a
14376 (vectorizable_mask_load_store): Update accordingly.
14377 (vectorizable_store): Likewise.
14378 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14382 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14383 IFN_MASK_LOAD calls here rather than...
14384 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14387 Alan Hayward <alan.hayward@arm.com>
14388 David Sherwood <david.sherwood@arm.com>
14390 * expmed.c (extract_bit_field_1): For vector extracts,
14391 fall back to extract_bit_field_as_subreg if vec_extract
14394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14395 Alan Hayward <alan.hayward@arm.com>
14396 David Sherwood <david.sherwood@arm.com>
14398 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14399 they are variable or constant sized.
14400 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14401 slots for constant-sized data.
14403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14404 Alan Hayward <alan.hayward@arm.com>
14405 David Sherwood <david.sherwood@arm.com>
14407 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14408 handling COND_EXPRs with boolean comparisons, try to find a better
14409 basis for the mask type than the boolean itself.
14411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14413 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14414 is calculated and how it can be overridden.
14415 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14416 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14418 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14422 Alan Hayward <alan.hayward@arm.com>
14423 David Sherwood <david.sherwood@arm.com>
14425 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14426 Remove the mode argument.
14427 (aarch64_simd_valid_immediate): Remove the mode and inverse
14429 * config/aarch64/iterators.md (bitsize): New iterator.
14430 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14431 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14432 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14433 aarch64_simd_valid_immediate.
14434 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14435 (aarch64_reg_or_bic_imm): Likewise.
14436 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14437 with an insn_type enum and msl with a modifier_type enum.
14438 Replace element_width with a scalar_mode. Change the shift
14439 to unsigned int. Add constructors for scalar_float_mode and
14440 scalar_int_mode elements.
14441 (aarch64_vect_float_const_representable_p): Delete.
14442 (aarch64_can_const_movi_rtx_p)
14443 (aarch64_simd_scalar_immediate_valid_for_move)
14444 (aarch64_simd_make_constant): Update call to
14445 aarch64_simd_valid_immediate.
14446 (aarch64_advsimd_valid_immediate_hs): New function.
14447 (aarch64_advsimd_valid_immediate): Likewise.
14448 (aarch64_simd_valid_immediate): Remove mode and inverse
14449 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14450 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14451 and aarch64_float_const_representable_p on the result.
14452 (aarch64_output_simd_mov_immediate): Remove mode argument.
14453 Update call to aarch64_simd_valid_immediate and use of
14454 simd_immediate_info.
14455 (aarch64_output_scalar_simd_mov_immediate): Update call
14458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14459 Alan Hayward <alan.hayward@arm.com>
14460 David Sherwood <david.sherwood@arm.com>
14462 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14463 (mode_nunits): Likewise CONST_MODE_NUNITS.
14464 * machmode.def (ADJUST_NUNITS): Document.
14465 * genmodes.c (mode_data::need_nunits_adj): New field.
14466 (blank_mode): Update accordingly.
14467 (adj_nunits): New variable.
14468 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14470 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14471 listed in adj_nunits.
14472 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14473 listed in adj_nunits. Don't emit case statements for such modes.
14474 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14475 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14476 nothing if adj_nunits is nonnull.
14477 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14478 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14479 (emit_mode_fbit): Update use of print_maybe_const_decl.
14480 (emit_move_size): Likewise. Treat the array as non-const
14482 (emit_mode_adjustments): Handle adj_nunits.
14484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14486 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14487 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14488 (VECTOR_MODES): Use it.
14489 (make_vector_modes): Take the prefix as an argument.
14491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14492 Alan Hayward <alan.hayward@arm.com>
14493 David Sherwood <david.sherwood@arm.com>
14495 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14496 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14497 for MODE_VECTOR_BOOL.
14498 * machmode.def (VECTOR_BOOL_MODE): Document.
14499 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14500 (make_vector_bool_mode): New function.
14501 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14503 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14504 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14506 * stor-layout.c (int_mode_for_mode): Likewise.
14507 * tree.c (build_vector_type_for_mode): Likewise.
14508 * varasm.c (output_constant_pool_2): Likewise.
14509 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14510 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14511 for MODE_VECTOR_BOOL.
14512 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14513 of mode class checks.
14514 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14515 instead of a list of mode class checks.
14516 (expand_vector_scalar_condition): Likewise.
14517 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14520 Alan Hayward <alan.hayward@arm.com>
14521 David Sherwood <david.sherwood@arm.com>
14523 * machmode.h (mode_size): Change from unsigned short to
14525 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14526 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14527 or if measurement_type is not polynomial.
14528 (fixed_size_mode::includes_p): Check for constant-sized modes.
14529 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14530 return a poly_uint16 rather than an unsigned short.
14531 (emit_mode_size): Change the type of mode_size from unsigned short
14532 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14533 (emit_mode_adjustments): Cope with polynomial vector sizes.
14534 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14536 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14538 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14539 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14540 * caller-save.c (setup_save_areas): Likewise.
14541 (replace_reg_with_saved_mem): Likewise.
14542 * calls.c (emit_library_call_value_1): Likewise.
14543 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14544 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14545 (gen_lowpart_for_combine): Likewise.
14546 * convert.c (convert_to_integer_1): Likewise.
14547 * cse.c (equiv_constant, cse_insn): Likewise.
14548 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14549 (cselib_subst_to_values): Likewise.
14550 * dce.c (word_dce_process_block): Likewise.
14551 * df-problems.c (df_word_lr_mark_ref): Likewise.
14552 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14553 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14554 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14555 (rtl_for_decl_location): Likewise.
14556 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14557 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14558 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14559 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14560 (expand_expr_real_1): Likewise.
14561 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14562 (pad_below): Likewise.
14563 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14564 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14565 * ira.c (get_subreg_tracking_sizes): Likewise.
14566 * ira-build.c (ira_create_allocno_objects): Likewise.
14567 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14568 (ira_sort_regnos_for_alter_reg): Likewise.
14569 * ira-costs.c (record_operand_costs): Likewise.
14570 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14571 (resolve_simple_move): Likewise.
14572 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14573 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14574 (lra_constraints): Likewise.
14575 (CONST_POOL_OK_P): Reject variable-sized modes.
14576 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14577 (add_pseudo_to_slot, lra_spill): Likewise.
14578 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14579 * optabs-query.c (get_best_extraction_insn): Likewise.
14580 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14581 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14582 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14583 * recog.c (offsettable_address_addr_space_p): Likewise.
14584 * regcprop.c (maybe_mode_change): Likewise.
14585 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14586 * regrename.c (build_def_use): Likewise.
14587 * regstat.c (dump_reg_info): Likewise.
14588 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14589 (find_reloads, find_reloads_subreg_address): Likewise.
14590 * reload1.c (eliminate_regs_1): Likewise.
14591 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14592 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14593 (simplify_binary_operation_1, simplify_subreg): Likewise.
14594 * targhooks.c (default_function_arg_padding): Likewise.
14595 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14596 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14597 (verify_gimple_assign_ternary): Likewise.
14598 * tree-inline.c (estimate_move_cost): Likewise.
14599 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14600 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14601 (get_address_cost_ainc): Likewise.
14602 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14603 (vect_supportable_dr_alignment): Likewise.
14604 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14605 (vectorizable_reduction): Likewise.
14606 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14607 (vectorizable_operation, vectorizable_load): Likewise.
14608 * tree.c (build_same_sized_truth_vector_type): Likewise.
14609 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14610 * var-tracking.c (emit_note_insn_var_location): Likewise.
14611 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14612 (ADDR_VEC_ALIGN): Likewise.
14614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14615 Alan Hayward <alan.hayward@arm.com>
14616 David Sherwood <david.sherwood@arm.com>
14618 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14620 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14621 or if measurement_type is polynomial.
14622 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14623 * combine.c (make_extraction): Likewise.
14624 * dse.c (find_shift_sequence): Likewise.
14625 * dwarf2out.c (mem_loc_descriptor): Likewise.
14626 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14627 (extract_bit_field, extract_low_bits): Likewise.
14628 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14629 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14630 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14631 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14632 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14633 * reload.c (find_reloads): Likewise.
14634 * reload1.c (alter_reg): Likewise.
14635 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14636 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14637 * tree-if-conv.c (predicate_mem_writes): Likewise.
14638 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14639 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14640 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14641 * valtrack.c (dead_debug_insert_temp): Likewise.
14642 * varasm.c (mergeable_constant_section): Likewise.
14643 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14646 Alan Hayward <alan.hayward@arm.com>
14647 David Sherwood <david.sherwood@arm.com>
14649 * expr.c (expand_assignment): Cope with polynomial mode sizes
14650 when assigning to a CONCAT.
14652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14653 Alan Hayward <alan.hayward@arm.com>
14654 David Sherwood <david.sherwood@arm.com>
14656 * machmode.h (mode_precision): Change from unsigned short to
14658 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14660 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14661 or if measurement_type is not polynomial.
14662 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14663 in which the mode is already known to be a scalar_int_mode.
14664 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14665 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14667 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14668 for GET_MODE_PRECISION.
14669 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14670 for GET_MODE_PRECISION.
14671 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14673 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14674 (expand_field_assignment, make_extraction): Likewise.
14675 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14676 (get_last_value): Likewise.
14677 * convert.c (convert_to_integer_1): Likewise.
14678 * cse.c (cse_insn): Likewise.
14679 * expr.c (expand_expr_real_1): Likewise.
14680 * lra-constraints.c (simplify_operand_subreg): Likewise.
14681 * optabs-query.c (can_atomic_load_p): Likewise.
14682 * optabs.c (expand_atomic_load): Likewise.
14683 (expand_atomic_store): Likewise.
14684 * ree.c (combine_reaching_defs): Likewise.
14685 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14686 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14687 * tree.h (type_has_mode_precision_p): Likewise.
14688 * ubsan.c (instrument_si_overflow): Likewise.
14690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14691 Alan Hayward <alan.hayward@arm.com>
14692 David Sherwood <david.sherwood@arm.com>
14694 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14695 polynomial numbers of units.
14696 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14697 (valid_vector_subparts_p): New function.
14698 (build_vector_type): Remove temporary shim and take the number
14699 of units as a poly_uint64 rather than an int.
14700 (build_opaque_vector_type): Take the number of units as a
14701 poly_uint64 rather than an int.
14702 * tree.c (build_vector_from_ctor): Handle polynomial
14703 TYPE_VECTOR_SUBPARTS.
14704 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14705 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14706 (build_vector_from_val): If the number of units is variable,
14707 use build_vec_duplicate_cst for constant operands and
14708 VEC_DUPLICATE_EXPR otherwise.
14709 (make_vector_type): Remove temporary is_constant ().
14710 (build_vector_type, build_opaque_vector_type): Take the number of
14711 units as a poly_uint64 rather than an int.
14712 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14714 * cfgexpand.c (expand_debug_expr): Likewise.
14715 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14716 (store_constructor, expand_expr_real_1): Likewise.
14717 (const_scalar_mask_from_tree): Likewise.
14718 * fold-const-call.c (fold_const_reduction): Likewise.
14719 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14720 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14721 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14722 (fold_relational_const): Likewise.
14723 (native_interpret_vector): Likewise. Change the size from an
14724 int to an unsigned int.
14725 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14726 TYPE_VECTOR_SUBPARTS.
14727 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14728 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14729 duplicating a non-constant operand into a variable-length vector.
14730 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14731 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14732 * ipa-icf.c (sem_variable::equals): Likewise.
14733 * match.pd: Likewise.
14734 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14735 * print-tree.c (print_node): Likewise.
14736 * stor-layout.c (layout_type): Likewise.
14737 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14738 * tree-cfg.c (verify_gimple_comparison): Likewise.
14739 (verify_gimple_assign_binary): Likewise.
14740 (verify_gimple_assign_ternary): Likewise.
14741 (verify_gimple_assign_single): Likewise.
14742 * tree-pretty-print.c (dump_generic_node): Likewise.
14743 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14744 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14745 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14746 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14747 (vect_shift_permute_load_chain): Likewise.
14748 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14749 (expand_vector_condition, optimize_vector_constructor): Likewise.
14750 (lower_vec_perm, get_compute_type): Likewise.
14751 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14752 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14753 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14754 (vect_recog_mask_conversion_pattern): Likewise.
14755 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14756 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14757 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14758 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14759 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14760 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14761 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14762 (supportable_widening_operation): Likewise.
14763 (supportable_narrowing_operation): Likewise.
14764 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14766 * varasm.c (output_constant): Likewise.
14768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14769 Alan Hayward <alan.hayward@arm.com>
14770 David Sherwood <david.sherwood@arm.com>
14772 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14773 so that both the length == 3 and length != 3 cases set up their
14774 own permute vectors. Add comments explaining why we know the
14775 number of elements is constant.
14776 (vect_permute_load_chain): Likewise.
14778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14779 Alan Hayward <alan.hayward@arm.com>
14780 David Sherwood <david.sherwood@arm.com>
14782 * machmode.h (mode_nunits): Change from unsigned char to
14784 (ONLY_FIXED_SIZE_MODES): New macro.
14785 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14786 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14787 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14789 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14790 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14791 or if measurement_type is not polynomial.
14792 * genmodes.c (ZERO_COEFFS): New macro.
14793 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14795 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14796 Use ZERO_COEFFS when emitting initializers.
14797 * data-streamer.h (bp_pack_poly_value): New function.
14798 (bp_unpack_poly_value): Likewise.
14799 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14800 for GET_MODE_NUNITS.
14801 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14802 for GET_MODE_NUNITS.
14803 * tree.c (make_vector_type): Remove temporary shim and make
14804 the real function take the number of units as a poly_uint64
14805 rather than an int.
14806 (build_vector_type_for_mode): Handle polynomial nunits.
14807 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14808 * emit-rtl.c (const_vec_series_p_1): Likewise.
14809 (gen_rtx_CONST_VECTOR): Likewise.
14810 * fold-const.c (test_vec_duplicate_folding): Likewise.
14811 * genrecog.c (validate_pattern): Likewise.
14812 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14813 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14814 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14815 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14816 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14817 * rtlanal.c (subreg_get_info): Likewise.
14818 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14819 (vect_grouped_load_supported): Likewise.
14820 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14821 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14822 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14823 (simplify_const_unary_operation, simplify_binary_operation_1)
14824 (simplify_const_binary_operation, simplify_ternary_operation)
14825 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14826 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14827 instead of CONST_VECTOR_NUNITS.
14828 * varasm.c (output_constant_pool_2): Likewise.
14829 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14830 explicit-encoded elements in the XVEC for variable-length vectors.
14832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14834 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14837 Alan Hayward <alan.hayward@arm.com>
14838 David Sherwood <david.sherwood@arm.com>
14840 * coretypes.h (fixed_size_mode): Declare.
14841 (fixed_size_mode_pod): New typedef.
14842 * builtins.h (target_builtins::x_apply_args_mode)
14843 (target_builtins::x_apply_result_mode): Change type to
14844 fixed_size_mode_pod.
14845 * builtins.c (apply_args_size, apply_result_size, result_vector)
14846 (expand_builtin_apply_args_1, expand_builtin_apply)
14847 (expand_builtin_return): Update accordingly.
14849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14851 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14852 * cselib.c (cselib_hash_rtx): Likewise.
14853 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14854 CONST_VECTOR encoding.
14856 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14857 Jeff Law <law@redhat.com>
14860 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14861 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14862 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14863 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14866 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14867 explicitly probe *sp in a noreturn function if there were any callee
14868 register saves or frame pointer is needed.
14870 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14873 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14874 BLKmode for ternary, binary or unary expressions.
14877 * var-tracking.c (delete_vta_debug_insn): New inline function.
14878 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14879 insns from get_insns () to NULL instead of each bb separately.
14880 Use delete_vta_debug_insn. No longer static.
14881 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14882 delete_vta_debug_insns callers.
14883 * rtl.h (delete_vta_debug_insns): Declare.
14884 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14885 instead of variable_tracking_main.
14887 2018-01-03 Martin Sebor <msebor@redhat.com>
14889 PR tree-optimization/83603
14890 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14891 arguments past the endof the argument list in functions declared
14892 without a prototype.
14893 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14894 Avoid checking when arguments are null.
14896 2018-01-03 Martin Sebor <msebor@redhat.com>
14899 * doc/extend.texi (attribute const): Fix a typo.
14900 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14901 issuing -Wsuggest-attribute for void functions.
14903 2018-01-03 Martin Sebor <msebor@redhat.com>
14905 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14906 offset_int::from instead of wide_int::to_shwi.
14907 (maybe_diag_overlap): Remove assertion.
14908 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14909 * gimple-ssa-sprintf.c (format_directive): Same.
14910 (parse_directive): Same.
14911 (sprintf_dom_walker::compute_format_length): Same.
14912 (try_substitute_return_value): Same.
14914 2018-01-03 Jeff Law <law@redhat.com>
14916 PR middle-end/83654
14917 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14918 non-constant residual for zero at runtime and avoid probing in
14919 that case. Reorganize code for trailing problem to mirror handling
14922 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14924 PR tree-optimization/83501
14925 * tree-ssa-strlen.c (get_string_cst): New.
14926 (handle_char_store): Call get_string_cst.
14928 2018-01-03 Martin Liska <mliska@suse.cz>
14930 PR tree-optimization/83593
14931 * tree-ssa-strlen.c: Include tree-cfg.h.
14932 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14933 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14934 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14936 (strlen_dom_walker::before_dom_children): Call
14937 gimple_purge_dead_eh_edges. Dump tranformation with details
14939 (strlen_dom_walker::before_dom_children): Update call by adding
14940 new argument cleanup_eh.
14941 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14943 2018-01-03 Martin Liska <mliska@suse.cz>
14946 * cif-code.def (VARIADIC_THUNK): New enum value.
14947 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14950 2018-01-03 Jan Beulich <jbeulich@suse.com>
14952 * sse.md (mov<mode>_internal): Tighten condition for when to use
14953 vmovdqu<ssescalarsize> for TI and OI modes.
14955 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14957 Update copyright years.
14959 2018-01-03 Martin Liska <mliska@suse.cz>
14962 * ipa-visibility.c (function_and_variable_visibility): Skip
14963 functions with noipa attribure.
14965 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14967 * gcc.c (process_command): Update copyright notice dates.
14968 * gcov-dump.c (print_version): Ditto.
14969 * gcov.c (print_version): Ditto.
14970 * gcov-tool.c (print_version): Ditto.
14971 * gengtype.c (create_file): Ditto.
14972 * doc/cpp.texi: Bump @copying's copyright year.
14973 * doc/cppinternals.texi: Ditto.
14974 * doc/gcc.texi: Ditto.
14975 * doc/gccint.texi: Ditto.
14976 * doc/gcov.texi: Ditto.
14977 * doc/install.texi: Ditto.
14978 * doc/invoke.texi: Ditto.
14980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14982 * vector-builder.h (vector_builder::m_full_nelts): Change from
14983 unsigned int to poly_uint64.
14984 (vector_builder::full_nelts): Update prototype accordingly.
14985 (vector_builder::new_vector): Likewise.
14986 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14987 (vector_builder::operator ==): Likewise.
14988 (vector_builder::finalize): Likewise.
14989 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14990 Take the number of elements as a poly_uint64 rather than an
14992 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14993 from unsigned int to poly_uint64.
14994 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14995 (vec_perm_indices::new_vector): Likewise.
14996 (vec_perm_indices::length): Likewise.
14997 (vec_perm_indices::nelts_per_input): Likewise.
14998 (vec_perm_indices::input_nelts): Likewise.
14999 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15000 number of elements per input as a poly_uint64 rather than an
15001 unsigned int. Use the original encoding for variable-length
15002 vectors, rather than clamping each individual element.
15003 For the second and subsequent elements in each pattern,
15004 clamp the step and base before clamping their sum.
15005 (vec_perm_indices::series_p): Handle polynomial element counts.
15006 (vec_perm_indices::all_in_range_p): Likewise.
15007 (vec_perm_indices_to_tree): Likewise.
15008 (vec_perm_indices_to_rtx): Likewise.
15009 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15010 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15011 (tree_vector_builder::new_binary_operation): Handle polynomial
15012 element counts. Return false if we need to know the number
15013 of elements at compile time.
15014 * fold-const.c (fold_vec_perm): Punt if the number of elements
15015 isn't known at compile time.
15017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15019 * vec-perm-indices.h (vec_perm_builder): Change element type
15020 from HOST_WIDE_INT to poly_int64.
15021 (vec_perm_indices::element_type): Update accordingly.
15022 (vec_perm_indices::clamp): Handle polynomial element_types.
15023 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15024 (vec_perm_indices::all_in_range_p): Likewise.
15025 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15027 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15028 polynomial vec_perm_indices element types.
15029 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15030 * fold-const.c (fold_vec_perm): Likewise.
15031 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15032 * tree-vect-generic.c (lower_vec_perm): Likewise.
15033 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15034 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15035 element type to HOST_WIDE_INT.
15037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15038 Alan Hayward <alan.hayward@arm.com>
15039 David Sherwood <david.sherwood@arm.com>
15041 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15042 rather than an int. Use plus_constant.
15043 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15044 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15047 Alan Hayward <alan.hayward@arm.com>
15048 David Sherwood <david.sherwood@arm.com>
15050 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15051 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
15058 mode sizes. Require a constant size for BLKmode parameters
15059 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15060 forces a parameter to be padded at the lsb end in order to
15061 fill a complete number of words, require the parameter size
15062 to be ordered wrt UNITS_PER_WORD.
15064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15065 Alan Hayward <alan.hayward@arm.com>
15066 David Sherwood <david.sherwood@arm.com>
15068 * reload1.c (spill_stack_slot_width): Change element type
15069 from unsigned int to poly_uint64_pod.
15070 (alter_reg): Treat mode sizes as polynomial.
15072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15073 Alan Hayward <alan.hayward@arm.com>
15074 David Sherwood <david.sherwood@arm.com>
15076 * reload.c (complex_word_subreg_p): New function.
15077 (reload_inner_reg_of_subreg, push_reload): Use it.
15079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15080 Alan Hayward <alan.hayward@arm.com>
15081 David Sherwood <david.sherwood@arm.com>
15083 * lra-constraints.c (process_alt_operands): Reject matched
15084 operands whose sizes aren't ordered.
15085 (match_reload): Refer to this check here.
15087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15088 Alan Hayward <alan.hayward@arm.com>
15089 David Sherwood <david.sherwood@arm.com>
15091 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15092 that the mode size is in the set {1, 2, 4, 8, 16}.
15094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15095 Alan Hayward <alan.hayward@arm.com>
15096 David Sherwood <david.sherwood@arm.com>
15098 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15099 Use plus_constant instead of gen_rtx_PLUS.
15101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15102 Alan Hayward <alan.hayward@arm.com>
15103 David Sherwood <david.sherwood@arm.com>
15105 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15106 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15107 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15108 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15109 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15110 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15111 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15112 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15113 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15114 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15116 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15117 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15118 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15119 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15120 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15121 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15122 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15123 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15124 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15125 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15127 * expr.c (emit_move_resolve_push): Treat the input and result
15128 of PUSH_ROUNDING as a poly_int64.
15129 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15130 (emit_push_insn): Likewise.
15131 * lra-eliminations.c (mark_not_eliminable): Likewise.
15132 * recog.c (push_operand): Likewise.
15133 * reload1.c (elimination_effects): Likewise.
15134 * rtlanal.c (nonzero_bits1): Likewise.
15135 * calls.c (store_one_arg): Likewise. Require the padding to be
15136 known at compile time.
15138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15139 Alan Hayward <alan.hayward@arm.com>
15140 David Sherwood <david.sherwood@arm.com>
15142 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15143 Use plus_constant instead of gen_rtx_PLUS.
15145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15146 Alan Hayward <alan.hayward@arm.com>
15147 David Sherwood <david.sherwood@arm.com>
15149 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15150 rather than an int.
15152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15153 Alan Hayward <alan.hayward@arm.com>
15154 David Sherwood <david.sherwood@arm.com>
15156 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15157 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15158 via stack temporaries. Treat the mode size as polynomial too.
15160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15161 Alan Hayward <alan.hayward@arm.com>
15162 David Sherwood <david.sherwood@arm.com>
15164 * expr.c (expand_expr_real_2): When handling conversions involving
15165 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15166 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15167 as a poly_uint64 too.
15169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15170 Alan Hayward <alan.hayward@arm.com>
15171 David Sherwood <david.sherwood@arm.com>
15173 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15176 Alan Hayward <alan.hayward@arm.com>
15177 David Sherwood <david.sherwood@arm.com>
15179 * combine.c (can_change_dest_mode): Handle polynomial
15180 REGMODE_NATURAL_SIZE.
15181 * expmed.c (store_bit_field_1): Likewise.
15182 * expr.c (store_constructor): Likewise.
15183 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15184 and polynomial REGMODE_NATURAL_SIZE.
15185 (gen_lowpart_common): Likewise.
15186 * reginfo.c (record_subregs_of_mode): Likewise.
15187 * rtlanal.c (read_modify_subreg_p): Likewise.
15189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15190 Alan Hayward <alan.hayward@arm.com>
15191 David Sherwood <david.sherwood@arm.com>
15193 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15194 numbers of elements.
15196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15197 Alan Hayward <alan.hayward@arm.com>
15198 David Sherwood <david.sherwood@arm.com>
15200 * match.pd: Cope with polynomial numbers of vector elements.
15202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15203 Alan Hayward <alan.hayward@arm.com>
15204 David Sherwood <david.sherwood@arm.com>
15206 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15207 in a POINTER_PLUS_EXPR.
15209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15210 Alan Hayward <alan.hayward@arm.com>
15211 David Sherwood <david.sherwood@arm.com>
15213 * omp-simd-clone.c (simd_clone_subparts): New function.
15214 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15215 (ipa_simd_modify_function_body): Likewise.
15217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15218 Alan Hayward <alan.hayward@arm.com>
15219 David Sherwood <david.sherwood@arm.com>
15221 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15222 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15223 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15224 (expand_vector_condition, vector_element): Likewise.
15225 (subparts_gt): New function.
15226 (get_compute_type): Use subparts_gt.
15227 (count_type_subparts): Delete.
15228 (expand_vector_operations_1): Use subparts_gt instead of
15229 count_type_subparts.
15231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15232 Alan Hayward <alan.hayward@arm.com>
15233 David Sherwood <david.sherwood@arm.com>
15235 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15236 (vect_compile_time_alias): ...this new function. Do the calculation
15237 on poly_ints rather than trees.
15238 (vect_prune_runtime_alias_test_list): Update call accordingly.
15240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15241 Alan Hayward <alan.hayward@arm.com>
15242 David Sherwood <david.sherwood@arm.com>
15244 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15246 (vect_schedule_slp_instance): Likewise.
15248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15249 Alan Hayward <alan.hayward@arm.com>
15250 David Sherwood <david.sherwood@arm.com>
15252 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15253 constant and extern definitions for variable-length vectors.
15254 (vect_get_constant_vectors): Note that the number of units
15255 is known to be constant.
15257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15258 Alan Hayward <alan.hayward@arm.com>
15259 David Sherwood <david.sherwood@arm.com>
15261 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15262 of units as polynomial. Choose between WIDE and NARROW based
15265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15266 Alan Hayward <alan.hayward@arm.com>
15267 David Sherwood <david.sherwood@arm.com>
15269 * tree-vect-stmts.c (simd_clone_subparts): New function.
15270 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15273 Alan Hayward <alan.hayward@arm.com>
15274 David Sherwood <david.sherwood@arm.com>
15276 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15277 vectors as polynomial. Use build_index_vector for
15278 IFN_GOMP_SIMD_LANE.
15280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15281 Alan Hayward <alan.hayward@arm.com>
15282 David Sherwood <david.sherwood@arm.com>
15284 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15285 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15286 for variable-length vectors.
15287 (vectorizable_mask_load_store): Treat the number of units as
15288 polynomial, asserting that it is constant if the condition has
15289 already been enforced.
15290 (vectorizable_store, vectorizable_load): Likewise.
15292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15293 Alan Hayward <alan.hayward@arm.com>
15294 David Sherwood <david.sherwood@arm.com>
15296 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15297 of units as polynomial. Punt if we can't tell at compile time
15298 which vector contains the final result.
15300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15301 Alan Hayward <alan.hayward@arm.com>
15302 David Sherwood <david.sherwood@arm.com>
15304 * tree-vect-loop.c (vectorizable_induction): Treat the number
15305 of units as polynomial. Punt on SLP inductions. Use an integer
15306 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15307 cast of such a series for variable-length floating-point
15310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15311 Alan Hayward <alan.hayward@arm.com>
15312 David Sherwood <david.sherwood@arm.com>
15314 * tree.h (build_index_vector): Declare.
15315 * tree.c (build_index_vector): New function.
15316 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15317 of units as polynomial, forcibly converting it to a constant if
15318 vectorizable_reduction has already enforced the condition.
15319 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15320 to create a {1,2,3,...} vector.
15321 (vectorizable_reduction): Treat the number of units as polynomial.
15322 Choose vectype_in based on the largest scalar element size rather
15323 than the smallest number of units. Enforce the restrictions
15326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15327 Alan Hayward <alan.hayward@arm.com>
15328 David Sherwood <david.sherwood@arm.com>
15330 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15331 number of units as polynomial.
15333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15334 Alan Hayward <alan.hayward@arm.com>
15335 David Sherwood <david.sherwood@arm.com>
15337 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15338 * target.def (autovectorize_vector_sizes): Return the vector sizes
15339 by pointer, using vector_sizes rather than a bitmask.
15340 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15341 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15342 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15344 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15345 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15346 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15347 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15348 * omp-general.c (omp_max_vf): Likewise.
15349 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15350 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15351 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15352 * tree-vect-slp.c (vect_slp_bb): Likewise.
15353 * doc/tm.texi: Regenerate.
15354 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15356 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15357 the vector size as a poly_uint64 rather than an unsigned int.
15358 (current_vector_size): Change from an unsigned int to a poly_uint64.
15359 (get_vectype_for_scalar_type): Update accordingly.
15360 * tree.h (build_truth_vector_type): Take the size and number of
15361 units as a poly_uint64 rather than an unsigned int.
15362 (build_vector_type): Add a temporary overload that takes
15363 the number of units as a poly_uint64 rather than an unsigned int.
15364 * tree.c (make_vector_type): Likewise.
15365 (build_truth_vector_type): Take the number of units as a poly_uint64
15366 rather than an unsigned int.
15368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15369 Alan Hayward <alan.hayward@arm.com>
15370 David Sherwood <david.sherwood@arm.com>
15372 * target.def (get_mask_mode): Take the number of units and length
15373 as poly_uint64s rather than unsigned ints.
15374 * targhooks.h (default_get_mask_mode): Update accordingly.
15375 * targhooks.c (default_get_mask_mode): Likewise.
15376 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15377 * doc/tm.texi: Regenerate.
15379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15380 Alan Hayward <alan.hayward@arm.com>
15381 David Sherwood <david.sherwood@arm.com>
15383 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15384 * omp-general.c (omp_max_vf): Likewise.
15385 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15386 (expand_omp_simd): Handle polynomial safelen.
15387 * omp-low.c (omplow_simd_context): Add a default constructor.
15388 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15389 (lower_rec_simd_input_clauses): Update accordingly.
15390 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15397 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15398 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15399 (vect_analyze_slp_cost): Likewise.
15400 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15401 (vect_model_load_cost): Likewise.
15403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15404 Alan Hayward <alan.hayward@arm.com>
15405 David Sherwood <david.sherwood@arm.com>
15407 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15408 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15409 from an unsigned int * to a poly_uint64_pod *.
15410 (calculate_unrolling_factor): New function.
15411 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15418 from an unsigned int to a poly_uint64.
15419 (_loop_vec_info::slp_unrolling_factor): Likewise.
15420 (_loop_vec_info::vectorization_factor): Change from an int
15422 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15423 (vect_get_num_vectors): New function.
15424 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15425 (vect_get_num_copies): Use vect_get_num_vectors.
15426 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15427 to an unsigned int *.
15428 (vect_analyze_data_refs): Change min_vf from an int * to a
15430 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15431 than an unsigned HOST_WIDE_INT.
15432 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15433 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15434 to an unsigned int *.
15435 (vect_analyze_data_ref_dependences): Likewise.
15436 (vect_compute_data_ref_alignment): Handle polynomial vf.
15437 (vect_enhance_data_refs_alignment): Likewise.
15438 (vect_prune_runtime_alias_test_list): Likewise.
15439 (vect_shift_permute_load_chain): Likewise.
15440 (vect_supportable_dr_alignment): Likewise.
15441 (dependence_distance_ge_vf): Take the vectorization factor as a
15442 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15443 (vect_analyze_data_refs): Change min_vf from an int * to a
15445 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15446 vfm1 as a poly_uint64 rather than an int. Make the same change
15447 for the returned bound_scalar.
15448 (vect_gen_vector_loop_niters): Handle polynomial vf.
15449 (vect_do_peeling): Likewise. Update call to
15450 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15451 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15453 * tree-vect-loop.c (vect_determine_vectorization_factor)
15454 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15455 (vect_get_known_peeling_cost): Likewise.
15456 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15457 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15458 (vect_transform_loop): Likewise. Use the lowest possible VF when
15459 updating the upper bounds of the loop.
15460 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15461 rather than an int.
15462 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15463 polynomial unroll factors.
15464 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15465 (vect_make_slp_decision): Likewise.
15466 (vect_supported_load_permutation_p): Likewise, and polynomial
15468 (vect_analyze_slp_cost): Handle polynomial vf.
15469 (vect_slp_analyze_node_operations): Likewise.
15470 (vect_slp_analyze_bb_1): Likewise.
15471 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15472 than an unsigned HOST_WIDE_INT.
15473 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15474 (vectorizable_load): Handle polynomial vf.
15475 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15477 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15480 Alan Hayward <alan.hayward@arm.com>
15481 David Sherwood <david.sherwood@arm.com>
15483 * match.pd: Handle bit operations involving three constants
15484 and try to fold one pair.
15486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15488 * tree-vect-loop-manip.c: Include gimple-fold.h.
15489 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15490 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15491 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15492 Add a path that uses a step of VF instead of 1, but disable it
15494 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15495 and niters_no_overflow parameters. Update calls to
15496 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15497 Create a new SSA name if the latter choses to use a ste other
15498 than zero, and return it via niters_vector_mult_vf_var.
15499 * tree-vect-loop.c (vect_transform_loop): Update calls to
15500 vect_do_peeling, vect_gen_vector_loop_niters and
15501 slpeel_make_loop_iterate_ntimes.
15502 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15503 (vect_gen_vector_loop_niters): Update declarations after above changes.
15505 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15507 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15508 128-bit round to integer instructions.
15509 (ceil<mode>2): Likewise.
15510 (btrunc<mode>2): Likewise.
15511 (round<mode>2): Likewise.
15513 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15515 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15516 unaligned VSX load/store on P8/P9.
15517 (expand_block_clear): Allow the use of unaligned VSX
15518 load/store on P8/P9.
15520 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15522 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15524 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15525 swap associated with both a load and a store.
15527 2018-01-02 Andrew Waterman <andrew@sifive.com>
15529 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15530 * config/riscv/riscv.md (clear_cache): Use it.
15532 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15534 * web.c: Remove out-of-date comment.
15536 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15538 * expr.c (fixup_args_size_notes): Check that any existing
15539 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15540 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15541 (emit_single_push_insn): ...here.
15543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15545 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15546 (const_vector_encoded_nelts): New function.
15547 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15548 (const_vector_int_elt, const_vector_elt): Declare.
15549 * emit-rtl.c (const_vector_int_elt_1): New function.
15550 (const_vector_elt): Likewise.
15551 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15552 of CONST_VECTOR_ELT.
15554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15556 * expr.c: Include rtx-vector-builder.h.
15557 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15558 directly on the tree encoding.
15559 (const_vector_from_tree): Likewise.
15560 * optabs.c: Include rtx-vector-builder.h.
15561 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15562 sequence of "u" values.
15563 * vec-perm-indices.c: Include rtx-vector-builder.h.
15564 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15565 directly on the vec_perm_indices encoding.
15567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15569 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15570 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15571 * rtx-vector-builder.h: New file.
15572 * rtx-vector-builder.c: Likewise.
15573 * rtl.h (rtx_def::u2): Add a const_vector field.
15574 (CONST_VECTOR_NPATTERNS): New macro.
15575 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15576 (CONST_VECTOR_DUPLICATE_P): Likewise.
15577 (CONST_VECTOR_STEPPED_P): Likewise.
15578 (CONST_VECTOR_ENCODED_ELT): Likewise.
15579 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15580 (unwrap_const_vec_duplicate): Likewise.
15581 (const_vec_series_p): Check for a non-duplicated vector encoding.
15582 Say that the function only returns true for integer vectors.
15583 * emit-rtl.c: Include rtx-vector-builder.h.
15584 (gen_const_vec_duplicate_1): Delete.
15585 (gen_const_vector): Call gen_const_vec_duplicate instead of
15586 gen_const_vec_duplicate_1.
15587 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15588 (gen_const_vec_duplicate): Use rtx_vector_builder.
15589 (gen_const_vec_series): Likewise.
15590 (gen_rtx_CONST_VECTOR): Likewise.
15591 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15592 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15593 Build a new vector rather than modifying a CONST_VECTOR in-place.
15594 (handle_special_swappables): Update call accordingly.
15595 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15596 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15597 Build a new vector rather than modifying a CONST_VECTOR in-place.
15598 (handle_special_swappables): Update call accordingly.
15600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15602 * simplify-rtx.c (simplify_const_binary_operation): Use
15603 CONST_VECTOR_ELT instead of XVECEXP.
15605 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15607 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15608 the selector elements to be different from the data elements
15609 if the selector is a VECTOR_CST.
15610 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15611 ssizetype for the selector.
15613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15615 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15616 before testing each element individually.
15617 * tree-vect-generic.c (lower_vec_perm): Likewise.
15619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15621 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15622 * selftest-run-tests.c (selftest::run_tests): Call it.
15623 * vector-builder.h (vector_builder::operator ==): New function.
15624 (vector_builder::operator !=): Likewise.
15625 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15626 (vec_perm_indices::all_from_input_p): New function.
15627 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15628 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15629 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15630 instead of reading the VECTOR_CST directly. Detect whether both
15631 vector inputs are the same before constructing the vec_perm_indices,
15632 and update the number of inputs argument accordingly. Use the
15633 utility functions added above. Only construct sel2 if we need to.
15635 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15637 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15638 the broadcast of the low byte.
15639 (expand_mult_highpart): Use an explicit encoding for the permutes.
15640 * optabs-query.c (can_mult_highpart_p): Likewise.
15641 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15642 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15643 (vectorizable_bswap): Likewise.
15644 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15645 explicit encoding for the power-of-2 permutes.
15646 (vect_permute_store_chain): Likewise.
15647 (vect_grouped_load_supported): Likewise.
15648 (vect_permute_load_chain): Likewise.
15650 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15652 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15653 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15654 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15655 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15656 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15657 (vect_gen_perm_mask_any): Likewise.
15659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15661 * int-vector-builder.h: New file.
15662 * vec-perm-indices.h: Include int-vector-builder.h.
15663 (vec_perm_indices): Redefine as an int_vector_builder.
15664 (auto_vec_perm_indices): Delete.
15665 (vec_perm_builder): Redefine as a stand-alone class.
15666 (vec_perm_indices::vec_perm_indices): New function.
15667 (vec_perm_indices::clamp): Likewise.
15668 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15669 (vec_perm_indices::new_vector): New function.
15670 (vec_perm_indices::new_expanded_vector): Update for new
15671 vec_perm_indices class.
15672 (vec_perm_indices::rotate_inputs): New function.
15673 (vec_perm_indices::all_in_range_p): Operate directly on the
15674 encoded form, without computing elided elements.
15675 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15676 encoding. Update for new vec_perm_indices class.
15677 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15678 the given vec_perm_builder.
15679 (expand_vec_perm_var): Update vec_perm_builder constructor.
15680 (expand_mult_highpart): Use vec_perm_builder instead of
15681 auto_vec_perm_indices.
15682 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15683 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15684 or double series encoding as appropriate.
15685 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15686 vec_perm_indices instead of auto_vec_perm_indices.
15687 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15688 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15689 (vect_permute_store_chain): Likewise.
15690 (vect_grouped_load_supported): Likewise.
15691 (vect_permute_load_chain): Likewise.
15692 (vect_shift_permute_load_chain): Likewise.
15693 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15694 (vect_transform_slp_perm_load): Likewise.
15695 (vect_schedule_slp_instance): Likewise.
15696 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15697 (vectorizable_mask_load_store): Likewise.
15698 (vectorizable_bswap): Likewise.
15699 (vectorizable_store): Likewise.
15700 (vectorizable_load): Likewise.
15701 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15702 vec_perm_indices instead of auto_vec_perm_indices. Use
15703 tree_to_vec_perm_builder to read the vector from a tree.
15704 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15705 vec_perm_builder instead of a vec_perm_indices.
15706 (have_whole_vector_shift): Use vec_perm_builder and
15707 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15708 truncation to calc_vec_perm_mask_for_shift.
15709 (vect_create_epilog_for_reduction): Likewise.
15710 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15711 from auto_vec_perm_indices to vec_perm_indices.
15712 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15713 instead of changing individual elements.
15714 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15715 the vector in d.perm.
15716 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15717 from auto_vec_perm_indices to vec_perm_indices.
15718 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15719 instead of changing individual elements.
15720 (arm_vectorize_vec_perm_const): Use new_vector to install
15721 the vector in d.perm.
15722 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15723 Update vec_perm_builder constructor.
15724 (rs6000_expand_interleave): Likewise.
15725 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15726 (rs6000_expand_interleave): Likewise.
15728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15730 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15731 to qimode could truncate the indices.
15732 * optabs.c (expand_vec_perm_var): Likewise.
15734 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15736 * Makefile.in (OBJS): Add vec-perm-indices.o.
15737 * vec-perm-indices.h: New file.
15738 * vec-perm-indices.c: Likewise.
15739 * target.h (vec_perm_indices): Replace with a forward class
15741 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15742 * optabs.h: Include vec-perm-indices.h.
15743 (expand_vec_perm): Delete.
15744 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15745 (expand_vec_perm_const): Declare.
15746 * target.def (vec_perm_const_ok): Replace with...
15747 (vec_perm_const): ...this new hook.
15748 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15749 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15750 * doc/tm.texi: Regenerate.
15751 * optabs.def (vec_perm_const): Delete.
15752 * doc/md.texi (vec_perm_const): Likewise.
15753 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15754 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15755 expand_vec_perm for constant permutation vectors. Assert that
15756 the mode of variable permutation vectors is the integer equivalent
15757 of the mode that is being permuted.
15758 * optabs-query.h (selector_fits_mode_p): Declare.
15759 * optabs-query.c: Include vec-perm-indices.h.
15760 (selector_fits_mode_p): New function.
15761 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15762 is defined, instead of checking whether the vec_perm_const_optab
15763 exists. Use targetm.vectorize.vec_perm_const instead of
15764 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15765 fit in the vector mode before using a variable permute.
15766 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15767 vec_perm_indices instead of an rtx.
15768 (expand_vec_perm): Replace with...
15769 (expand_vec_perm_const): ...this new function. Take the selector
15770 as a vec_perm_indices rather than an rtx. Also take the mode of
15771 the selector. Update call to shift_amt_for_vec_perm_mask.
15772 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15773 Use vec_perm_indices::new_expanded_vector to expand the original
15774 selector into bytes. Check whether the indices fit in the vector
15775 mode before using a variable permute.
15776 (expand_vec_perm_var): Make global.
15777 (expand_mult_highpart): Use expand_vec_perm_const.
15778 * fold-const.c: Includes vec-perm-indices.h.
15779 * tree-ssa-forwprop.c: Likewise.
15780 * tree-vect-data-refs.c: Likewise.
15781 * tree-vect-generic.c: Likewise.
15782 * tree-vect-loop.c: Likewise.
15783 * tree-vect-slp.c: Likewise.
15784 * tree-vect-stmts.c: Likewise.
15785 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15787 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15788 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15789 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15790 (aarch64_vectorize_vec_perm_const): ...this new function.
15791 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15792 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15793 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15794 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15795 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15796 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15797 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15799 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15800 check for NEON modes.
15801 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15802 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15803 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15804 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15806 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15807 the old VEC_PERM_CONST conditions.
15808 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15809 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15810 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15811 (ia64_vectorize_vec_perm_const_ok): Merge into...
15812 (ia64_vectorize_vec_perm_const): ...this new function.
15813 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15814 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15815 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15816 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15817 * config/mips/mips.c (mips_expand_vec_perm_const)
15818 (mips_vectorize_vec_perm_const_ok): Merge into...
15819 (mips_vectorize_vec_perm_const): ...this new function.
15820 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15821 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15822 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15823 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15824 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15825 (rs6000_expand_vec_perm_const): Delete.
15826 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15828 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15829 (altivec_expand_vec_perm_const_le): Take each operand individually.
15830 Operate on constant selectors rather than rtxes.
15831 (altivec_expand_vec_perm_const): Likewise. Update call to
15832 altivec_expand_vec_perm_const_le.
15833 (rs6000_expand_vec_perm_const): Delete.
15834 (rs6000_vectorize_vec_perm_const_ok): Delete.
15835 (rs6000_vectorize_vec_perm_const): New function.
15836 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15837 an element count and rtx array.
15838 (rs6000_expand_extract_even): Update call accordingly.
15839 (rs6000_expand_interleave): Likewise.
15840 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15841 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15842 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15843 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15844 (rs6000_expand_vec_perm_const): Delete.
15845 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15846 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15847 (altivec_expand_vec_perm_const_le): Take each operand individually.
15848 Operate on constant selectors rather than rtxes.
15849 (altivec_expand_vec_perm_const): Likewise. Update call to
15850 altivec_expand_vec_perm_const_le.
15851 (rs6000_expand_vec_perm_const): Delete.
15852 (rs6000_vectorize_vec_perm_const_ok): Delete.
15853 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15854 reference to the SPE evmerge intructions.
15855 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15856 an element count and rtx array.
15857 (rs6000_expand_extract_even): Update call accordingly.
15858 (rs6000_expand_interleave): Likewise.
15859 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15860 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15862 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15866 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15867 vector mode and that that mode matches the mode of the data
15869 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15870 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15871 directly using expand_vec_perm_1 when forcing selectors into
15873 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15877 * optabs-query.h (can_vec_perm_p): Delete.
15878 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15879 * optabs-query.c (can_vec_perm_p): Split into...
15880 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15881 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15882 particular selector is valid.
15883 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15884 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15885 (vect_grouped_load_supported): Likewise.
15886 (vect_shift_permute_load_chain): Likewise.
15887 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15888 (vect_transform_slp_perm_load): Likewise.
15889 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15890 (vectorizable_bswap): Likewise.
15891 (vect_gen_perm_mask_checked): Likewise.
15892 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15893 implementations of variable permutation vectors into account
15894 when deciding which selector to use.
15895 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15896 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15897 with a false third argument.
15898 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15899 to test whether the constant selector is valid and can_vec_perm_var_p
15900 to test whether a variable selector is valid.
15902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15904 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15905 * optabs-query.c (can_vec_perm_p): Likewise.
15906 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15907 instead of vec_perm_indices.
15908 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15909 (vect_gen_perm_mask_checked): Likewise,
15910 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15911 (vect_gen_perm_mask_checked): Likewise,
15913 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15915 * optabs-query.h (qimode_for_vec_perm): Declare.
15916 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15917 (qimode_for_vec_perm): ...this new function.
15918 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15920 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15922 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15923 does not have a conditional at the top.
15925 2018-01-02 Richard Biener <rguenther@suse.de>
15927 * ipa-inline.c (big_speedup_p): Fix expression.
15929 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15932 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15935 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15939 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15940 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15941 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15942 cond_taken_branch_cost 3->4.
15944 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15946 PR tree-optimization/83581
15947 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15948 TODO_cleanup_cfg if any changes have been made.
15950 PR middle-end/83608
15951 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15952 convert_modes if target mode has the right side, but different mode
15955 PR middle-end/83609
15956 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15957 last argument when extracting from CONCAT. If either from_real or
15958 from_imag is NULL, use expansion through memory. If result is not
15959 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15960 the parts directly to inner mode, if even that fails, use expansion
15963 PR middle-end/83623
15964 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15965 check for bswap in mode rather than HImode and use that in expand_unop
15968 Copyright (C) 2018 Free Software Foundation, Inc.
15970 Copying and distribution of this file, with or without modification,
15971 are permitted in any medium without royalty provided the copyright
15972 notice and this notice are preserved.