1 2018-06-08 Martin Liska <mliska@suse.cz>
3 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5 (hsa_function_summary::hsa_function_summary): Use the new enum
7 (hsa_gpu_implementation_p): Use hsa_summaries::get.
8 * hsa-gen.c (hsa_get_host_function): Likewise.
9 (get_brig_function_name): Likewise.
10 * ipa-hsa.c (process_hsa_functions): Likewise.
11 (ipa_hsa_write_summary): Likewise.
12 * symbol-summary.h (symtab_duplication): Use ::get function/
15 2018-06-08 Martin Liska <mliska@suse.cz>
17 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
19 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
20 (hsa_register_kernel): Likewise.
21 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
22 * hsa-gen.c (hsa_get_host_function): Likewise.
23 (get_brig_function_name): Likewise.
24 (generate_hsa): Likewise.
25 (pass_gen_hsail::execute): Likewise.
26 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
27 (devirtualization_time_bonus): Likewise.
28 (ipcp_propagate_stage): Likewise.
29 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
30 (edge_set_predicate): Likewise.
31 (evaluate_conditions_for_known_args): Likewise.
32 (evaluate_properties_for_edge): Likewise.
33 (ipa_fn_summary::reset): Likewise.
34 (ipa_fn_summary_t::duplicate): Likewise.
35 (dump_ipa_call_summary): Likewise.
36 (ipa_dump_fn_summary): Likewise.
37 (analyze_function_body): Likewise.
38 (compute_fn_summary): Likewise.
39 (estimate_edge_devirt_benefit): Likewise.
40 (estimate_edge_size_and_time): Likewise.
41 (estimate_calls_size_and_time): Likewise.
42 (estimate_node_size_and_time): Likewise.
43 (inline_update_callee_summaries): Likewise.
44 (remap_edge_change_prob): Likewise.
45 (remap_edge_summaries): Likewise.
46 (ipa_merge_fn_summary_after_inlining): Likewise.
47 (ipa_update_overall_fn_summary): Likewise.
48 (read_ipa_call_summary): Likewise.
49 (inline_read_section): Likewise.
50 (write_ipa_call_summary): Likewise.
51 (ipa_fn_summary_write): Likewise.
52 (ipa_free_fn_summary): Likewise.
53 * ipa-hsa.c (process_hsa_functions): Likewise.
54 (ipa_hsa_write_summary): Likewise.
55 (ipa_hsa_read_section): Likewise.
56 * ipa-icf.c (sem_function::merge): Likewise.
57 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
58 (do_estimate_edge_time): Likewise.
59 (estimate_size_after_inlining): Likewise.
60 (estimate_growth): Likewise.
61 (growth_likely_positive): Likewise.
62 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
63 (inline_call): Likewise.
64 * ipa-inline.c (caller_growth_limits): Likewise.
65 (can_inline_edge_p): Likewise.
66 (can_inline_edge_by_limits_p): Likewise.
67 (compute_uninlined_call_time): Likewise.
68 (compute_inlined_call_time): Likewise.
69 (want_inline_small_function_p): Likewise.
70 (edge_badness): Likewise.
71 (update_caller_keys): Likewise.
72 (update_callee_keys): Likewise.
73 (recursive_inlining): Likewise.
74 (inline_small_functions): Likewise.
75 (inline_to_all_callers_1): Likewise.
76 (dump_overall_stats): Likewise.
77 (early_inline_small_functions): Likewise.
78 (early_inliner): Likewise.
79 * ipa-inline.h (estimate_edge_growth): Likewise.
80 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
81 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
82 * ipa-prop.h (IPA_NODE_REF): Likewise.
83 (IPA_EDGE_REF): Likewise.
84 * ipa-pure-const.c (malloc_candidate_p): Likewise.
85 (propagate_malloc): Likewise.
86 * ipa-split.c (execute_split_functions): Likewise.
87 * symbol-summary.h: Rename get to get_create.
89 (get_create): Likewise.
90 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
92 2018-06-08 Martin Liska <mliska@suse.cz>
94 * symbol-summary.h (release): Move definition out of class
96 (symtab_removal): Likewise.
97 (symtab_duplication): Likewise.
99 2018-06-08 Martin Liska <mliska@suse.cz>
101 * symbol-summary.h (function_summary): Move constructor
102 implementation out of class declaration.
104 (symtab_insertion): Likewise.
105 (symtab_removal): Likewise.
106 (symtab_duplication): Likewise.
109 2018-06-08 Martin Liska <mliska@suse.cz>
111 * Makefile.in: Remove support for MPX (macros, related functions,
112 fields in cgraph_node, ...).
113 * builtin-types.def (BT_BND): Likewise.
114 (BT_FN_BND_CONST_PTR): Likewise.
115 (BT_FN_CONST_PTR_BND): Likewise.
116 (BT_FN_VOID_PTR_BND): Likewise.
117 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
118 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
119 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
120 (expand_builtin_mempcpy_with_bounds): Likewise.
121 (expand_builtin_memset_with_bounds): Likewise.
122 (expand_builtin_memset_args): Likewise.
123 (std_expand_builtin_va_start): Likewise.
124 (expand_builtin): Likewise.
125 (expand_builtin_with_bounds): Likewise.
126 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
127 (DEF_LIB_BUILTIN_CHKP): Likewise.
128 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
129 (DEF_CHKP_BUILTIN): Likewise.
130 (BUILT_IN_MEMCPY): Likewise.
131 (BUILT_IN_MEMMOVE): Likewise.
132 (BUILT_IN_MEMPCPY): Likewise.
133 (BUILT_IN_MEMSET): Likewise.
134 (BUILT_IN_STPCPY): Likewise.
135 (BUILT_IN_STRCAT): Likewise.
136 (BUILT_IN_STRCHR): Likewise.
137 (BUILT_IN_STRCPY): Likewise.
138 (BUILT_IN_STRLEN): Likewise.
139 (BUILT_IN_MEMCPY_CHK): Likewise.
140 (BUILT_IN_MEMMOVE_CHK): Likewise.
141 (BUILT_IN_MEMPCPY_CHK): Likewise.
142 (BUILT_IN_MEMSET_CHK): Likewise.
143 (BUILT_IN_STPCPY_CHK): Likewise.
144 (BUILT_IN_STRCAT_CHK): Likewise.
145 (BUILT_IN_STRCPY_CHK): Likewise.
146 * calls.c (store_bounds): Likewise.
147 (emit_call_1): Likewise.
148 (special_function_p): Likewise.
149 (maybe_warn_nonstring_arg): Likewise.
150 (initialize_argument_information): Likewise.
151 (finalize_must_preallocate): Likewise.
152 (compute_argument_addresses): Likewise.
153 (expand_call): Likewise.
154 * cfgexpand.c (expand_call_stmt): Likewise.
155 (expand_return): Likewise.
156 (expand_gimple_stmt_1): Likewise.
157 (pass_expand::execute): Likewise.
158 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
159 (cgraph_node::remove): Likewise.
160 (cgraph_node::dump): Likewise.
161 (cgraph_node::verify_node): Likewise.
162 * cgraph.h (chkp_function_instrumented_p): Likewise.
163 (symtab_node::get_alias_target): Likewise.
164 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
165 (cgraph_local_p): Likewise.
166 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
167 (cgraph_edge::rebuild_references): Likewise.
168 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
169 (walk_polymorphic_call_targets): Likewise.
170 (cgraph_node::expand_thunk): Likewise.
171 (symbol_table::output_weakrefs): Likewise.
172 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
173 (ix86_handle_option): Likewise.
174 * config/i386/constraints.md: Likewise.
175 * config/i386/i386-builtin-types.def (BND): Likewise.
179 * config/i386/i386-builtin.def (BDESC_END): Likewise.
180 (BDESC_FIRST): Likewise.
182 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
183 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
184 * config/i386/i386.c (enum reg_class): Likewise.
185 (ix86_target_string): Likewise.
186 (ix86_option_override_internal): Likewise.
187 (ix86_conditional_register_usage): Likewise.
188 (ix86_valid_target_attribute_inner_p): Likewise.
189 (ix86_set_indirect_branch_type): Likewise.
190 (ix86_set_current_function): Likewise.
191 (ix86_function_arg_regno_p): Likewise.
192 (init_cumulative_args): Likewise.
193 (ix86_function_arg_advance): Likewise.
194 (ix86_function_arg): Likewise.
195 (ix86_pass_by_reference): Likewise.
196 (ix86_function_value_regno_p): Likewise.
197 (ix86_function_value_1): Likewise.
198 (ix86_function_value_bounds): Likewise.
199 (ix86_return_in_memory): Likewise.
200 (ix86_setup_incoming_vararg_bounds): Likewise.
201 (ix86_va_start): Likewise.
202 (indirect_thunk_need_prefix): Likewise.
203 (print_reg): Likewise.
204 (ix86_print_operand): Likewise.
205 (ix86_expand_call): Likewise.
206 (ix86_output_function_return): Likewise.
207 (reg_encoded_number): Likewise.
208 (BDESC_VERIFYS): Likewise.
209 (ix86_init_mpx_builtins): Likewise.
210 (ix86_init_builtins): Likewise.
211 (ix86_emit_cmove): Likewise.
212 (ix86_emit_move_max): Likewise.
213 (ix86_expand_builtin): Likewise.
214 (ix86_builtin_mpx_function): Likewise.
215 (ix86_get_arg_address_for_bt): Likewise.
216 (ix86_load_bounds): Likewise.
217 (ix86_store_bounds): Likewise.
218 (ix86_load_returned_bounds): Likewise.
219 (ix86_store_returned_bounds): Likewise.
220 (ix86_class_likely_spilled_p): Likewise.
221 (ix86_hard_regno_mode_ok): Likewise.
222 (x86_order_regs_for_local_alloc): Likewise.
223 (ix86_mitigate_rop): Likewise.
224 (ix86_bnd_prefixed_insn_p): Likewise.
225 (ix86_mpx_bound_mode): Likewise.
226 (ix86_make_bounds_constant): Likewise.
227 (ix86_initialize_bounds): Likewise.
228 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
229 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
230 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
231 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
232 (TARGET_CHKP_BOUND_MODE): Likewise.
233 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
234 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
235 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
236 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
237 * config/i386/i386.h (TARGET_MPX): Likewise.
238 (TARGET_MPX_P): Likewise.
239 (VALID_BND_REG_MODE): Likewise.
240 (FIRST_BND_REG): Likewise.
241 (LAST_BND_REG): Likewise.
242 (enum reg_class): Likewise.
243 (BND_REG_P): Likewise.
244 (BND_REGNO_P): Likewise.
246 (ADJUST_INSN_LENGTH): Likewise.
247 * config/i386/i386.md: Likewise.
248 * config/i386/i386.opt: Likewise.
249 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
251 (LINK_MPX): Likewise.
252 (MPX_SPEC): Likewise.
253 (LIBMPX_SPEC): Likewise.
254 (LIBMPXWRAPPERS_SPEC): Likewise.
255 (CHKP_SPEC): Likewise.
256 * config/i386/predicates.md: Likewise.
257 * dbxout.c (dbxout_type): Likewise.
258 * doc/extend.texi: Likewise.
259 * doc/invoke.texi: Likewise.
260 * doc/md.texi: Likewise.
261 * doc/tm.texi: Likewise.
262 * doc/tm.texi.in: Likewise.
263 * dwarf2out.c (is_base_type): Likewise.
264 (gen_formal_types_die): Likewise.
265 (gen_subprogram_die): Likewise.
266 (gen_type_die_with_usage): Likewise.
267 (gen_decl_die): Likewise.
268 (dwarf2out_late_global_decl): Likewise.
269 * expr.c (expand_assignment): Likewise.
270 (emit_storent_insn): Likewise.
271 (store_expr_with_bounds): Likewise.
272 (store_expr): Likewise.
273 (expand_expr_real_1): Likewise.
274 * expr.h (store_expr_with_bounds): Likewise.
275 * function.c (use_register_for_decl): Likewise.
276 (struct bounds_parm_data): Likewise.
277 (assign_parms_augmented_arg_list): Likewise.
278 (assign_parm_find_entry_rtl): Likewise.
279 (assign_parm_is_stack_parm): Likewise.
280 (assign_parm_load_bounds): Likewise.
281 (assign_bounds): Likewise.
282 (assign_parms): Likewise.
283 (expand_function_start): Likewise.
284 * gcc.c (CHKP_SPEC): Likewise.
285 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
286 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
287 (wrestrict_dom_walker::check_call): Likewise.
288 * gimple.c (gimple_build_call_from_tree): Likewise.
289 * gimple.h (enum gf_mask): Likewise.
290 (gimple_call_with_bounds_p): Likewise.
291 (gimple_call_set_with_bounds): Likewise.
292 * gimplify.c (gimplify_init_constructor): Likewise.
293 * ipa-cp.c (initialize_node_lattices): Likewise.
294 (propagate_constants_across_call): Likewise.
295 (find_more_scalar_values_for_callers_subset): Likewise.
296 * ipa-hsa.c (process_hsa_functions): Likewise.
297 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
298 * ipa-icf.c (sem_function::merge): Likewise.
299 * ipa-inline.c (early_inliner): Likewise.
300 * ipa-pure-const.c (warn_function_noreturn): Likewise.
301 (warn_function_cold): Likewise.
302 (propagate_pure_const): Likewise.
303 * ipa-ref.h (enum GTY): Likewise.
304 * ipa-split.c (find_retbnd): Likewise.
305 (consider_split): Likewise.
306 (split_function): Likewise.
307 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
308 * ipa.c (walk_polymorphic_call_targets): Likewise.
309 (symbol_table::remove_unreachable_nodes): Likewise.
310 (process_references): Likewise.
311 (cgraph_build_static_cdtor_1): Likewise.
312 * lto-cgraph.c (lto_output_node): Likewise.
313 (output_refs): Likewise.
314 (compute_ltrans_boundary): Likewise.
315 (input_overwrite_node): Likewise.
316 (input_node): Likewise.
317 (input_cgraph_1): Likewise.
318 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
319 * passes.c (pass_manager::execute_early_local_passes): Likewise.
320 (class pass_chkp_instrumentation_passes): Likewise.
321 (make_pass_chkp_instrumentation_passes): Likewise.
322 * passes.def: Likewise.
323 * rtl.h (struct GTY): Likewise.
324 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
325 * stor-layout.c (layout_type): Likewise.
326 * symtab.c: Likewise.
327 * target.def: Likewise.
328 * targhooks.c (default_chkp_bound_type): Likewise.
329 (default_chkp_bound_mode): Likewise.
330 (default_builtin_chkp_function): Likewise.
331 (default_chkp_function_value_bounds): Likewise.
332 (default_chkp_make_bounds_constant): Likewise.
333 (default_chkp_initialize_bounds): Likewise.
334 * targhooks.h (default_chkp_bound_type): Likewise.
335 (default_chkp_bound_mode): Likewise.
336 (default_builtin_chkp_function): Likewise.
337 (default_chkp_function_value_bounds): Likewise.
338 (default_chkp_make_bounds_constant): Likewise.
339 (default_chkp_initialize_bounds): Likewise.
340 * toplev.c (compile_file): Likewise.
341 (process_options): Likewise.
342 * tree-core.h (DEF_BUILTIN): Likewise.
343 (DEF_BUILTIN_CHKP): Likewise.
344 * tree-inline.c (declare_return_variable): Likewise.
345 (remap_gimple_stmt): Likewise.
347 (initialize_inlined_parameters): Likewise.
348 (expand_call_inline): Likewise.
349 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
350 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
351 (make_pass_ipa_chkp_produce_thunks): Likewise.
352 (make_pass_chkp): Likewise.
353 (make_pass_chkp_opt): Likewise.
354 (make_pass_chkp_instrumentation_passes): Likewise.
355 * tree-pretty-print.c (dump_generic_node): Likewise.
356 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
357 * tree-ssa-dce.c (propagate_necessity): Likewise.
358 (eliminate_unnecessary_stmts): Likewise.
359 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
360 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
361 * tree-ssa-sccvn.h: Likewise.
362 * tree-ssa-strlen.c (get_string_length): Likewise.
363 (valid_builtin_call): Likewise.
364 (adjust_last_stmt): Likewise.
365 (handle_builtin_strchr): Likewise.
366 (handle_builtin_strcpy): Likewise.
367 (handle_builtin_stxncpy): Likewise.
368 (handle_builtin_memcpy): Likewise.
369 (handle_builtin_strcat): Likewise.
370 (strlen_check_and_optimize_stmt): Likewise.
371 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
372 * tree-streamer-in.c: Likewise.
373 * tree-streamer.c (record_common_node): Likewise.
374 * tree.c (tree_code_size): Likewise.
375 (wide_int_to_tree_1): Likewise.
376 (type_contains_placeholder_1): Likewise.
377 (build_common_tree_nodes): Likewise.
378 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
379 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
380 (POINTER_BOUNDS_P): Likewise.
381 (BOUNDED_TYPE_P): Likewise.
382 (BOUNDED_P): Likewise.
383 (CALL_WITH_BOUNDS_P): Likewise.
384 (pointer_bounds_type_node): Likewise.
385 * value-prof.c (gimple_ic): Likewise.
386 * var-tracking.c (vt_add_function_parameters): Likewise.
387 * varasm.c (make_decl_rtl): Likewise.
388 (assemble_start_function): Likewise.
389 (output_constant): Likewise.
390 (maybe_assemble_visibility): Likewise.
391 * varpool.c (ctor_for_folding): Likewise.
392 * chkp-builtins.def: Remove.
393 * ipa-chkp.c: Remove.
394 * ipa-chkp.h: Remove.
395 * rtl-chkp.c: Remove.
396 * rtl-chkp.h: Remove.
397 * tree-chkp-opt.c: Remove.
398 * tree-chkp.c: Remove.
399 * tree-chkp.h: Remove.
401 2018-06-07 Carl Love <cel@us.ibm.com>
403 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
404 vextract_fp_from_shortl): Add BE support.
406 2018-06-07 Paul Koning <ni1d@arrl.net>
408 * compare-elim.c (try_merge_compare): Don't merge compare if
409 address contains a side effect.
410 (try_eliminate_compare): Likewise.
412 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
414 * config.gcc: Support "tremont".
415 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
416 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
418 * config/i386/i386.c (m_TREMONT): Define.
419 (processor_target_table): Add "tremont".
420 (PTA_TREMONT): Define.
421 (ix86_lea_outperforms): Add TARGET_TREMONT.
422 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
423 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
424 and M_INTEL_GOLDMONT_PLUS.
425 (fold_builtin_cpu): Add "tremont".
426 (ix86_add_stmt_cost): Add TARGET_TREMONT.
427 (ix86_option_override_internal): Add "tremont".
428 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
429 (processor_type): Add PROCESSOR_TREMONT.
430 * config/i386/x86-tune.def: Add m_TREMONT.
431 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
433 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
435 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
436 symbol defined for msp430i* devices to be lower case.
438 gcc/testsuite/gcc.target/msp430/
439 * msp430i-device-symbol.c: New test.
440 * msp430f-device-symbol.c: New test.
441 * msp430.h: New test header file.
443 2018-06-07 Richard Biener <rguenther@suse.de>
445 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
446 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
447 Properly wrap signed arithmetic if overflow wraps.
449 2018-06-07 Jakub Jelinek <jakub@redhat.com>
451 PR tree-optimization/69615
452 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
453 of a cast from a same precision integral SSA_NAME in a bb dominated
454 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
455 cast to utype if rhs2 has already a compatible type.
457 2018-06-07 Richard Biener <rguenther@suse.de>
459 PR tree-optimization/85935
460 * graphite-scop-detection.c (find_params_in_bb): Analyze
461 condition operands with respect to the correct loop. Assert
462 the analysis doesn't fail.
464 2018-06-04 Carl Love <cel@us.ibm.com>
466 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
467 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
468 as it is slightly cheaper.
469 (first_match_or_eos_index_<mode>):
470 Calculate index using natural element order.
471 (first_match_index_<mode>):
472 Calculate index using natural element order.
473 (first_match_or_eos_index_<mode>):
474 Calculate index using natural order.
475 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
477 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
478 P9V_BUILTIN_VCLZLSBB_V16QI.
479 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
482 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
484 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
485 indentation and line wrap for many prototypes. Add missing
486 @smallexample directives around block of prototypes for vec_xl and
489 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
491 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
492 track if we pass or return IEEE 128-bit floating point.
493 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
494 C++ mangling that is compatible with GCC 8.1.
495 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
496 (init_cumulative_args): Note if we pass or return IEEE 128-bit
497 floating point types.
498 (rs6000_function_arg_advance_1): Likewise.
499 (rs6000_mangle_type): Optionally generate mangled names that match
500 what GCC 8.1 generated for IEEE 128-bit floating point types.
501 (rs6000_globalize_decl_name): If we have an external function that
502 passes or returns IEEE 128-bit types, generate a weak reference
503 from the mangled name used in GCC 8.1 to the current mangled
505 (rs6000_init_builtins): Make __ibm128 use the long double type if
506 long double is IBM extended double. Make __float128 use the long
507 double type if long double is IEEE 128-bit.
510 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
511 macro for __ibm128 built-in functions.
512 (PACK_IF): Add __ibm128 pack/unpack functions.
513 (UNPACK_IF): Likewise.
514 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
515 enable long double built-in functions if long double is IEEE
516 128-bit floating point.
517 (rs6000_invalid_builtin): Update long double built-in function
519 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
520 functions, adjust the built-in function to use the long double
521 built-in function if __ibm128 and long double are the same type.
522 * doc/extend.texi (PowerPC builtins): Update documention for
523 __builtin_{,un}pack_longdouble. Add documentation for
524 __builtin_{,un}pack_ibm128.
526 2018-06-06 Jim Wilson <jimw@sifive.com>
528 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
529 (struct machine_function): New field interrupt_mode.
530 (riscv_handle_type_attribute): New function. Add forward declaration.
531 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
532 (riscv_expand_epilogue): Check interrupt_mode field.
533 (riscv_set_current_function): Check interrupt attribute args and
534 set interrupt_mode field.
535 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
536 (riscv_sret, riscv_uret): New.
537 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
538 new arguments to interrupt attribute.
540 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
543 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
544 Don't handle -mcpu=power8 if -mpower9-vector is also used.
546 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
548 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
549 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
550 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
551 several redundant entries.
553 2018-06-06 David Malcolm <dmalcolm@redhat.com>
555 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
556 type from "rtx" to "rtx_insn *".
557 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
558 for local "call_insn", removing cast.
559 (ix86_expand_call): Likewise, introducing a "call_insn" local.
561 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
563 PR tree-optimization/86066
564 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
565 for BIT_INSERT_EXPR stores.
567 2018-06-06 Richard Biener <rguenther@suse.de>
569 PR tree-optimization/86062
570 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
572 of to be offsetted base.
574 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
576 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
577 to be static and remove check on interrupt attribute name.
579 gcc/testsuite/gcc.target/msp430/
580 * function-attributes-4.c: New test.
581 * static-interrupts.c: New test.
583 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
585 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
586 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
588 2018-06-05 Steve Ellcey <sellcey@cavium.com>
591 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
593 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
594 Remove second argument, change how error is called.
595 (aarch64_layout_arg): Remove second argument from
596 aarch64_err_no_fpadvsimd call.
597 (aarch64_init_cumulative_args): Ditto.
598 (aarch64_gimplify_va_arg_expr): Ditto.
599 * config/aarch64/aarch64.md (mov<mode>): Ditto.
601 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
603 * config/i386/i386.md (simple_return_indirect_internal): New expander.
604 (*simple_return_indirect_internal<mode>): Rename from
605 simple_return_indirect_internal. Use W mode iterator.
606 (rstorssp): New expander.
607 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
608 (clrssbsy): New expander.
609 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
611 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
613 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
615 (cmse_check_pointed_object): Likewise.
617 2018-06-05 Martin Liska <mliska@suse.cz>
619 PR gcov-profile/47618
620 * doc/invoke.texi: Document how -fprofile-dir format
623 2018-06-05 Richard Biener <rguenther@suse.de>
625 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
626 removal pretend DOM info isn't available so we do not update
627 it and only remove edges, not dominated blocks. Actually free
628 DOM info in case we removed something. Remove unreachable blocks.
629 (mfb_keep_latches): Work with either DOM info or marked backedges.
630 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
631 first. Mark backedges if DOM info isn't available.
632 (Re-)compute DOM info after cleanup_control_flow_pre.
634 2018-06-05 Richard Biener <rguenther@suse.de>
636 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
637 (locus_discrim_hasher::hash): Adjust.
638 (locus_discrim_hasher::equal): Likewise.
639 (next_discriminator_for_locus): Work on line directly.
640 (same_line_p): Pass in expanded locus1 as well.
641 (assign_discriminators): Avoid redundant location expansions.
643 2018-06-05 Richard Biener <rguenther@suse.de>
645 PR tree-optimization/86046
646 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
647 if required after clearing TREE_ADDRESSABLE.
649 2018-06-05 Richard Biener <rguenther@suse.de>
651 PR tree-optimization/86047
652 * tree-ssa-loop.c (for_each_index): Glob handling of all
653 decls and constants and really handle all of them.
655 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
658 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
659 qualifier_void_pointer and qualifier_const_void_pointer.
660 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
661 (arm_init_builtins): Handle the above.
662 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
663 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
664 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
667 2018-06-05 Martin Liska <mliska@suse.cz>
669 * auto-profile.c (read_autofdo_file): Do not use
670 gcov_ctr_summary struct.
671 (afdo_callsite_hot_enough_for_early_inline): Likewise.
672 * coverage.c (struct counts_entry): Likewise.
673 (read_counts_file): Read just single summary entry.
674 (get_coverage_counts): Use gcov_summary struct.
675 * coverage.h (get_coverage_counts): Likewise.
676 * gcov-dump.c (dump_working_sets): Likewise.
677 (tag_summary): Dump just single summary.
678 * gcov-io.c (gcov_write_summary): Write just histogram
680 (gcov_read_summary): Read just single summary.
681 (compute_working_sets): Use gcov_summary struct.
682 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
683 of GCOV_COUNTERS_SUMMABLE.
684 (GCOV_COUNTERS_SUMMABLE): Remove.
685 (GCOV_FIRST_VALUE_COUNTER): Replace with
686 GCOV_COUNTER_V_INTERVAL.
687 (struct gcov_ctr_summary): Remove.
688 (struct gcov_summary): Directly use fields of former
690 (compute_working_sets): Use gcov_summary struct.
691 * gcov.c (read_count_file): Do not use ctrs fields.
692 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
694 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
696 * profile.c: Likewise.
697 * profile.h: Likewise.
699 2018-06-05 Martin Liska <mliska@suse.cz>
701 PR gcov-profile/84846
702 * gcov.c (output_lines): Print working directory only
703 in intermediate format.
705 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
707 * config/s390/s390-builtin-types.def: Add void function type.
708 * config/s390/s390-builtins.def: Use the function type for the
711 2018-06-04 Jim Wilson <jimw@sifive.com>
713 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
715 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
716 and maybe_eh_return. Change regno to unsigned int. Use new args to
717 handle EH_RETURN_DATA_REGNO registers properly.
718 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
719 (riscv_expand_epilogue): Update comment. Change argument name and
720 type. Update code to use new name and type. Pass new args to
721 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
723 * config/riscv/riscv.md (NORMAL_RETURN): New.
724 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
725 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
726 (eh_return): Call gen_eh_return_internal and emit barrier.
727 (eh_return_internal): Call riscv_expand_epilogue.
729 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
731 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
732 bit_insertion field and declare can_be_merged_into method.
733 (merged_store_group::can_be_merged_into): New method.
734 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
735 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
736 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
738 2018-06-04 Richard Biener <rguenther@suse.de>
740 PR tree-optimization/85955
741 * builtins.c (fold_builtin_sincos): Convert pointers to
742 destination to appropriate type before dereferencing.
744 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
746 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
748 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
750 * expr.c (expand_expr_real_1): Force the operand into memory if
751 its TYPE_MODE is BLKmode and if there is no integer mode for
752 the number of bits being extracted.
754 2018-06-04 Jakub Jelinek <jakub@redhat.com>
758 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
759 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
761 2018-06-04 Richard Biener <rguenther@suse.de>
763 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
764 (cleanup_tree_cfg_noloop): ... single caller. Do
765 start_recording_case_labels later.
767 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
769 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
770 to _IMMINTRIN_H_INCLUDED.
771 * config/i386/pconfigintrin.h: Ditto.
772 * config/i386/waitpkgintrin.h: Ditto.
773 * config/i386/immintrin.h: Add includes for sgxintrin.h,
774 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
775 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
776 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
777 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
778 waitpkgintrin.h and cldemoteintrin.h.
780 2018-06-04 Richard Biener <rguenther@suse.de>
782 PR tree-optimization/86038
783 * tracer.c (find_best_successor): Check probability for
784 being initialized, bail out if not.
786 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
789 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
790 of bits to ignore when comparing architectures.
792 2018-06-04 Jakub Jelinek <jakub@redhat.com>
794 PR tree-optimization/69615
795 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
796 maximum or minimum of the type, try to merge it also as if
797 range1 is + [-, x - 1] or + [x + 1, -].
800 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
802 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
804 PR tree-optimization/86034
805 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
806 the unsigned bitfield type in a bit insertion sequence if it does not
807 have a larger precision than the bitfield size.
808 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
810 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
812 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
814 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
816 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
817 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
818 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
819 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
821 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
823 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
824 Disable -fdelete-null-pointer-checks for ELF toolchain.
826 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
827 Kito Cheng <kito.cheng@gmail.com>
829 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
830 (nds32le-*-*, nds32be-*-*): Integrate checking process.
831 (nds32*-*-*): Add glibc and uclibc conditions.
832 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
833 (TARGET_EXCEPT_UNWIND_INFO): Define.
834 * config/nds32/elf.h: New file.
835 * config/nds32/linux.h: New file.
836 * config/nds32/nds32-elf.opt: New file.
837 * config/nds32/nds32-linux.opt: New file.
838 * config/nds32/nds32-fp-as-gp.c
839 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
840 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
842 (nds32_asm_file_end): Ditto.
843 (nds32_print_operand): Ditto.
844 (nds32_insert_attributes): Ditto.
845 (nds32_init_libfuncs): New function.
846 (TARGET_HAVE_TLS): Define.
847 (TARGET_INIT_LIBFUNCS): Define.
848 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
850 (TARGET_ELF): Apply different mcmodel setting.
851 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
852 been migrated into elf.h and linux.h files.
853 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
854 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
855 (mcmodel): The content has been migrated into nds32-elf.opt and
856 nds32-linux.opt files.
857 * config/nds32/t-elf: New file.
858 * config/nds32/t-linux: New file.
860 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
861 Shiva Chen <shiva0217@gmail.com>
863 * config/nds32/constants.md (unspec_volatile_element): Add
864 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
865 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
867 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
868 (make_pass_nds32_fp_as_gp): Declare.
869 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
871 (nds32_asm_function_end_prologue): Remove unused asm output.
872 (nds32_asm_function_begin_epilogue): Remove unused asm output.
873 (nds32_asm_file_start): Output necessary fp_as_gp information.
874 (nds32_option_override): Adjust register usage.
875 (nds32_expand_prologue): Consider fp_as_gp situation.
876 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
877 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
880 (simple_return): Ditto.
881 (omit_fp_begin): Output special directive for fp_as_gp.
882 (omit_fp_end): Output special directive for fp_as_gp.
883 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
884 mforbid-fp-as-gp): New options.
886 2018-06-01 Mark Wielaard <mark@klomp.org>
888 * dwarf2out.c (dwarf2out_finish): Remove generation of
891 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
893 * gimple-ssa-store-merging.c: Include gimple-fold.h.
894 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
895 (struct merged_store_group): Add bit_insertion field.
896 (dump_char_array): Use standard hexadecimal format.
897 (merged_store_group::merged_store_group): Set bit_insertion to false.
898 (merged_store_group::apply_stores): Use optimal buffer size. Deal
899 with BIT_INSERT_EXPR stores. Move up code updating the mask and
900 also print the mask in the dump file.
901 (pass_store_merging::gate): Minor tweak.
902 (imm_store_chain_info::coalesce_immediate): Fix wrong association
903 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
904 stores with INTEGER_CST stores.
905 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
906 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
907 and use it throughout. Generate bit insertion sequences if need be.
908 (pass_store_merging::process_store): Remove redundant condition.
909 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
911 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
913 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
914 the 128-bit floating point types. Fix function comment.
916 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
918 * config/aarch64/aarch64-simd.md
919 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
921 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
924 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
926 PR tree-optimization/85989
927 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
929 (backprop::intersect_uses): Check it when deciding whether this
930 is a backedge reference.
931 (backprop::process_block): Add each phi to m_visited_phis
932 after visiting it, then clear it at the end.
934 2018-06-01 Richard Biener <rguenther@suse.de>
936 * tree-vectorizer.h (vect_dr_stmt): New function.
937 (vect_get_load_cost): Adjust.
938 (vect_get_store_cost): Likewise.
939 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
940 Use vect_dr_stmt instead of DR_SMTT.
941 (vect_record_base_alignments): Likewise.
942 (vect_calculate_target_alignment): Likewise.
943 (vect_compute_data_ref_alignment): Likewise and make static.
944 (vect_update_misalignment_for_peel): Likewise.
945 (vect_verify_datarefs_alignment): Likewise.
946 (vector_alignment_reachable_p): Likewise.
947 (vect_get_data_access_cost): Likewise. Pass down
948 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
949 (vect_get_peeling_costs_all_drs): Likewise.
950 (vect_peeling_hash_get_lowest_cost): Likewise.
951 (vect_enhance_data_refs_alignment): Likewise.
952 (vect_find_same_alignment_drs): Likewise.
953 (vect_analyze_data_refs_alignment): Likewise.
954 (vect_analyze_group_access_1): Likewise.
955 (vect_analyze_group_access): Likewise.
956 (vect_analyze_data_ref_access): Likewise.
957 (vect_analyze_data_ref_accesses): Likewise.
958 (vect_vfa_segment_size): Likewise.
959 (vect_small_gap_p): Likewise.
960 (vectorizable_with_step_bound_p): Likewise.
961 (vect_prune_runtime_alias_test_list): Likewise.
962 (vect_analyze_data_refs): Likewise.
963 (vect_supportable_dr_alignment): Likewise.
964 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
965 (vect_gen_prolog_loop_niters): Likewise.
966 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
967 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
969 (vect_recog_mask_conversion_pattern): Likewise.
970 (vect_try_gather_scatter_pattern): Likewise.
971 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
972 to vect_get_store_cost.
973 (vect_get_store_cost): Get stmt_info instead of DR.
974 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
975 (vect_get_load_cost): Get stmt_info instead of DR.
977 2018-06-01 Richard Biener <rguenther@suse.de>
980 * gimple-fold.c (var_decl_component_p): Also allow offsetted
981 vars wrapped in MEM_REFs.
983 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
985 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
986 Fix subreg tests so that we only return a choice between
987 GENERAL_REGS and FP_REGS if the original classes included both.
989 2018-06-01 Richard Biener <rguenther@suse.de>
992 * tree-ssa-structalias.c (get_function_part_constraint):
993 Handle NULL fi->decl.
994 (find_func_aliases_for_call): Properly handle indirect
996 (find_func_clobbers): Likewise.
997 (ipa_pta_execute): Likewise.
998 (create_variable_info_for): For functions that are ifunc_resolver
999 resolve to a varinfo that contains the result of the resolver call.
1000 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1003 2018-05-31 Michael Collison <michael.collison@arm.com>
1005 * config/aarch64/aarch64.md:
1006 (*fix_to_zero_extenddfdi2): New pattern.
1007 * gcc.target/aarch64/fix_extend1.c: New testcase.
1009 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1013 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1014 and BUILT_IN_STRNCMP_EQ.
1015 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1016 BUILT_IN_STRNCMP_EQ.
1017 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1018 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1019 (gimple_fold_builtin): Likewise.
1020 * tree-ssa-strlen.c (compute_string_length): New function.
1021 (determine_min_obsize): New function.
1022 (handle_builtin_string_cmp): New function to handle calls to
1023 string compare functions.
1024 (strlen_optimize_stmt): Add handling to builtin string compare
1026 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1027 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1028 * tree.c (build_common_builtin_nodes): Add new defines of
1029 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1031 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1034 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1035 functions with naked attribute.
1037 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1039 * config/i386/sse.md (avx_vec_concat<mode>):
1040 Substitute concat_tg_mode mode attribute with xtg_mode.
1041 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1042 (concat_tg_mode): Remove mode attribute.
1044 2018-05-31 Martin Sebor <msebor@redhat.com>
1047 * calls.c (alloc_max_size): Correct a logic error/typo.
1048 Treat excessive arguments as infinite. Warn for invalid arguments.
1049 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1051 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1054 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1055 and movx for Haswell.
1057 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1058 Cesar Philippidis <cesar@codesourcery.com>
1061 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1062 when emitting error on private/firstprivate reductions.
1063 * omp-low.c (lower_omp_target): Avoid reference-type processing
1064 on pointers for firstprivate clause.
1066 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1068 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1071 * config/aarch64/aarch64-simd.md
1072 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1073 (aarch64_ld1_x3_<mode>): Likewise
1074 (aarch64_st1x2<VALLDIF:mode>): Likewise
1075 (aarch64_st1_x2_<mode>): Likewise
1076 (aarch64_st1x3<VALLDIF:mode>): Likewise
1077 (aarch64_st1_x3_<mode>): Likewise
1078 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1079 (vld1_s8_x3): Likewise.
1080 (vld1_u16_x3): Likewise.
1081 (vld1_s16_x3): Likewise.
1082 (vld1_u32_x3): Likewise.
1083 (vld1_s32_x3): Likewise.
1084 (vld1_u64_x3): Likewise.
1085 (vld1_s64_x3): Likewise.
1086 (vld1_f16_x3): Likewise.
1087 (vld1_f32_x3): Likewise.
1088 (vld1_f64_x3): Likewise.
1089 (vld1_p8_x3): Likewise.
1090 (vld1_p16_x3): Likewise.
1091 (vld1_p64_x3): Likewise.
1092 (vld1q_u8_x3): Likewise.
1093 (vld1q_s8_x3): Likewise.
1094 (vld1q_u16_x3): Likewise.
1095 (vld1q_s16_x3): Likewise.
1096 (vld1q_u32_x3): Likewise.
1097 (vld1q_s32_x3): Likewise.
1098 (vld1q_u64_x3): Likewise.
1099 (vld1q_s64_x3): Likewise.
1100 (vld1q_f16_x3): Likewise.
1101 (vld1q_f32_x3): Likewise.
1102 (vld1q_f64_x3): Likewise.
1103 (vld1q_p8_x3): Likewise.
1104 (vld1q_p16_x3): Likewise.
1105 (vld1q_p64_x3): Likewise.
1106 (vst1_s64_x2): Likewise.
1107 (vst1_u64_x2): Likewise.
1108 (vst1_f64_x2): Likewise.
1109 (vst1_s8_x2): Likewise.
1110 (vst1_p8_x2): Likewise.
1111 (vst1_s16_x2): Likewise.
1112 (vst1_p16_x2): Likewise.
1113 (vst1_s32_x2): Likewise.
1114 (vst1_u8_x2): Likewise.
1115 (vst1_u16_x2): Likewise.
1116 (vst1_u32_x2): Likewise.
1117 (vst1_f16_x2): Likewise.
1118 (vst1_f32_x2): Likewise.
1119 (vst1_p64_x2): Likewise.
1120 (vst1q_s8_x2): Likewise.
1121 (vst1q_p8_x2): Likewise.
1122 (vst1q_s16_x2): Likewise.
1123 (vst1q_p16_x2): Likewise.
1124 (vst1q_s32_x2): Likewise.
1125 (vst1q_s64_x2): Likewise.
1126 (vst1q_u8_x2): Likewise.
1127 (vst1q_u16_x2): Likewise.
1128 (vst1q_u32_x2): Likewise.
1129 (vst1q_u64_x2): Likewise.
1130 (vst1q_f16_x2): Likewise.
1131 (vst1q_f32_x2): Likewise.
1132 (vst1q_f64_x2): Likewise.
1133 (vst1q_p64_x2): Likewise.
1134 (vst1_s64_x3): Likewise.
1135 (vst1_u64_x3): Likewise.
1136 (vst1_f64_x3): Likewise.
1137 (vst1_s8_x3): Likewise.
1138 (vst1_p8_x3): Likewise.
1139 (vst1_s16_x3): Likewise.
1140 (vst1_p16_x3): Likewise.
1141 (vst1_s32_x3): Likewise.
1142 (vst1_u8_x3): Likewise.
1143 (vst1_u16_x3): Likewise.
1144 (vst1_u32_x3): Likewise.
1145 (vst1_f16_x3): Likewise.
1146 (vst1_f32_x3): Likewise.
1147 (vst1_p64_x3): Likewise.
1148 (vst1q_s8_x3): Likewise.
1149 (vst1q_p8_x3): Likewise.
1150 (vst1q_s16_x3): Likewise.
1151 (vst1q_p16_x3): Likewise.
1152 (vst1q_s32_x3): Likewise.
1153 (vst1q_s64_x3): Likewise.
1154 (vst1q_u8_x3): Likewise.
1155 (vst1q_u16_x3): Likewise.
1156 (vst1q_u32_x3): Likewise.
1157 (vst1q_u64_x3): Likewise.
1158 (vst1q_f16_x3): Likewise.
1159 (vst1q_f32_x3): Likewise.
1160 (vst1q_f64_x3): Likewise.
1161 (vst1q_p64_x3): Likewise.
1163 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1165 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1166 user_label_prefix to name.
1168 * tree-core.h: Update comment about the format of NAME string
1169 passed to handler in attribute_spec.
1171 * config/msp430/msp430.md: Remove erroneous subreg expression from
1172 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1173 zero_extend{q,h}isi2.
1175 2018-05-30 Borislav Petkov <bp@suse.de>
1177 * doc/extend.texi: Document some architecture specific
1178 constraints and sort entries.
1180 2018-05-30 Martin Sebor <msebor@redhat.com>
1183 * builtins.c (expand_builtin_stpcpy_1): New function.
1184 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1185 only if the former succeeds.
1187 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1189 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1192 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1194 * doc/invoke.texi (-flinker-output): Document
1196 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1198 * passes.c (ipa_write_summaries): Only modify statements if body
1200 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1201 incrementally linking.
1202 (ipa_passes): Likewise.
1203 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1204 pass down resolution info.
1205 * common.opt (flag_incremental_link): Update info.
1206 * gcc.c (plugin specs): Turn flinker-output=* to
1207 -plugin-opt=-linker-output-known
1208 * toplev.c (compile_file): Also cut compilation when doing incremental
1210 * flag-types. (enum lto_partition_model): Add
1211 LTO_LINKER_OUTPUT_NOLTOREL.
1212 (invoke.texi): Add -flinker-output docs.
1213 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1214 link same way as WPA; do not stream in dead initializers.
1216 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1217 darf2out_register_external_decl): Support incremental link.
1219 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1221 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1223 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1225 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1226 it down to simple_object_copy_lto_debug_sections.
1227 (run_gcc): Determine incremental LTO link time and configure
1228 lto1 into non-wpa mode, disable renaming of debug sections.
1230 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1232 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1233 descriptions of various incorrectly documented functions.
1235 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1238 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1239 address check not strict.
1241 2018-05-30 Richard Biener <rguenther@suse.de>
1243 PR tree-optimization/85964
1244 * tracer.c (better_p): Drop initialized count check, we only
1245 call the function with initialized counts now.
1246 (find_best_successor): Do find a best edge if one
1247 has uninitialized count.
1248 (find_best_predecessor): Likewise. Do BB frequency check only
1249 if count is initialized.
1251 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1253 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1254 (aarch64_ldrstr_offset_compare): New.
1255 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1256 load/store orderings.
1257 (aarch64_gen_adjusted_ldpstp): Likewise.
1259 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1261 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1262 Check for subset of GENERAL_REGS and FP_REGS.
1263 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1266 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1268 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1269 and wi::to_poly_offset. Add the current offset and then check
1270 whether the sum fits, rather than using an unchecked addition of
1271 a checked term. Check for a shwi rather than a uhwi.
1272 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1273 (store_constructor): Use poly_int_tree_p.
1274 (expand_expr_real_1): Likewise.
1275 * function.c (assign_temp): Likewise.
1276 * fold-const.c (const_binop): Use poly_int_tree_p and
1278 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1280 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1281 to_poly_offset for MEM offsets.
1282 * ipa-icf.c (sem_variable::equals): Likewise.
1283 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1284 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1285 wi::to_poly_offset for BIT_FIELD_REF offsets.
1286 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1288 * var-tracking.c (emit_note_insn_var_location): Use
1289 tree_to_poly_uint64.
1291 2018-05-29 Jim Wilson <jimw@sifive.com>
1293 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1295 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1298 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1299 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1301 (sse4_1_round<mode>2): Use nonimmediate_operand
1302 for operand 1 predicate.
1304 2018-05-29 Martin Sebor <msebor@redhat.com>
1305 Richard Biener <rguenther@suse.de>
1308 * calls.c (get_size_range): Call determine_value_range instead
1309 of get_value_range..
1310 * tree-vrp.h (determine_value_range): Declared new function.
1311 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1313 2018-05-29 Richard Biener <rguenther@suse.de>
1315 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1316 sure to use non-pattern stmts for get_earlier_stmt arguments.
1317 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1318 called on pattern stmts.
1319 (get_later_stmt): Likewise.
1321 2018-05-29 Martin Liska <mliska@suse.cz>
1323 PR gcov-profile/85759
1324 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1327 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1329 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1331 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1335 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1336 VEC_PACK_FLOAT_EXPR): New tree codes.
1337 * tree-pretty-print.c (op_code_prio): Handle
1338 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1339 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1340 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1341 * tree-inline.c (estimate_operator_cost): Likewise.
1342 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1343 * fold-const.c (const_binop): Likewise.
1344 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1345 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1346 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1347 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1348 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1349 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1350 * expr.c (expand_expr_real_2): Likewise.
1351 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1352 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1353 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1355 * optabs.c (expand_widen_pattern_expr): For
1356 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1357 sign from result type rather than operand's type.
1358 (expand_binop_directly): For vec_packu_float_optab and
1359 vec_packs_float_optab allow result type to be different from operand's
1361 * optabs-tree.c (optab_for_tree_code): Handle
1362 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1363 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1364 * tree-vect-generic.c (expand_vector_operations_1): Handle
1365 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1366 VEC_PACK_FLOAT_EXPR.
1367 * tree-vect-stmts.c (supportable_widening_operation): Handle
1369 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1370 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1371 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1372 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1373 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1375 (vec_pack<floatprefix>_float_<mode>): New expander.
1376 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1378 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1379 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1380 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1381 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1382 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1384 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1385 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1386 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1387 VEC_PACK_FLOAT_EXPR): Document.
1389 2018-05-29 Richard Biener <rguenther@suse.de>
1391 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1393 (stmt_vec_info_vec): Make pointer.
1394 (init_stmt_vec_info_vec): Remove.
1395 (free_stmt_vec_info_vec): Likewise.
1396 (set_stmt_vec_info_vec): New function.
1397 (free_stmt_vec_infos): Likewise.
1398 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1399 (set_vinfo_for_stmt): Likewise.
1400 (get_earlier_stmt): Likewise.
1401 (get_later_stmt): Likewise.
1402 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1403 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1404 (vec_info::~vec_info): Free stmt_vec_infos.
1405 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1406 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1407 (pass_slp_vectorize::execute): Likewise.
1408 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1409 (free_stmt_vec_info_vec): Likewise.
1410 (set_stmt_vec_info_vec): New function.
1411 (free_stmt_vec_infos): Likewise.
1412 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1413 the global stmt_vec_info_vec.
1414 * tree-parloops.c (gather_scalar_reductions): Use
1415 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1418 2018-05-29 Richard Biener <rguenther@suse.de>
1420 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1422 2018-05-29 Martin Liska <mliska@suse.cz>
1423 David Malcolm <dmalcolm@redhat.com>
1425 * vec.c (test_reverse): New.
1426 (vec_c_tests): Add new test.
1427 * vec.h (vl_ptr>::reverse): New function.
1429 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1431 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1433 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1436 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1438 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1440 2018-05-28 Richard Biener <rguenther@suse.de>
1442 PR tree-optimization/85933
1443 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1444 look at stmts marked as vectorizable.
1446 2018-05-28 Richard Biener <rguenther@suse.de>
1448 PR tree-optimization/85934
1449 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1450 vector boolean check before scalar optimization.
1452 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1454 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1457 2018-05-28 Mark Wielaard <mark@klomp.org>
1459 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1460 if it is an expression containing a minus sign.
1462 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1464 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1466 2018-05-27 Paul Koning <ni1d@arrl.net>
1468 * config/pdp11/pdp11.md (truncsihi2): Remove.
1470 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1471 Chung-Ju Wu <jasonwucj@gmail.com>
1473 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1475 (unaligned_store_dw): Ditto.
1476 * config/nds32/nds32-memory-manipulation.c
1477 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1478 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1479 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1480 (emit_setmem_word_loop): Rename to ...
1481 (emit_setmem_doubleword_loop): ... this.
1482 (nds32_gen_dup_4_byte_to_word_value): New function.
1483 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1484 (nds32_expand_setmem_loop): Refine implementation.
1485 (nds32_expand_setmem_loop_v3m): Ditto.
1486 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1489 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1491 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1493 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1495 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1496 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1497 machine->attr_no_prologue_p.
1498 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1499 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1500 (nds32_expand_epilogue): Consider attr_naked_p.
1501 (nds32_expand_epilogue_v3pop): Likewise.
1502 (nds32_can_use_return_insn): Likewise.
1503 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1504 attr_no_prologue_p fields.
1505 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1507 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1510 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1512 * config/i386/sse.md
1513 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1515 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1517 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1519 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1521 (*<floatsuffix>floatv2div2sf2): Rename to ...
1522 (*float<floatunssuffix>v2div2sf2): ... this.
1523 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1524 (float<floatunssuffix>v2div2sf2_mask): ... this.
1525 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1526 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1527 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1529 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1531 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1533 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1535 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1537 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1539 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1540 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1541 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1542 gen_ufix_truncv8dfv8si2.
1543 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1544 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1545 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1546 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1547 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1548 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1549 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1550 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1552 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1556 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1558 2018-05-25 Jim Wilson <jimw@sifive.com>
1560 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1561 * config/riscv/riscv.c (struct machine_function): Add
1562 interrupt_handler_p and attribute_checked_p fields.
1563 (riscv_attribute_table): Add interrupt.
1564 (riscv_interrupt_type_p): New.
1565 (riscv_save_reg_p): Save extra regs for interrupt handler.
1566 (riscv_use_save_libcall): Return false for interrupt handler.
1567 (riscv_first_stack_step): Add forward declaration.
1568 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1569 for interrupt handler with large frame. Use it for saved reg list.
1570 (riscv_expand_prologue): Move flag_stack_usage_info support to
1571 eliminate duplication.
1572 (riscv_expand_epilogue): Generate mret for interrupt handler.
1573 (riscv_epilogue_uses): New.
1574 (riscv_can_use_return_insn): Return false for interrupt handler.
1575 (riscv_function_ok_for_sibcall): Likewise.
1576 (riscv_set_current_function): Add interrupt handler support.
1577 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1578 * config/riscv/riscv.md (UNSPECV_MRET): New.
1580 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1582 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1584 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1586 PR tree-optimization/85712
1587 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1588 this candidate has already been replaced in-situ by a copy.
1590 2018-05-25 Jason Merrill <jason@redhat.com>
1592 PR c++/80485 - inline function non-zero address.
1593 * symtab.c (nonzero_address): Check DECL_COMDAT.
1595 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1598 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1599 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1600 extension of SImode operation. Use const123_operand predicate.
1601 (*saddsi_1): Remove.
1602 (*saddl_se_1): Ditto.
1604 (*ssubl_se_1): Ditto.
1605 * config/alpha/predicates.md (const123_operand): New predicate.
1606 * config/alpha/constraints.md (P): Use IN_RANGE.
1608 2018-05-25 Richard Biener <rguenther@suse.de>
1610 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1612 (ref_maybe_used_by_stmt_p): Likewise.
1613 (stmt_may_clobber_ref_p): Likewise.
1614 (stmt_may_clobber_ref_p_1): Likewise.
1615 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1617 (ref_maybe_used_by_stmt_p): Likewise.
1618 (stmt_may_clobber_ref_p): Likewise.
1619 (stmt_may_clobber_ref_p_1): Likewise.
1620 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1621 the alias oracle to disambiguate DRs with stmts DR analysis
1623 (vect_analyze_data_refs): Do not give up on not analyzable
1624 DRs for BB vectorization. Remove code truncating the dataref
1627 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1630 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1631 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1632 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1634 2018-05-25 Richard Biener <rguenther@suse.de>
1636 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1637 function, combining stmt data ref gathering and fatal analysis
1639 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1640 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1641 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1642 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1643 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1645 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1647 PR tree-optimization/85720
1648 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1649 SCC if all partitions are builtins.
1650 (version_loop_by_alias_check): New parameter. Generate cancelable
1651 runtime alias check if all partitions are builtins.
1652 (distribute_loop): Update call to above function.
1654 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1656 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1657 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1658 (parm_default_def_partition_arg): Ditto.
1659 (set_parm_default_def_partition): Ditto.
1660 (get_parm_default_def_partitions): Ditto and make it static.
1661 (get_undefined_value_partitions): Ditto and make it static.
1662 (remove_ssa_form): Refactor call to init_var_map here.
1663 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1664 computation for loop region.
1665 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1666 (register_default_def): Delete.
1667 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1668 (parm_default_def_partition_arg): Ditto.
1669 (set_parm_default_def_partition): Ditto.
1670 (get_parm_default_def_partitions): Ditto and make it static.
1671 (get_undefined_value_partitions): Ditto and make it static.
1672 (coalesce_with_default, coalesce_with_default): Update comment.
1673 (create_coalesce_list_for_region): New func factored out from
1674 create_outofssa_var_map.
1675 (populate_coalesce_list_for_outofssa): New func factored out from
1676 create_outofssa_var_map and coalesce_ssa_name.
1677 (create_outofssa_var_map): Delete.
1678 (coalesce_ssa_name): Refactor to support live range computation.
1679 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1680 (get_parm_default_def_partitions): Delete.
1681 (get_undefined_value_partitions): Ditto.
1682 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1683 computation for loop region.
1684 (new_tree_live_info, loe_visit_block): Ditto.
1685 (live_worklist, set_var_live_on_entry): Ditto.
1686 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1687 * tree-ssa-live.h (struct _var_map): New fields.
1688 (init_var_map): Change decl.
1689 (region_contains_p): New.
1691 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1693 * tree-ssa-live.h (live_merge_and_clear): Delete.
1695 2018-05-25 Richard Biener <rguenther@suse.de>
1698 * tree-dump.c (dequeue_and_dump): Remove access to removed
1699 operand 2 of a SWITCH_EXPR.
1701 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1703 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1704 multiplication and division.
1705 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1706 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1707 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1708 (cond_udiv_optab, cond_umod_optab): New optabs.
1709 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1710 (IFN_COND_RDIV): New internal functions.
1711 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1712 TRUNC_MOD_EXPR and RDIV_EXPR.
1713 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1714 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1716 (SVE_INT_BINARY): Include mult.
1717 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1718 (optab, sve_int_op): Handle mult.
1719 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1721 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1722 for SVE_INT_BINARY_SD.
1724 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1726 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1727 (optab, sve_int_op): Handle div and udiv.
1728 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1729 for SVE_INT_BINARY_SD.
1730 (*<optab><mode>3): New insn for the same.
1732 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1734 * tree-vect-patterns.c: Include predict.h.
1735 (vect_recog_divmod_pattern): Restrict check for division support
1736 to when optimizing for size.
1738 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1740 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1741 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1742 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1743 (gimple_match_op::set_op): Likewise.
1744 (gimple_resimplify4): Declare.
1745 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1746 (expr::gen_transform): Likewise.
1747 (decision_tree::gen): Generate a simplification routine for 4 operands.
1748 * gimple-match-head.c (gimple_simplify): Add an overload for
1749 4 operands. In the top-level function, handle up to 4 call
1750 arguments and call gimple_resimplify4.
1751 (gimple_resimplify4): New function.
1752 (build_call_internal): Pass a fourth operand.
1753 (maybe_push_to_seq): Likewise.
1754 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1755 Fold VEC_COND_EXPRs of an operation and a default value into
1756 an IFN_COND_* function if possible.
1757 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1759 (SVE_COND_FP_BINARY): Include them.
1760 (optab, sve_fp_op): Handle them.
1761 (SVE_INT_BINARY_REV): New code iterator.
1762 (SVE_COND_FP_BINARY_REV): New int iterator.
1763 (commutative): New int attribute.
1764 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1766 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1768 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1769 (*cond_<optab><mode>): New patterns for reversed operands.
1771 2018-05-25 Richard Biener <rguenther@suse.de>
1773 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1774 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1775 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1776 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1777 * tree-vect-data-refs.c (everywhere): Adjust users.
1778 * tree-vect-loop.c (everywhere): Likewise.
1779 * tree-vect-slp.c (everywhere): Likewise.
1780 * tree-vect-stmts.c (everywhere): Likewise.
1781 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1783 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1785 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1787 (gcc_cv_as_section_exclude): ... this.
1788 Try Solaris as #exclude syntax.
1789 * configure: Regenerate.
1790 * config.in: Regenerate.
1791 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1793 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1794 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1796 * varasm.c (default_elf_asm_named_section): Don't check if
1797 HAVE_GAS_SECTION_EXCLUDE is defined.
1799 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1801 * doc/md.texi: Update the documentation of the cond_* optabs
1802 to mention the new final operand. Fix GET_MODE_NUNITS call.
1803 Describe the scalar case too.
1804 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1805 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1807 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1808 (get_conditional_internal_fn): Update comment.
1809 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1810 accumulator value as a final argument to conditional functions.
1811 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1812 a define_expand and add an "else" operand. Assert for now that
1813 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1814 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1815 (*cond_<optab><mode>): New patterns.
1816 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1817 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1818 (UNSPEC_COND_EOR): Delete.
1819 (optab): Remove associated mappings.
1820 (SVE_INT_BINARY): New code iterator.
1821 (sve_int_op): Remove int attribute and add "minus" to the code
1823 (SVE_COND_INT_OP): Delete.
1824 (SVE_COND_FP_OP): Rename to...
1825 (SVE_COND_FP_BINARY): ...this.
1827 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1829 * optabs.c (can_reuse_operands_p): New function.
1830 (maybe_legitimize_operands): Try to reuse the results for
1833 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1835 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1836 Add {q} suffix to insn mnemonic.
1838 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1840 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1841 (msp430_warn_func_return): New.
1843 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1845 * fold-const.c (tree_nonzero_bits): New function.
1846 * fold-const.h (tree_nonzero_bits): Likewise.
1847 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1848 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1850 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1854 * varasm.c (assemble_alias): Check ifunc_resolver only on
1857 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1860 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1861 when memory input operand is handled.
1863 2018-05-24 Luis Machado <luis.machado@linaro.org>
1865 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1867 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1869 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1871 * match.pd: Delay FMA folds until after vectorization.
1873 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1876 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1877 address check not strict.
1879 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1881 * gimple-match.h (gimple_match_op): New class.
1882 (mprts_hook): Replace parameters with a gimple_match_op *.
1883 (maybe_build_generic_op): Likewise.
1884 (gimple_simplified_result_is_gimple_val): Replace parameters with
1885 a const gimple_match_op *.
1886 (gimple_simplify): Replace code_helper * and tree * parameters with
1887 a gimple_match_op * parameter.
1888 (gimple_resimplify1): Replace code_helper *, tree and tree *
1889 parameters with a gimple_match_op * parameter.
1890 (gimple_resimplify2): Likewise.
1891 (gimple_resimplify3): Likewise.
1892 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1893 parameters with a gimple_match_op * parameter.
1894 * gimple-match-head.c (gimple_simplify): Change prototypes of
1895 auto-generated functions to take a gimple_match_op * instead of
1896 separate code_helper * and tree * parameters. Make the same
1897 change in the top-level overload and update calls to the
1898 gimple_resimplify routines. Update calls to the auto-generated
1899 functions and to maybe_push_res_to_seq in the publicly-facing
1900 operation-specific gimple_simplify overloads.
1901 (gimple_match_op::MAX_NUM_OPS): Define.
1902 (gimple_resimplify1): Replace rcode and ops with a single res_op
1903 parameter. Update call to gimple_simplify.
1904 (gimple_resimplify2): Likewise.
1905 (gimple_resimplify3): Likewise.
1906 (mprts_hook): Replace parameters with a gimple_match_op *.
1907 (maybe_build_generic_op): Likewise.
1908 (build_call_internal): Replace type, nargs and ops with
1909 a gimple_match_op *.
1910 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1911 with a single gimple_match_op *. Update calls to mprts_hook,
1912 build_call_internal and gimple_simplified_result_is_gimple_val.
1913 Factor out code that is common to the tree_code and combined_fn cases.
1914 * genmatch.c (expr::gen_transform): Replace tem_code and
1915 tem_ops with a gimple_match_op called tem_op. Update calls
1916 to the gimple_resimplify functions and maybe_push_res_to_seq.
1917 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1918 res_ops. Update call to the gimple_resimplify functions.
1919 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1920 (decision_tree::gen): Make the functions take a gimple_match_op *
1921 called res_op instead of separate res_code and res_ops parameters.
1922 Update call accordingly.
1923 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1924 and ops with a single res_op parameter. Update calls to
1925 maybe_build_generic_op and maybe_push_res_to_seq.
1926 (fold_stmt_1): Update calls to gimple_simplify and
1927 replace_stmt_with_simplification.
1928 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1929 and gimple_simplified_result_is_gimple_val.
1930 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1932 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1933 with a gimple_match_op *.
1934 (vn_nary_build_or_lookup): Likewise. Update call to
1935 vn_nary_build_or_lookup_1.
1936 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1937 gimple_match_op *. Update calls to the gimple_resimplify routines
1938 and to gimple_simplified_result_is_gimple_val.
1939 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1940 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1941 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1942 (visit_nary_op): Likewise.
1943 (visit_reference_op_load): Likewise.
1945 2018-05-23 Luis Machado <luis.machado@linaro.org>
1947 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1948 modifier for printing the step amount.
1950 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1953 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1956 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1958 * doc/sourcebuild.texi (Endianness): New subsubsection.
1960 2018-05-23 Luis Machado <luis.machado@linaro.org>
1962 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1963 <prefetch_dynamic_strides>: New const bool field.
1964 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1965 prefetch_dynamic_strides.
1966 (exynosm1_prefetch_tune): Likewise.
1967 (thunderxt88_prefetch_tune): Likewise.
1968 (thunderx_prefetch_tune): Likewise.
1969 (thunderx2t99_prefetch_tune): Likewise.
1970 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1972 (aarch64_override_options_internal): Update to set
1973 PARAM_PREFETCH_DYNAMIC_STRIDES.
1974 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1975 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1976 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1977 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1978 prefetch-dynamic-strides setting.
1980 2018-05-23 Luis Machado <luis.machado@linaro.org>
1982 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1983 <minimum_stride>: New const int field.
1984 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1985 minimum_stride field defaulting to -1.
1986 (exynosm1_prefetch_tune): Likewise.
1987 (thunderxt88_prefetch_tune): Likewise.
1988 (thunderx_prefetch_tune): Likewise.
1989 (thunderx2t99_prefetch_tune): Likewise.
1990 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1991 <default_opt_level>: Set to 3.
1992 (aarch64_override_options_internal): Update to set
1993 PARAM_PREFETCH_MINIMUM_STRIDE.
1994 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1995 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1996 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1997 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1998 stride is constant and is below the minimum stride threshold.
2000 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2002 * config/arm/arm-cpus.in (mode26): Delete.
2003 (armv4): Delete mode26 reference.
2004 * config/arm/arm.c (arm_configure_build_target): Delete use of
2007 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2009 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2011 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2012 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2014 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2015 (floatunsdidf2): Ditto.
2017 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2019 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2020 (fixuns_trunc<mode>si2_avx512f): Ditto.
2021 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2022 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2023 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2025 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2027 PR rtl-optimization/79985
2028 * df-scan.c (df_insn_refs_collect): Remove special case for
2029 global registers and asm statements.
2031 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2033 * extend.texi (Global Register Variables): Rewrite the bullet list.
2034 Note that the register is available for allocation. Note that access
2035 via inline asm must use constraints. Add note about async-signal
2036 handlers. Remove paragraph about automagic register selection.
2038 2018-05-23 Richard Biener <rguenther@suse.de>
2040 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2041 of fixed offset from memset VN.
2043 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2045 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2047 (alloc_cand_and_find_basis): Initialize first_interp field.
2048 (slsr_process_mul): Modify first_interp field.
2049 (slsr_process_add): Likewise.
2050 (slsr_process_cast): Modify first_interp field for each new
2052 (slsr_process_copy): Likewise.
2053 (dump_candidate): Dump first_interp field.
2054 (replace_mult_candidate): Process all interpretations, not just
2056 (replace_rhs_if_not_dup): Likewise.
2057 (replace_one_candidate): Likewise.
2059 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2061 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2063 (aarch64_needs_frame_chain): New function.
2064 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2066 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2069 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2070 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2071 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2072 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2073 as true for strict-align.
2074 (aarch64_can_inline_p): Perform checks even when callee has no
2075 attributes to check for strict alignment.
2076 * doc/extend.texi (AArch64 Function Attributes): Document
2078 * doc/invoke.texi: (AArch64 Options): Likewise.
2080 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2082 PR tree-optimization/85853
2083 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2084 the handling of the root of the node to...
2085 (vect_slp_analyze_node_operations_1): ...this new function,
2086 and run the whole thing with the child nodes' def types
2087 set according to their SLP node's def type.
2089 2018-05-23 Richard Biener <rguenther@suse.de>
2092 * tree-data-ref.c (create_runtime_alias_checks): Defer
2093 and ignore overflow warnings.
2095 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2097 PR tree-optimization/85822
2098 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2101 2018-05-23 Richard Biener <rguenther@suse.de>
2103 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2104 memset constants via native_interpret_expr.
2106 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2109 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2111 (cgraph_node::create_alias): Likewise.
2112 (cgraph_node::get_availability): Check ifunc_resolver instead
2113 of looking up ifunc attribute.
2114 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2115 * varasm.c (do_assemble_alias): Likewise.
2116 (assemble_alias): Likewise.
2117 (default_binds_local_p_3): Likewise.
2118 * cgraph.h (cgraph_node): Add ifunc_resolver.
2119 (cgraph_node::only_called_directly_or_aliased_p): Return false
2121 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2123 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2124 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2125 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2126 instead of looking up ifunc attribute.
2128 2018-05-22 Luis Machado <luis.machado@linaro.org>
2130 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2132 2018-05-22 Martin Sebor <msebor@redhat.com>
2135 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2136 only when expasion succeeds.
2137 (expand_builtin_strcmp): Same.
2138 (expand_builtin_strncmp): Same.
2140 2018-05-22 Martin Sebor <msebor@redhat.com>
2142 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2144 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2145 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2147 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2148 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2150 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2151 Add check for legitimate_address.
2152 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2153 (aarch64_swap_ldrstr_operands): New.
2154 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2157 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2158 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2160 * config/aarch64/aarch64.md: New patterns to generate stp
2162 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2163 single words and double words.
2164 (load_pair_sw, load_pair_dw): Likewise.
2165 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2167 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2169 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2170 for different mode ldpstp and add peephole for merged zero stores.
2172 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2174 (aarch64_gen_store_pair): Rename calls to match new patterns.
2175 (aarch64_gen_load_pair): Rename calls to match new patterns.
2176 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2177 (load_pair<DREG:mode><DREG2:mode>): ... This.
2178 (store_pair<mode>): Rename to...
2179 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2180 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2182 (V_INT_EQUIV): Handle SImode.
2183 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2186 2018-05-22 Martin Sebor <msebor@redhat.com>
2189 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2190 or ajust the presumed bound on an operation to avoid unnecessary
2193 2018-05-22 Martin Sebor <msebor@redhat.com>
2195 PR tree-optimization/85826
2196 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2197 assuming that a DECL necesarily has a constant size.
2199 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2202 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2204 2018-05-22 Richard Biener <rguenther@suse.de>
2206 PR tree-optimization/85834
2207 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2208 non-constant and non-zero memset arguments.
2210 2018-05-22 Martin Liska <mliska@suse.cz>
2213 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2215 2018-05-22 Richard Biener <rguenther@suse.de>
2217 PR tree-optimization/85863
2218 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2219 comparisons when vectype is specified.
2220 (vectorizable_condition): Do not specify vectype for
2221 vect_is_simple_cond when SLP vectorizing.
2223 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2226 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2227 define __ibm128 as long double.
2228 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2229 as a distinct type when IEEE 128-bit support is enabled.
2230 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2231 128-bit types to use the correct functions.
2232 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2233 convert between 128-bit floating point types that have different
2234 modes but the same representation, instead of using gen_lowpart to
2236 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2238 (IFKF_reg): New attributes to give the register constraints for
2240 (extend<mode>tf2_internal): New insns to mark an explicit
2241 conversion between 128-bit floating point types that have a
2242 different mode but share the same representation.
2244 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2246 PR tree-optimization/85814
2247 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2248 a null return from get_strinfo when unsharing the next
2249 strinfo in the chain.
2251 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2254 * varasm.c (weak_finish): Clean up weak_decls.
2256 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2258 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2259 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2260 UNSPEC_UADALP values.
2261 * config/aarch64/iterators.md (ABAL): New int iterator.
2264 (sur): Add mappings for the above.
2265 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2267 (aarch64_<sur>abal<mode>_4): Likewise.
2268 (aarch64_<sur>adalp<mode>_3): Likewise.
2269 (<sur>sadv16qi): New define_expand.
2271 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2273 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2274 (*movdf_internal): Ditto.
2275 (*rcpsf2_sse): Ditto.
2276 (*rsqrtsf2_sse): Ditto.
2277 (*sqrt<mode>2_sse): Ditto.
2279 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2281 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2283 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2284 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2285 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2286 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2288 * config/aarch64/arm_neon.h: Likewise.
2289 * config/aarch64/iterators.md (VQ_I): New.
2291 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2293 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2295 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2298 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2300 * config/nds32/constraints.md (S): New constraint.
2301 * config/nds32/nds32.md (call_internal): Use constraint S.
2302 (call_value_internal): Likewise.
2303 (sibcall_internal): Likewise.
2304 (sibcall_value_internal): Likewise.
2306 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2307 Chung-Ju Wu <jasonwucj@gmail.com>
2309 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2312 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2313 Chung-Ju Wu <jasonwucj@gmail.com>
2315 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2316 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2317 (nds32_rtx_costs_impl): Simplify.
2318 (nds32_address_cost_impl): Simplify.
2319 (nds32_init_rtx_costs): New function.
2320 (nds32_rtx_costs_speed_prefer): Likewise.
2321 (nds32_rtx_costs_size_prefer): Likewise.
2322 (nds32_address_cost_speed_prefer): Likewise.
2323 (nds32_address_cost_speed_fwprop): Likewise.
2324 (nds32_address_cost_size_prefer): Likewise.
2325 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2326 * config/nds32/nds32.c (nds32_option_override): Use
2327 nds32_init_rtx_costs function.
2329 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2331 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2332 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2333 (TARGET_PIPELINE_N8): Likewise.
2334 (TARGET_PIPELINE_N10): Likewise.
2335 (TARGET_PIPELINE_N13): Likewise.
2336 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2338 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2340 * config/nds32/nds32-fpu.md: Update copyright year.
2342 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2344 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2346 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2348 * config/nds32/nds32.c
2349 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2350 * config/nds32/nds32.opt (minline-asm-r15): New option.
2352 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2354 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2356 * config/nds32/nds32.md (abssi2): New pattern.
2358 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2360 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2361 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2362 Merge insn pattern from sse_cvtsi2ss<round_name> and
2363 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2364 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2365 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2366 using SWI48 mode iterator.
2367 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2368 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2369 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2370 pattern from sse_cvttss2si<round_saeonly_name>
2371 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2372 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2373 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2374 using SWI48 mode iterator.
2375 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2376 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2377 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2378 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2379 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2380 using SWI48 mode iterator.
2381 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2382 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2383 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2384 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2385 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2386 SWI48 mode iterator.
2387 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2388 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2389 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2390 pattern from sse_cvttsd2si<round_saeonly_name>
2391 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2393 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2395 * config/nds32/nds32-md-auxiliary.c
2396 (nds32_valid_smw_lwm_base_p): Refine.
2397 (nds32_output_smw_single_word): Refine.
2398 (nds32_output_smw_double_word): New.
2399 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2401 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2403 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2404 (nds32_output_stack_pop): Refine.
2405 (nds32_expand_unaligned_load): Refine.
2406 (nds32_expand_unaligned_store): Refine.
2408 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2409 Chung-Ju Wu <jasonwucj@gmail.com>
2411 * config/nds32/constants.md: Add TP_REGNUM constant.
2412 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2413 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2415 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2416 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2417 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2418 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2419 and PIC code generation.
2420 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2422 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2424 * config/nds32/nds32.md: Support TLS and PIC.
2425 * config/nds32/nds32.c: Support TLS and PIC.
2426 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2427 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2430 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2432 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2433 mode with E_ prefix.
2435 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2436 Chung-Ju Wu <jasonwucj@gmail.com>
2438 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2439 * config/nds32/nds32-md-auxiliary.c
2440 (symbolic_reference_mentioned_p): New.
2441 (nds32_legitimize_ict_address): New.
2442 (nds32_expand_ict_move): New.
2443 (nds32_indirect_call_referenced_p): New.
2444 (nds32_symbol_binds_local_p): Delete.
2445 (nds32_long_call_p): Modify.
2446 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2447 * config/nds32/nds32-protos.h
2448 (symbolic_reference_mentioned_p): Declare.
2449 (nds32_legitimize_ict_address): Declare.
2450 (nds32_expand_ict_move): Declare.
2451 (nds32_indirect_call_referenced_p): Declare.
2452 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2453 (nds32_relax_group): Use nds32_ict_const_p as condition.
2454 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2455 (nds32_asm_file_start): Output ict_model directive in asm code.
2456 (nds32_legitimate_address_p): Consider indirect call.
2457 (nds32_print_operand): Consider indirect call.
2458 (nds32_print_operand_address): Consider indirect call.
2459 (nds32_insert_attributes): Handle "indirect_call" attribute.
2460 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2461 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2462 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2463 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2464 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2465 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2466 (TARGET_ICT_MODEL_SMALL): Define.
2467 (TARGET_ICT_MODEL_LARGE): Define.
2468 * config/nds32/nds32.md (movsi): Consider ict model.
2469 (call, call_value): Consider ict model.
2470 (sibcall, sibcall_value): Consider ict model.
2471 * config/nds32/nds32.opt (mict-model): New option.
2472 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2475 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2476 Monk Chiang <sh.chiang04@gmail.com>
2477 Jim Wilson <jimw@sifive.com>
2479 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2480 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2481 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2482 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2483 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2484 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2485 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2486 compute save_libcall_adjustment properly.
2487 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2488 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2489 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2490 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2491 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2492 (ABI_SPEC): Handle mabi=ilp32e.
2493 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2494 (RVE): Add RVE mask.
2495 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2496 <-march>: Add rv32e as an example.
2498 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2501 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2502 (intra_create_variable_infos): Handle C++ constructors.
2504 2018-05-18 Martin Liska <mliska@suse.cz>
2506 * passes.def: Remove a redundant pass.
2508 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2511 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2513 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2515 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2517 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2519 (armv2, armv2a, armv3, armv3m): Delete architectures.
2520 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2521 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2522 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2524 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2525 (*mulsidi3adddi): Likewise.
2526 (mulsidi3): Likewise.
2527 (*mulsidi3_nov6): Likewise.
2528 (umulsidi3): Likewise.
2529 (umulsidi3_nov6): Likewise.
2530 (umaddsidi4): Likewise.
2531 (*umulsidi3adddi): Likewise.
2532 (smulsi3_highpart): Likewise.
2533 (*smulsi3_highpart_nov6): Likewise.
2534 (umulsi3_highpart): Likewise.
2535 (*umulsi3_highpart_nov6): Likewise.
2536 * config/arm/arm.h (arm_arch3m): Delete.
2537 * config/arm/arm.c (arm_arch3m): Delete.
2538 (arm_option_override_internal): Update armv3-related comment.
2539 (arm_configure_build_target): Delete use of isa_bit_mode32.
2540 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2541 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2542 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2544 (mulusa3): Likewise.
2545 * config/arm/arm-protos.h (arm_arch3m): Delete.
2546 * config/arm/arm-tables.opt: Regenerate.
2547 * config/arm/arm-tune.md: Likewise.
2548 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2549 deleted architectures.
2551 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2553 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2554 (armv5t, armv5te): New features.
2555 (ARMv5, ARMv5e): Delete fgroups.
2556 (ARMv5t, ARMv5te): Adjust for above changes.
2558 (armv5, armv5e): Delete arches.
2559 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2561 (*call_reg_arm): Likewise.
2562 (*call_value_reg_armv5): Likewise.
2563 (*call_value_reg_arm): Likewise.
2564 (*call_symbol): Likewise.
2565 (*call_value_symbol): Likewise.
2566 (*sibcall_insn): Likewise.
2567 (*sibcall_value_insn): Likewise.
2569 (prefetch): Likewise.
2570 (define_split and define_peephole2 dependent on arm_arch5):
2572 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2574 (TARGET_ARM_QBIT): Likewise.
2575 (TARGET_DSP_MULTIPLY): Likewise.
2576 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2577 (arm_arch5, arm_arch5e): Delete.
2578 (arm_arch5t, arm_arch5te): Declare.
2579 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2580 (arm_arch5t): Declare.
2581 (arm_option_reconfigure_globals): Update for the above.
2582 (arm_options_perform_arch_sanity_checks): Update comment, replace
2583 use of arm_arch5 with arm_arch5t.
2584 (use_return_insn): Likewise.
2585 (arm_emit_call_insn): Likewise.
2586 (output_return_instruction): Likewise.
2587 (arm_final_prescan_insn): Likewise.
2588 (arm_coproc_builtin_available): Likewise.
2589 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2590 arm_arch5e with arm_arch5t and arm_arch5te.
2591 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2592 (arm_arch5t, arm_arch5te): Declare.
2593 * config/arm/arm-tables.opt: Regenerate.
2594 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2595 * config/arm/t-multilib: Likewise.
2596 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2597 instead of arm_arch5.
2598 (*call_reg_thumb1): Likewise.
2599 (*call_value_reg_thumb1_v5): Likewise.
2600 (*call_value_reg_thumb1): Likewise.
2601 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2603 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2605 2018-05-18 Martin Liska <mliska@suse.cz>
2607 PR gcov-profile/84846
2608 * doc/gcov.texi: Document -t option of gcov tool.
2610 2018-05-18 Martin Liska <mliska@suse.cz>
2612 PR gcov-profile/84846
2613 * gcov.c (print_usage): Add new -t option.
2614 (process_args): Handle the option.
2615 (generate_results): Use stdout as output when requested by
2618 2018-05-18 Martin Liska <mliska@suse.cz>
2620 PR gcov-profile/84846
2621 * coverage.c (coverage_init): Write PWD to .gcno file.
2622 * doc/gcov.texi: Document how working directory is printed.
2623 * gcov-dump.c (dump_gcov_file): Print PWD.
2624 * gcov.c (output_intermediate_file): Likewise.
2625 (read_graph_file): Read PWD string.
2626 (output_lines): Print PWD.
2628 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2631 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2632 for retval and return false if all args to phi are zero.
2634 2018-05-18 Richard Biener <rguenther@suse.de>
2636 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2638 (evrp_dom_walker::before_dom_children): Call it.
2640 2018-05-18 Richard Biener <rguenther@suse.de>
2642 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2643 results when processing array refs with variable index.
2645 2018-05-18 Toon Moene <toon@moene.org>
2647 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2648 directly after that of -floop-interchange. Indicate that both
2649 options are enabled by default when specifying -O3.
2651 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2653 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2654 iterator. Delete separate integer-mode vec_set<mode> expander.
2655 (aarch64_simd_vec_setv2di): Delete.
2656 (vec_setv2di): Delete.
2657 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2658 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2659 the "w, r" alternative.
2661 2018-05-18 Martin Liska <mliska@suse.cz>
2663 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2664 * tree-pass.h (make_pass_lower_switch_O0): New function.
2665 * tree-switch-conversion.c (node_has_low_bound): Remove.
2666 (node_has_high_bound): Likewise.
2667 (node_is_bounded): Likewise.
2668 (class pass_lower_switch): Make it a template type and create
2670 (pass_lower_switch::execute): Add template argument.
2671 (make_pass_lower_switch): New function.
2672 (make_pass_lower_switch_O0): New function.
2673 (do_jump_if_equal): Remove.
2674 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2675 all the hard work to tree optimization passes.
2677 2018-05-18 Martin Liska <mliska@suse.cz>
2679 * dbgcnt.c (limit_low): Renamed from limit.
2680 (limit_high): New variable.
2681 (dbg_cnt_is_enabled): Check for upper limit.
2682 (dbg_cnt): Adjust dumping.
2683 (dbg_cnt_set_limit_by_index): Add new argument for high
2685 (dbg_cnt_set_limit_by_name): Likewise.
2686 (dbg_cnt_process_single_pair): Parse new format.
2687 (dbg_cnt_process_opt): Use strtok.
2688 (dbg_cnt_list_all_counters): Remove 'value' and add
2690 * doc/invoke.texi: Document changes.
2692 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2694 * doc/sourcebuild.texi (scalar_all_fma): Document.
2695 * tree.def (FMA_EXPR): Delete.
2696 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2697 * internal-fn.c (ternary_direct): New macro.
2698 (expand_ternary_optab_fn): Likewise.
2699 (direct_ternary_optab_supported_p): Likewise.
2700 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2701 * builtins.c (fold_builtin_fma): Delete.
2702 (fold_builtin_3): Don't call it.
2703 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2704 * expr.c (expand_expr_real_2): Likewise.
2705 * fold-const.c (operand_equal_p): Likewise.
2706 (fold_ternary_loc): Likewise.
2707 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2708 * gimple.c (DEFTREECODE): Likewise.
2709 * gimplify.c (gimplify_expr): Likewise.
2710 * optabs-tree.c (optab_for_tree_code): Likewise.
2711 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2712 * tree-eh.c (operation_could_trap_p): Likewise.
2713 (stmt_could_throw_1_p): Likewise.
2714 * tree-inline.c (estimate_operator_cost): Likewise.
2715 * tree-pretty-print.c (dump_generic_node): Likewise.
2716 (op_code_prio): Likewise.
2717 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2718 * tree-ssa-operands.c (get_expr_operands): Likewise.
2719 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2720 * fold-const-call.h (fold_fma): Delete.
2721 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2722 CFN_FNMA and CFN_FNMS.
2724 * genmatch.c (combined_fn): New enum.
2725 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2726 (commutative_op): New function.
2727 (commutate): Use it. Handle more than 2 operands.
2728 (dt_operand::gen_gimple_expr): Use commutative_op.
2729 (parser::parse_expr): Allow :c to be used with non-binary
2730 operators if the commutative operand is known.
2731 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2732 CFN_FMS, CFN_FNMA and CFN_FNMS.
2733 (backprop::process_assign_use): Remove FMA_EXPR handling.
2734 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2735 (gen_hsa_fma): New function.
2736 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2737 IFN_FNMA and IFN_FNMS.
2738 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2739 * gimple-fold.h (follow_all_ssa_edges): Declare.
2740 * gimple-fold.c (follow_all_ssa_edges): New function.
2741 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2742 gimple_build interface and use follow_all_ssa_edges to fold the result.
2743 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2744 instead of checking for optabs directly.
2745 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2746 rather than FMA_EXPRs.
2747 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2748 call to IFN_FMA instead of an FMA_EXPR.
2750 2018-05-17 Jim Wilson <jimw@sifive.com>
2752 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2753 sign extended promoted subreg, and the range does not have the sign bit
2754 set, then do a sign extend.
2756 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2757 test, check for sign extended subreg and/or constant operands, and
2758 do a sign extend in that case.
2760 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2762 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2763 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2765 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2766 Change logics_shift_reg to logics_shift_imm.
2767 (thunderx2t99_fp_loadpair_basic): Delete.
2768 (thunderx2t99_fp_storepair_basic): Delete.
2769 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2770 (thunderx2t99_asimd_polynomial): Delete.
2771 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2772 and neon_fp_mul_d_scalar_q.
2773 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2774 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2775 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2776 (thunderx2t99_asimd_lut): Add missing tbl types.
2777 (thunderx2t99_asimd_ext): Delete.
2778 (thunderx2t99_asimd_load1_1_mult): Delete.
2779 (thunderx2t99_asimd_load1_2_mult): Delete.
2780 (thunderx2t99_asimd_load1_ldp): New.
2781 (thunderx2t99_asimd_load1): New.
2782 (thunderx2t99_asimd_load2): Add missing *load2* types.
2783 (thunderx2t99_asimd_load3): New.
2784 (thunderx2t99_asimd_load4): New.
2785 (thunderx2t99_asimd_store1_1_mult): Delete.
2786 (thunderx2t99_asimd_store1_2_mult): Delete.
2787 (thunderx2t99_asimd_store2_mult): Delete.
2788 (thunderx2t99_asimd_store2_onelane): Delete.
2789 (thunderx2t99_asimd_store_stp): New.
2790 (thunderx2t99_asimd_store1): New.
2791 (thunderx2t99_asimd_store2): New.
2792 (thunderx2t99_asimd_store3): New.
2793 (thunderx2t99_asimd_store4): New.
2795 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2797 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2798 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2800 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2801 Segher Boessenkool <segher@kernel.crashing.org>
2804 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2807 2018-05-17 Richard Biener <rguenther@suse.de>
2809 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2810 for pruning loop and prune defs feeding only already visited PHIs.
2812 2018-05-17 Richard Biener <rguenther@suse.de>
2814 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2816 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2817 Richard Biener <rguenther@suse.de>
2819 PR tree-optimization/85793
2820 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2821 for VMAT_ELEMENTWISE.
2823 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2825 * internal-fn.h (lookup_internal_fn): Declare
2826 * internal-fn.c (lookup_internal_fn): New function.
2827 * gimple.c (gimple_build_call_from_tree): Handle calls to
2829 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2830 internal function names.
2831 * tree-pretty-print.c (dump_generic_node): Likewise.
2832 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2834 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2836 * gimple-fold.h (gimple_build): Make the function forms take
2837 combined_fn rather than built_in_function.
2838 (gimple_simplify): Likewise.
2839 * gimple-match-head.c (gimple_simplify): Likewise.
2840 * gimple-fold.c (gimple_build): Likewise.
2841 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2842 rather than gimple_build_call_internal.
2843 (get_initial_defs_for_reduction): Likewise.
2844 (vect_create_epilog_for_reduction): Likewise.
2845 (vectorizable_live_operation): Likewise.
2847 2018-05-17 Martin Liska <mliska@suse.cz>
2849 * gimple-ssa-sprintf.c (format_directive): Do not use
2850 space in between 'G_' and '('.
2852 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2855 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2856 even if the mask is not all ones.
2859 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2861 (ix86_gimple_fold_builtin): Likewise.
2864 * config/i386/i386.c: Include tree-vector-builder.h.
2865 (ix86_vector_shift_count): New function.
2866 (ix86_fold_builtin): Fold shift builtins by scalar count.
2867 (ix86_gimple_fold_builtin): Likewise.
2869 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2870 _mm512_setzero): New intrinsics.
2872 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2873 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2875 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2876 code generation for cases where splatting a value is not useful.
2877 * simplify-rtx.c (simplify_ternary_operation): Simplify
2878 vec_merge across a vec_duplicate and a paradoxical subreg forming
2879 a vector mode to a vec_concat.
2881 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2883 * config.gcc: Support "goldmont-plus".
2884 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2886 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2887 PROCESSOR_GOLDMONT_PLUS.
2888 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2889 (processor_target_table): Add "goldmont-plus".
2890 (PTA_GOLDMONT_PLUS): Define.
2891 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2892 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2893 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2894 (fold_builtin_cpu): Add "goldmont-plus".
2895 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2896 (ix86_option_override_internal): Add "goldmont-plus".
2897 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2898 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2899 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2900 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2902 2018-05-17 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/85757
2905 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2906 remove defs that only feed that PHI from further processing.
2908 2018-05-16 Jim Wilson <jimw@sifive.com>
2910 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2912 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2914 2018-05-16 Mark Wielaard <mark@klomp.org>
2916 * dwarf2out.c (count_index_strings): New function.
2917 (output_indirect_strings): Call count_index_strings and generate
2918 header for dwarf_version >= 5.
2920 2018-05-16 Mark Wielaard <mark@klomp.org>
2922 * dwarf2out.c (dwarf_FORM): New function.
2923 (set_indirect_string): Use dwarf_FORM.
2924 (reset_indirect_string): Likewise.
2925 (size_of_die): Likewise.
2926 (value_format): Likewise.
2927 (output_die): Likewise.
2928 (add_skeleton_AT_string): Likewise.
2929 (output_macinfo_op): Likewise.
2930 (index_string): Likewise.
2931 (output_index_string_offset): Likewise.
2932 (output_index_string): Likewise.
2933 (count_index_strings): Likewise.
2935 2018-05-16 Carl Love <cel@us.ibm.com>
2937 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2938 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2940 2018-05-16 Martin Jambor <mjambor@suse.cz>
2942 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2943 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2945 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2947 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2948 (fnma<mode>4): Likewise.
2949 (fms<mode>4): Likewise.
2950 (fnms<mode>4): Likewise.
2951 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2952 (aarch64_fnma<mode>4): Likewise.
2953 (aarch64_fms<mode>4): Likewise.
2954 (aarch64_fnms<mode>4): Likewise.
2955 (aarch64_fnmadd<mode>4): Likewise.
2957 2018-05-16 Jason Merrill <jason@redhat.com>
2959 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2961 2018-05-16 Richard Biener <rguenther@suse.de>
2963 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2964 (dump_stmt_cost): Declare.
2965 (add_stmt_cost): Dump cost we add.
2966 (add_stmt_costs): New function.
2967 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2969 (vect_analyze_stmt): Adjust prototype.
2970 (vectorizable_condition): Likewise.
2971 (vectorizable_live_operation): Likewise.
2972 (vectorizable_reduction): Likewise.
2973 (vectorizable_induction): Likewise.
2974 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2975 cost vector to pass to vectorizable_ and record afterwards.
2976 (vect_model_reduction_cost): Take cost vector argument and adjust.
2977 (vect_model_induction_cost): Likewise.
2978 (vectorizable_reduction): Likewise.
2979 (vectorizable_induction): Likewise.
2980 (vectorizable_live_operation): Likewise.
2981 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2982 SLP_TREE_NUMBER_OF_VEC_STMTS.
2983 (vect_analyze_slp_cost_1): Remove.
2984 (vect_analyze_slp_cost): Likewise.
2985 (vect_slp_analyze_node_operations): Take visited args and
2986 a target cost vector. Avoid processing already visited stmt sets.
2987 (vect_slp_analyze_operations): Use a local cost vector to gather
2988 costs and register those of non-discarded instances.
2989 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2990 (vect_schedule_slp_instance): Remove copying of
2991 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2993 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2994 adding cost. Record cost entry location.
2995 (vect_prologue_cost_for_slp_op): Function to compute cost of
2996 a constant or invariant generated for SLP vect in the prologue,
2997 split out from vect_analyze_slp_cost_1.
2998 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2999 (vect_model_promotion_demotion_cost): Likewise.
3000 (vect_model_store_cost): Likewise, make static.
3001 (vect_model_load_cost): Likewise.
3002 (vectorizable_bswap): Add cost vector arg and adjust.
3003 (vectorizable_call): Likewise.
3004 (vectorizable_simd_clone_call): Likewise.
3005 (vectorizable_conversion): Likewise.
3006 (vectorizable_assignment): Likewise.
3007 (vectorizable_shift): Likewise.
3008 (vectorizable_operation): Likewise.
3009 (vectorizable_store): Likewise.
3010 (vectorizable_load): Likewise.
3011 (vectorizable_condition): Likewise.
3012 (vectorizable_comparison): Likewise.
3013 (can_vectorize_live_stmts): Likewise.
3014 (vect_analyze_stmt): Likewise.
3015 (vect_transform_stmt): Adjust calls to vectorizable_*.
3016 * tree-vectorizer.c: Include gimple-pretty-print.h.
3017 (dump_stmt_cost): New function.
3019 2018-05-16 Richard Biener <rguenther@suse.de>
3021 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3022 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3023 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3024 (check_name): New callback.
3025 (dse_classify_store): Track cycles via a visited bitmap of PHI
3026 defs and simplify handling of in-loop and across loop dead stores
3027 and properly fail for loop-variant refs. Handle byte-tracking with
3028 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3031 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3033 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3034 (vect_get_mask_type_for_stmt): Likewise.
3035 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3037 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3038 to determine the statement's vector type and the vector type that
3039 should be used for calculating nunits. Deal with cases in which
3040 the type has to be deferred.
3041 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3042 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3043 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3044 (vect_determine_vf_for_stmt): New functions, split out from...
3045 (vect_determine_vectorization_factor): ...here.
3046 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3047 (vect_get_mask_type_for_stmt): New functions, split out from
3048 vect_determine_vectorization_factor.
3050 2018-05-16 Richard Biener <rguenther@suse.de>
3052 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3053 verify the [VEC_]COND_EXPR embedded comparison.
3055 2018-05-15 Martin Sebor <msebor@redhat.com>
3057 PR tree-optimization/85753
3058 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3059 RECORD_TYPE in addition to ARRAY_TYPE.
3061 2018-05-15 Martin Sebor <msebor@redhat.com>
3064 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3066 2018-05-15 Richard Biener <rguenther@suse.de>
3068 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3069 add by_clobber_p one. Change algorithm to collect all defs
3070 representing uses we need to walk and try reducing them to
3071 a single one before failing.
3072 (dse_dom_walker::dse_optimize_stmt): Adjust.
3074 2018-05-13 Mark Wielaard <mark@klomp.org>
3076 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3077 (size_of_loc_descr): Likewise.
3078 (output_loc_operands): Likewise.
3079 (output_loc_operands_raw): Likewise.
3080 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3081 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3082 (hash_loc_operands): Likewise.
3083 (compare_loc_operands): Likewise.
3085 2018-05-14 Mark Wielaard <mark@klomp.org>
3087 * dwarf2out.c (count_index_addrs): New function.
3088 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3090 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3092 PR tree-optimization/83648
3093 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3094 return value as malloc candidate.
3096 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3099 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3100 param as true in call to suggest_attribute.
3102 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3104 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3105 -mreadonly-in-sdata.
3107 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3109 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3111 (aarch64_crypto_aesd_fused): Likewise.
3113 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3115 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3116 (movsi_aarch64): Likewise.
3117 (load_pairsi): Likewise.
3118 (load_pairdi): Likewise.
3119 (store_pairsi): Likewise.
3120 (store_pairdi): Likewise.
3121 (load_pairsf): Likewise.
3122 (load_pairdf): Likewise.
3123 (store_pairsf): Likewise.
3124 (store_pairdf): Likewise.
3125 (zero_extend): Likewise.
3126 (trunc): Swap alternatives.
3127 (fcvt_target): Add '?' to prefer w over r.
3129 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3132 * config/i386/i386.md: Disallow non-commutative arithmetics in
3133 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3134 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3135 in the peephole2 before it.
3137 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3139 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3140 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3141 (ix86_handle_option): Handle -mcldemote.
3142 * config.gcc: New header.
3143 * config/i386/cldemoteintrin.h: New file.
3144 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3145 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3147 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3148 OPTION_MASK_ISA_CLDEMOTE.
3149 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3150 (ix86_valid_target_attribute_inner_p): Ditto.
3151 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3152 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3153 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3154 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3155 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3157 * config/i386/i386.opt: Add -mcldemote.
3158 * config/i386/x86intrin.h: New header.
3159 * doc/invoke.texi: Add -mcldemote.
3161 2018-05-14 Richard Biener <rguenther@suse.de>
3163 * doc/match-and-simplify.texi: Adjust :s documentation.
3165 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3167 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3168 intended memcpy size.
3169 (REORDER_45): Likewise.
3171 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3173 * sort.cc: New file.
3174 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3175 * vec.c (qsort_chk): Use gcc_qsort.
3176 * Makefile.in (OBJS-libcommon): Add sort.o.
3177 (build/sort.o): New target. Use it...
3178 (BUILD_RTL): ... here, and...
3179 (build/gencfn-macros): ... here, and...
3180 (build/genmatch): ... here.
3182 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3183 Chung-Ju Wu <jasonwucj@gmail.com>
3185 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3186 * config/nds32/nds32-graywolf.md: New file.
3187 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3188 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3190 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3191 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3192 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3193 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3194 * config/nds32/pipelines.md: Include n15 settings.
3196 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3197 Chung-Ju Wu <jasonwucj@gmail.com>
3199 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3200 * config/nds32/nds32-n13.md: New file.
3201 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3202 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3204 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3205 * config/nds32/nds32.md (pipeline_model): Add n13.
3206 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3207 * config/nds32/pipelines.md: Include n13 settings.
3209 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3210 Chung-Ju Wu <jasonwucj@gmail.com>
3212 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3213 * config/nds32/nds32-n10.md: New file.
3214 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3215 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3217 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3218 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3219 * config/nds32/nds32.md (pipeline_model): Add n10.
3220 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3221 * config/nds32/pipelines.md: Include n10 settings.
3223 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3224 Kito Cheng <kito.cheng@gmail.com>
3225 Chung-Ju Wu <jasonwucj@gmail.com>
3227 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3228 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3229 Add enum values for DSP extension instructions.
3230 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3232 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3233 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3235 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3236 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3237 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3238 * config/nds32/nds32-intrinsic.md: Likewise.
3239 * config/nds32/nds32_intrinsic.h: Likewise.
3240 * config/nds32/nds32-md-auxiliary.c: Likewise.
3241 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3242 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3243 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3244 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3245 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3246 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3247 TYPE_DMAC in switch statement.
3248 * config/nds32/nds32.c: New checking and implementation for DSP
3249 extension instructions.
3250 * config/nds32/nds32.h: Likewise.
3251 * config/nds32/nds32.md: Likewise.
3252 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3253 * config/nds32/predicates.md: Implement new predicates for DSP
3256 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3258 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3259 Reformat alternatives and attributes so it is easier to identify
3260 which constraints/attributes go with which instruction.
3261 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3262 (mov<mode>_softfloat32, FMOVE64): Likewise.
3263 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3264 (mov<mode>_softfloat64, FMOVE64): Likewise.
3266 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3268 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3270 (Basic PowerPC Built-in Functions): The new name of the
3271 subsection previously known as "PowerPC Built-in Functions".
3272 (Basic PowerPC Built-in Functions Available on all Configurations):
3274 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3275 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3276 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3277 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3279 2018-05-11 Martin Jambor <mjambor@suse.cz>
3282 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3285 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3288 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3290 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3292 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3293 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3294 (ix86_handle_option): Handle -mwaitpkg.
3295 * config.gcc: New header.
3296 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3297 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3298 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3300 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3301 OPTION_MASK_ISA_WAITPKG.
3302 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3303 (ix86_option_override_internal): Add PTA_WAITPKG.
3304 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3305 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3306 IX86_BUILTIN_TPAUSE.
3307 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3308 __builtin_ia32_umwait and __builtin_ia32_tpause.
3309 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3310 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3311 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3312 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3313 UNSPECV_TPAUSE): New.
3314 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3315 * config/i386/i386.opt: Add -mwaitpkg.
3316 * config/i386/waitpkgintrin.h: New file.
3317 * config/i386/x86intrin.h: New header.
3318 * doc/invoke.texi: Add -mwaitpkg.
3320 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3323 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3325 (cortex-m0): Use armv6s-m isa.
3326 (cortex-m0plus): Likewise.
3327 (cortex-m1): Likewise.
3328 (cortex-m0.small-multiply): Likewise.
3329 (cortex-m0plus.small-multiply): Likewise.
3330 (cortex-m1.small-multiply): Likewise.
3332 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3333 Jakub Jelinek <jakub@redhat.com>
3335 PR tree-optimization/85692
3336 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3337 source permute as well.
3339 2018-05-11 Martin Liska <mliska@suse.cz>
3342 * doc/extend.texi: Document LLVM style format for no_sanitize
3345 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3347 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3348 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3349 (mode_supports_vsx_dform_quad): Likewise.
3350 (mode_supports_vmx_dform): Move these functions to be next to the
3351 other mode_supports functions.
3352 (mode_supports_dq_form): Likewise.
3353 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3354 mode_supports_dq_form.
3355 (reg_offset_addressing_ok_p): Likewise.
3356 (offsettable_ok_by_alignment): Likewise.
3357 (rs6000_legitimate_offset_address_p): Likewise.
3358 (legitimate_lo_sum_address_p): Likewise.
3359 (rs6000_legitimize_address): Likewise.
3360 (rs6000_legitimize_reload_address): Likewise.
3361 (rs6000_secondary_reload_inner): Likewise.
3362 (rs6000_preferred_reload_class): Likewise.
3363 (rs6000_output_move_128bit): Likewise.
3365 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3367 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3368 Generate SImode target register for null target.
3369 <case IX86_BUILTIN_XGETBV>: Ditto.
3370 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3371 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3373 2018-05-10 Carl Love <cel@us.ibm.com>
3375 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3376 dcbtt and dcbtstt if operands[2] is 0.
3378 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3381 * config/i386/sse.md (usadv64qi): New expander.
3383 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3385 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3386 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3387 -maltivec=be support.
3388 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3389 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3390 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3391 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3392 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3393 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3394 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3395 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3396 altivec_vsumsws): Adjust.
3397 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3398 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3399 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3401 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3402 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3403 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3404 (altivec_lve<VI_char>x): Delete expand.
3405 (*altivec_lve<VI_char>x_internal): Rename to...
3406 (altivec_lve<VI_char>x): ... this.
3407 (altivec_lvxl_<mode>): Delete expand.
3408 (*altivec_lvxl_<mode>_internal): Rename to ...
3409 (altivec_lvxl_<mode>): ... this.
3410 (altivec_stvxl_<mode>): Delete expand.
3411 (*altivec_stvxl_<mode>_internal): Rename to ...
3412 (altivec_stvxl_<mode>): ... this.
3413 (altivec_stve<VI_char>x): Delete expand.
3414 (*altivec_stve<VI_char>x_internal): Rename to ...
3415 (altivec_stve<VI_char>x): ... this.
3416 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3417 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3418 reduc_plus_scal_<mode>): Adjust.
3419 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3421 (rs6000_cpu_cpp_builtins): Adjust.
3422 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3423 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3424 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3425 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3426 -maltivec=be support.
3427 (rs6000_split_vec_extract_var): Adjust.
3428 (rs6000_split_v4si_init): Adjust.
3429 (swap_selector_for_mode): Delete.
3430 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3431 altivec_expand_stvex_be): Delete.
3432 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3433 -maltivec=be support.
3434 (rs6000_gimple_fold_builtin): Ditto.
3435 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3437 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3438 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3439 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3440 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3441 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3442 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3443 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3444 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3445 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3446 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3447 anonymous split): Adjust.
3448 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3449 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3451 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3453 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3454 when --with-gxx-include-dir is also specified.
3455 * configure: Regenerate.
3457 2018-05-09 Jim Wilson <jimw@sifive.com>
3460 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3461 * config/riscv/t-withmultilib: New.
3462 * config/riscv/withmultilib.h: New.
3463 * doc/install.texi: Document RISC-V --with-multilib-list support.
3465 2018-05-09 Richard Biener <rguenther@suse.de>
3467 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3469 (vect_bb_vectorization_profitable_p): Adjust. Compute
3470 actual scalar cost using the cost vector and the add_stmt_cost
3473 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3475 PR rtl-optimization/85645
3476 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3477 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3479 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3481 PR rtl-optimization/85645
3482 * shrink-wrap.c (spread_components): Return a boolean saying if
3483 anything was changed.
3484 (try_shrink_wrapping_separate): Iterate spread_components until
3485 nothing changes anymore.
3487 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3489 PR rtl-optimization/85645
3490 * regrename.c (build_def_use): Also kill the chains that include the
3491 destination of a REG_CFA_REGISTER note.
3493 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3495 PR rtl-optimization/85645
3496 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3497 insn that has a REG_CFA_REGISTER note.
3499 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3501 * cfgexpand.c (expand_clobber): New function.
3502 (expand_gimple_stmt_1): Use it.
3503 * tree-vect-stmts.c (vect_clobber_variable): New function,
3505 (vectorizable_simd_clone_call): ...here.
3506 (vectorizable_store): Emit a clobber either side of an
3507 IFN_STORE_LANES sequence.
3508 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3510 2018-05-09 Tom de Vries <tom@codesourcery.com>
3513 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3514 (define_insn "trap_if_false"): Add exit after trap.
3516 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3518 PR rtl-optimization/85638
3519 * bb-reorder.c: Include common/common-target.h.
3520 (create_forwarder_block): New function extracted from...
3521 (fix_up_crossing_landing_pad): ...here. Rename into...
3522 (dw2_fix_up_crossing_landing_pad): ...this.
3523 (sjlj_fix_up_crossing_landing_pad): New function.
3524 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3525 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3526 from both partitions and exit the loop after one iteration.
3528 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3531 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3533 (Basic PowerPC Built-in Functions): The new name of the
3534 subsection previously known as "PowerPC Built-in Functions".
3535 (Basic PowerPC Built-in Functions Available on all Configurations):
3537 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3539 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3541 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3543 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3546 2018-05-08 Jim Wilson <jimw@sifive.com>
3548 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3549 (LD_EMUL_SUFFIX): New.
3550 (LINK_SPEC): Use it.
3552 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3554 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3556 (Basic PowerPC Built-in Functions): The new name of the
3557 subsection previously known as "PowerPC Built-in Functions".
3558 (Basic PowerPC Built-in Functions Available on all Configurations):
3560 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3562 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3564 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3566 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3569 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3572 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3573 after cmpelim optimization.
3575 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3577 * config.gcc: Support "goldmont".
3578 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3579 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3581 * config/i386/i386.c (m_GOLDMONT): Define.
3582 (processor_target_table): Add "goldmont".
3583 (PTA_GOLDMONT): Define.
3584 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3585 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3586 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3587 (fold_builtin_cpu): Add "goldmont".
3588 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3589 (ix86_option_override_internal): Add "goldmont".
3590 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3591 (processor_type): Add PROCESSOR_GOLDMONT.
3592 * config/i386/i386.md: Add CPU "glm".
3593 * config/i386/glm.md: New file.
3594 * config/i386/x86-tune.def: Add m_GOLDMONT.
3595 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3597 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3600 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3602 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3603 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3604 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3607 * config/i386/i386.c (ix86_fold_builtin): Handle
3608 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3611 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3612 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3614 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3617 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3618 (check_arch): Likewise.
3619 (check_fpu): Return the result rather than printing it.
3620 (end arch): Fix operator precedence.
3621 (end cpu): Likewise.
3622 (END): Print the result from check_fpu.
3624 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3625 Alan Hayward <alan.hayward@arm.com>
3626 David Sherwood <david.sherwood@arm.com>
3628 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3629 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3630 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3631 (*fcmuo<mode>_and): New patterns.
3633 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3635 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3636 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3637 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3638 (cmp_op, sve_imm_con): New code attributes.
3639 (SVE_COND_INT_CMP, imm_con): Delete.
3640 (cmp_op): Remove above unspecs from int attribute.
3641 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3643 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3644 comparison-specific unspecs.
3645 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3646 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3647 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3648 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3649 (*vec_fcm<cmp_op><mode>): Rename to...
3650 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3651 (*vec_fcmuo<mode>): Rename to...
3652 (*fcmuo<mode>): ...this and adjust likewise.
3653 (*pred_fcm<cmp_op><mode>): New pattern.
3654 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3655 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3657 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3659 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3660 (aarch64_emit_sve_predicated_cond): New function.
3661 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3662 (aarch64_emit_unspec_cond_or): Replace with...
3663 (aarch64_emit_sve_or_conds): ...this new function. Use
3664 aarch64_emit_sve_ptrue_op for the individual comparisons and
3665 aarch64_emit_binop to OR them together.
3666 (aarch64_emit_inverted_unspec_cond): Replace with...
3667 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3668 aarch64_emit_sve_ptrue_op for the comparison and
3669 aarch64_emit_unop to invert the result.
3670 (aarch64_expand_sve_vec_cmp_float): Update after the above
3671 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3673 2018-05-07 Nathan Sidwell <nathan@acm.org>
3675 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3676 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3677 (Backwards Compatibility): Likewise.
3679 2018-05-07 Luis Machado <luis.machado@linaro.org>
3683 2018-05-07 Luis Machado <luis.machado@linaro.org>
3685 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3686 <prefetch_dynamic_strides>: New const bool field.
3687 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3688 prefetch_dynamic_strides.
3689 (exynosm1_prefetch_tune): Likewise.
3690 (thunderxt88_prefetch_tune): Likewise.
3691 (thunderx_prefetch_tune): Likewise.
3692 (thunderx2t99_prefetch_tune): Likewise.
3693 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3695 (aarch64_override_options_internal): Update to set
3696 PARAM_PREFETCH_DYNAMIC_STRIDES.
3697 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3698 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3699 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3700 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3701 prefetch-dynamic-strides setting.
3703 2018-05-07 Luis Machado <luis.machado@linaro.org>
3705 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3706 <minimum_stride>: New const int field.
3707 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3708 minimum_stride field.
3709 (exynosm1_prefetch_tune): Likewise.
3710 (thunderxt88_prefetch_tune): Likewise.
3711 (thunderx_prefetch_tune): Likewise.
3712 (thunderx2t99_prefetch_tune): Likewise.
3713 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3714 (aarch64_override_options_internal): Update to set
3715 PARAM_PREFETCH_MINIMUM_STRIDE.
3716 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3717 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3718 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3719 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3720 stride is constant and is below the minimum stride threshold.
3722 2018-05-07 Luis Machado <luis.machado@linaro.org>
3724 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3727 2018-05-07 Luis Machado <luis.machado@linaro.org>
3729 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3730 <prefetch_dynamic_strides>: New const bool field.
3731 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3732 prefetch_dynamic_strides.
3733 (exynosm1_prefetch_tune): Likewise.
3734 (thunderxt88_prefetch_tune): Likewise.
3735 (thunderx_prefetch_tune): Likewise.
3736 (thunderx2t99_prefetch_tune): Likewise.
3737 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3739 (aarch64_override_options_internal): Update to set
3740 PARAM_PREFETCH_DYNAMIC_STRIDES.
3741 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3742 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3743 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3744 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3745 prefetch-dynamic-strides setting.
3747 2018-05-07 Luis Machado <luis.machado@linaro.org>
3749 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3750 <minimum_stride>: New const int field.
3751 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3752 minimum_stride field.
3753 (exynosm1_prefetch_tune): Likewise.
3754 (thunderxt88_prefetch_tune): Likewise.
3755 (thunderx_prefetch_tune): Likewise.
3756 (thunderx2t99_prefetch_tune): Likewise.
3757 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3758 (aarch64_override_options_internal): Update to set
3759 PARAM_PREFETCH_MINIMUM_STRIDE.
3760 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3761 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3762 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3763 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3764 stride is constant and is below the minimum stride threshold.
3766 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3769 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3770 the type is addressable. Don't force op into register if it has
3773 2018-05-05 Roland McGrath <mcgrathr@google.com>
3776 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3777 any section for which we don't know a specific type it should have,
3778 regardless of name. Previously this was done only for the exact
3779 names ".init_array", ".fini_array", and ".preinit_array".
3780 (default_elf_asm_named_section): Add comment about
3781 relationship with default_section_type_flags and SECTION_NOTYPE.
3782 (get_section): Don't consider it a type conflict if one side has
3783 SECTION_NOTYPE and the other doesn't, as long as neither has the
3784 SECTION_BSS et al used in the default_section_type_flags logic.
3786 2018-05-05 Tom de Vries <tom@codesourcery.com>
3789 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3790 (workaround_barsyncs): New function.
3791 (nvptx_reorg): Use workaround_barsyncs.
3792 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3793 (define_expand "nvptx_membar_cta"): New define_expand.
3794 (define_insn "*nvptx_membar_cta"): New insn.
3796 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3798 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3799 To improve optimization opportunities.
3800 * builtin-types.def: The new needed builtin types for the above.
3802 2018-05-04 Richard Biener <rguenther@suse.de>
3804 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3805 * gimple-ssa-store-merging.c
3806 (imm_store_chain_info::output_merged_store): Remove redundant create,
3807 release split_store vector contents on failure.
3808 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3809 scalar stmt vector on cache hit.
3811 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3813 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3815 * config.gcc (powerpc-xilinx-eabi*): Remove.
3816 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3818 (fusion_addis_mem_combo_load): Ditto.
3819 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3821 (rs6000_cpu_cpp_builtins): Ditto.
3822 * config/rs6000/rs6000-linux.c
3823 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3824 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3825 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3827 (rs6000_setup_reg_addr_masks): Ditto.
3828 (rs6000_init_hard_regno_mode_ok): Ditto.
3829 (rs6000_option_override_internal): Ditto.
3830 (legitimate_lo_sum_address_p): Ditto.
3831 (rs6000_legitimize_address): Ditto.
3832 (rs6000_legitimize_reload_address): Ditto.
3833 (rs6000_legitimate_address_p): Ditto.
3834 (abi_v4_pass_in_fpr): Ditto.
3835 (setup_incoming_varargs): Ditto.
3836 (rs6000_gimplify_va_arg): Ditto.
3837 (rs6000_split_multireg_move): Ditto.
3838 (rs6000_savres_strategy): Ditto.
3839 (rs6000_emit_prologue_components): Ditto.
3840 (rs6000_emit_epilogue_components): Ditto.
3841 (rs6000_emit_prologue): Ditto.
3842 (rs6000_emit_epilogue): Ditto.
3843 (rs6000_elf_file_end): Ditto.
3844 (rs6000_function_value): Ditto.
3845 (rs6000_libcall_value): Ditto.
3846 * config/rs6000/rs6000.h: Ditto.
3847 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3848 (TARGET_MINMAX): ... this. New.
3849 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3850 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3851 (*movsi_internal1_single): Delete.
3852 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3853 mfpu=, mxilinx-fpu): Delete.
3854 * config/rs6000/singlefp.h: Delete.
3855 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3856 * config/rs6000/t-rs6000: Ditto.
3857 * config/rs6000/t-xilinx: Delete.
3858 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3859 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3860 (VStype_simple): Delete.
3861 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3862 * config/rs6000/xfpu.h: Delete.
3863 * config/rs6000/xfpu.md: Delete.
3864 * config/rs6000/xilinx.h: Delete.
3865 * config/rs6000/xilinx.opt: Delete.
3866 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3867 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3869 2018-05-04 Tom de Vries <tom@codesourcery.com>
3872 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3875 2018-05-04 Richard Biener <rguenther@suse.de>
3878 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3879 (expand_complex_div_wide): Likewise.
3880 (expand_complex_operations_1): Likewise.
3881 (expand_complex_libcall): Preserve EH info of the original stmt.
3882 (tree_lower_complex): Handle removed blocks.
3883 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3884 on complex multiplication and division libcall builtins.
3886 2018-05-04 Richard Biener <rguenther@suse.de>
3889 * fold-const.c (negate_expr_p): Restrict negation of operand
3890 zero of a division to when we know that can happen without
3892 (fold_negate_expr_1): Likewise.
3894 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3897 * real.h (real_nextafter): Declare.
3898 * real.c (real_nextafter): New function.
3899 * fold-const-call.c (fold_const_nextafter): New function.
3900 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3901 CASE_CFN_NEXTTOWARD.
3902 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3903 even when arg1_mode is different from arg0_mode.
3905 2018-05-03 Nathan Sidwell <nathan@acm.org>
3907 * doc/extend.texi (Deprecated Features): Remove
3909 (Backwards Compatibility): Likewise.
3910 * doc/invoke.texi (C++ Language Options): Likewise.
3911 (C++ Dialect Options): Likewise.
3913 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3916 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3917 _mm512_mask_mullox_epi64): New intrinsics.
3919 2018-05-03 Tom de Vries <tom@codesourcery.com>
3922 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3923 dump files): Add offload-tree.
3925 2018-05-03 Richard Biener <rguenther@suse.de>
3927 PR tree-optimization/85615
3928 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3929 to loops not nested in BBs loop father to avoid creating multi-entry
3932 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3934 PR tree-optimization/70291
3935 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3936 arguments. Change return type to tree. Emit libcall as a new
3937 statement rather than replacing existing one when inplace_p is true.
3938 (expand_complex_multiplication_components): New function.
3939 (expand_complex_multiplication): Expand floating-point complex
3940 multiplication using the above.
3941 (expand_complex_division): Rename inner_type parameter to type.
3942 Update expand_complex_libcall call-site.
3943 (expand_complex_operations_1): Update expand_complex_multiplication
3944 and expand_complex_division call-sites.
3946 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3949 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3950 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3951 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3952 the highest significant bit of the shift count mask is clear. In
3953 check whether and[sq]i3 is needed verify that all significant bits
3954 of the shift count other than the highest are set.
3956 2018-05-02 Tom de Vries <tom@codesourcery.com>
3959 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3960 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3961 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3962 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3963 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3964 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3965 __builtin_goacc_parlevel_size.
3967 2018-05-02 Richard Biener <rguenther@suse.de>
3969 PR tree-optimization/85597
3970 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3971 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3974 2018-05-02 Tom de Vries <tom@codesourcery.com>
3977 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3978 dump files): Add ltrans-tree.
3980 2018-05-02 Tom de Vries <tom@codesourcery.com>
3983 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3984 dump files): Add wpa-ipa.
3986 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3988 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3989 powerpc*-*-linux*paired* target.
3990 * config/rs6000/750cl.h: Delete.
3991 * config/rs6000/paired.h: Delete.
3992 * config/rs6000/paired.md: Delete.
3993 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3995 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3996 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3997 comment. Remove paired float support.
3998 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3999 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4001 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4002 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4004 * config/rs6000/rs6000.c: Remove paired float support.
4005 (paired_expand_vector_init, paired_expand_vector_move,
4006 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4007 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4008 paired_expand_builtin, paired_expand_predicate_builtin,
4009 paired_init_builtins): Delete.
4010 * config/rs6000/rs6000.h: Remove paired float support.
4011 * config/rs6000/rs6000.md: Remove paired float support.
4012 (move_from_CR_ov_bit): Delete.
4013 * config/rs6000/rs6000.opt (mpaired): Delete.
4014 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4015 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4017 2018-05-02 Richard Biener <rguenther@suse.de>
4020 * gimplify.c (gimplify_save_expr): When in SSA form allow
4021 SAVE_EXPRs to compute to SSA vars.
4023 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4026 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4027 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4028 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4029 clobber operands[2], instead use a new pseudo. Formatting fixes.
4031 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4033 PR tree-optimization/85586
4034 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4035 exit early for statements in the same group if the accesses are
4038 2018-05-02 Tom de Vries <tom@codesourcery.com>
4041 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4044 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4046 PR tree-optimization/85143
4047 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4049 2018-05-01 Tom de Vries <tom@codesourcery.com>
4052 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4053 not found" error message.
4055 2018-05-01 Tom de Vries <tom@codesourcery.com>
4058 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4059 * vec.c (test_ordered_remove_if): New function.
4060 (vec_c_tests): Call test_ordered_remove_if.
4061 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4062 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4063 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4064 VEC_ORDERED_REMOVE_IF.
4066 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4068 PR tree-optimization/82665
4069 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4070 pointer subtraction where arguments come from a memchr call.
4072 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4074 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4075 --push-state --as-needed and --pop-state instead of --as-needed and
4076 --no-as-needed if ld supports it.
4077 * configure: Regenerated.
4080 * doc/install.texi2html: Replace _002d with - and _002a with * in
4081 generated html files using sed.
4083 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4086 * gcc-rich-location.c (blank_line_before_p): New function.
4087 (use_new_line): New function.
4088 (gcc_rich_location::add_fixit_insert_formatted): New function.
4089 * gcc-rich-location.h
4090 (gcc_rich_location::add_fixit_insert_formatted): New function.
4092 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4094 * selftest.c (assert_streq): Rename "expected" and "actual" to
4095 "val1" and "val2". Extend NULL-handling to cover both inputs
4096 symmetrically, while still requiring both to be non-NULL for a pass.
4097 * selftest.h (assert_streq): Rename "expected" and "actual" to
4099 (ASSERT_EQ): Likewise.
4100 (ASSERT_EQ_AT): Likewise.
4101 (ASSERT_KNOWN_EQ): Likewise.
4102 (ASSERT_KNOWN_EQ_AT): Likewise.
4103 (ASSERT_NE): Likewise.
4104 (ASSERT_MAYBE_NE): Likewise.
4105 (ASSERT_MAYBE_NE_AT): Likewise.
4106 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4107 the assertion to pass.
4108 (ASSERT_STREQ_AT): Likewise.
4110 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4112 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4113 interaction with -pie.
4115 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4117 * selftest.h: Fix alphabetization of per-source-file selftest
4120 2018-04-30 Jason Merrill <jason@redhat.com>
4122 PR c++/61982 - dead stores to destroyed objects.
4123 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4126 2018-04-30 Jason Merrill <jason@redhat.com>
4128 * tree.c (build_clobber): New.
4129 * tree.h: Declare it.
4130 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4132 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4134 * diagnostic-show-locus.c (layout::layout): Update for
4135 location_get_source_line returning a char_span.
4136 (struct char_span): Move to input.h.
4137 (struct correction): Update for fields in char_span becoming
4139 (struct source_line): Update for location_get_source_line
4140 returning a char_span.
4141 (layout::print_line): Likewise.
4142 * edit-context.c (edited_file::print_content): Likewise.
4143 (edited_file::print_diff_hunk): Likewise.
4144 (edited_file::print_run_of_changed_lines): Likewise.
4145 (edited_file::get_num_lines): Likewise.
4146 (edited_line::edited_line): Likewise.
4147 * final.c (asm_show_source): Likewise.
4148 * input.c (location_get_source_line): Convert return type
4149 from const char * to char_span, losing the final "line_len"
4151 (dump_location_info): Update for the above.
4152 (get_substring_ranges_for_loc): Likewise. Use a char_span
4153 when handling the literal within the line.
4154 (test_reading_source_line): Update for location_get_source_line
4155 returning a char_span.
4156 * input.h (class char_span): Move here from
4157 diagnostic-show-locus.c, converting from a struct to a class.
4158 Make data members private.
4159 (char_span::operator bool): New.
4160 (char_span::length): New.
4161 (char_span::get_buffer): New.
4162 (char_span::operator[]): New.
4163 (char_span::subspan): Make const.
4164 (char_span::xstrdup): New.
4165 (location_get_source_line): Convert return type from const char *
4166 to char_span, losing the final "line_size" param.
4168 2018-04-30 Jan Hubicka <jh@suse.cz>
4170 * lto-wrapper.c (ltrans_priorities): New static var.
4171 (cmp_priority): New.
4172 (run_gcc): Read priorities and if doing parallel build order
4173 the Makefile by them.
4175 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4177 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4179 2018-04-30 Richard Biener <rguenther@suse.de>
4181 * tree-cfg.c (verify_address): Remove base argument, add
4182 flag whether to check TREE_ADDRESSABLE and do that.
4183 (verify_expr): Remove.
4184 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4185 (verify_gimple_assign_single): Likewise.
4186 (verify_gimple_switch): Likewise.
4187 (verify_expr_location_1): Dereference tp once. Add (disabled)
4188 piece from verify_expr.
4189 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4191 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4193 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4195 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4197 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4198 (small_data_pattern): Likewise.
4199 (arc_rewrite_small_data): Likewise.
4200 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4201 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4202 (get_symbol_alignment): New function.
4203 (legitimate_small_data_address_p): Likewise.
4204 (legitimate_scaled_address): Update, call
4205 legitimate_small_data_address_p.
4206 (output_sdata): New static variable.
4207 (arc_print_operand): Update how we handle small data operands.
4208 (arc_print_operand_address): Likewise.
4209 (arc_legitimate_address_p): Update, use
4210 legitimate_small_data_address_p.
4211 (arc_rewrite_small_data_p): Remove.
4212 (arc_rewrite_small_data_1): Likewise.
4213 (arc_rewrite_small_data): Likewise.
4214 (small_data_pattern): Likewise.
4215 (compact_sda_memory_operand): Update to use
4216 legitimate_small_data_address_p and get_symbol_alignment.
4217 (prepare_move_operands): Don't rewite sdata pattern.
4218 (prepare_extend_operands): Remove.
4219 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4221 (zero_extendqisi2): Likewise.
4222 (zero_extendhisi2): Likewise.
4223 (extendqihi2): Likewise.
4224 (extendqisi2): Likewise.
4225 (extendhisi2): Likewise.
4229 * config/arc/constraints.md (Usd): Change it to memory constraint.
4231 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4233 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4234 as source of std instructions.
4235 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4236 allow 6-bit constants as source for store instructions.
4237 (movdi_insn): Update instruction pattern to allow 6-bit constants
4238 as source for store instructions.
4240 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4242 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4244 2018-04-30 Nathan Sidwell <nathan@acm.org>
4245 Sandra Loosemore <sandra@codesourcery.com>
4247 * dumpfile.c (dump_open): Allow '-' for stdout.
4248 * doc/invoke.texi (Developer Options): Document dump filename
4249 determination early. Document stdin/stdout selection.
4251 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4253 Microblaze Target: PIC data text relative
4255 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4256 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4258 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4259 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4260 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4261 New addressing mode for data-text relative position indepenedent code.
4262 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4263 'ADDRESS_SYMBOLIC_TXT_REL'.
4264 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4265 (microblaze_legitimate_pic_operand): Exclude function calls from
4266 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4267 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4269 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4270 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4271 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4272 for 'address + offset'.
4273 (microblaze_expand_prologue): Add new function prologue call for
4275 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4276 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4277 table in case of TARGET_PIC_DATA_TEXT_REL.
4278 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4279 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4280 Add new macros 'UNSPEC_TEXT',
4281 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4282 + exclude function calls from 'UNSPEC_PLT' in case of data text
4284 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4285 new target hook for generating address diff vector tables in case of
4287 * doc/tm.texi : Regenerate.
4288 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4289 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4290 of addr diff vector generation.
4291 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4292 target hook definition.
4293 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4294 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4295 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4296 option for data text relative.
4298 2018-04-30 Richard Biener <rguenther@suse.de>
4300 * tree-chrec.h (evolution_function_is_constant_p): Remove
4302 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4304 2018-04-30 Richard Biener <rguenther@suse.de>
4307 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4309 2018-04-30 Richard Biener <rguenther@suse.de>
4311 PR tree-optimization/28364
4312 PR tree-optimization/85275
4313 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4314 copying first exit test.
4316 2018-04-28 Mark Wielaard <mark@klomp.org>
4318 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4320 (dwarf_AT): Handle DW_AT_addr_base.
4321 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4323 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4326 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4327 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4328 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4329 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4331 2018-04-28 Richard Biener <rguenther@suse.de>
4333 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4334 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4335 to reflect use. Only add interesting stmts.
4337 2018-04-27 Martin Jambor <mjambor@suse.cz>
4340 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4341 the jump function allows for passing through aggregate values.
4343 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4345 * input.h (in_system_header_at): Convert from macro to inline
4347 (from_macro_expansion_at): Likewise.
4348 (from_macro_definition_at): Likewise.
4350 2018-04-27 Jeff Law <law@redhat.com>
4352 * config.gcc: Mark tile* targets as deprecated/obsolete.
4354 2018-04-27 Richard Biener <rguenther@suse.de>
4356 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4359 2018-04-27 Richard Biener <rguenther@suse.de>
4361 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4363 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4365 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4366 with Yd constraint. Set "preferred_for_speed" attribute from
4367 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4369 (*movdi_internal): Ditto.
4370 (movti_interunit splitters): Remove
4371 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4372 (movdi_interunit splitters): Ditto.
4373 * config/i386/constraints.md (Ye): Remove.
4374 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4376 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4379 * config/aarch64/constraints.md (Usg): Limit to 31.
4382 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4384 PR tree-optimization/85529
4385 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4386 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4387 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4388 zero extension or masking of the MSB bit.
4389 (optimize_range_tests): Add FIRST_BB argument, pass it through
4390 to optimize_range_tests_var_bound.
4391 (maybe_optimize_range_tests, reassociate_bb): Adjust
4392 optimize_range_tests callers.
4394 2018-04-26 Richard Biener <rguenther@suse.de>
4395 Jakub Jelinek <jakub@redhat.com>
4397 * cgraph.h (symbol_table): Just declare debug method here.
4398 * symtab.c (symbol_table::debug): Define.
4400 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4402 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4404 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4406 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4407 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4408 (*movdi_internal): Substitute Yi and Yj constraint with x
4409 and Ym and Yn constraint with y constraint. Update "isa"
4410 attribute and set "preferred_for_speed" attribute from
4411 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4412 (*movsi_internal): Ditto.
4413 (*movdf_internal): Ditto.
4414 (*movsf_internal): Ditto.
4415 (*zero_extendsidi2): Ditto.
4416 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4417 (sse2_loadld): Ditto.
4418 (*vec_extract<ssevecmodelower>_0): Ditto.
4419 (*vec_extractv4si_0_zext_sse4): Ditto.
4420 (vec_concatv2di): Ditto.
4421 (*vec_dup<mode>): Ditto.
4422 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4423 * config/i386/constraints.md (Yi): Remove.
4428 2018-04-26 Nathan Sidwell <nathan@acm.org>
4430 * dumpfile.c (dump_open): New.
4431 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4432 (dump_finish): Detect stdio/stderr by value not name.
4434 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4436 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4438 2018-04-26 Tom de Vries <tom@codesourcery.com>
4441 * config/nvptx/nvptx.c (verify_neutering_jumps)
4442 (verify_neutering_labels): New function
4443 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4445 2018-04-26 Tom de Vries <tom@codesourcery.com>
4448 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4449 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4452 2018-04-26 Richard Biener <rguenther@suse.de>
4453 Tom de Vries <tom@codesourcery.com>
4456 * lto-streamer-out.c (output_function): Fixup loops if required to match
4457 discovery done in the reader.
4459 2018-04-26 Richard Biener <rguenther@suse.de>
4461 PR tree-optimization/85116
4462 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4463 have a loop exit from the single latch predecessor. Remove
4464 case of header with just condition.
4465 (ch_base::copy_headers): Exclude infinite loops from any
4467 (pass_ch::execute): Record exits.
4469 2018-04-26 Richard Biener <rguenther@suse.de>
4471 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4472 prologue cost vector and pass it to vect_get_load_cost.
4473 (vect_get_peeling_costs_all_drs): Likewise.
4474 (vect_peeling_hash_get_lowest_cost): Likewise.
4475 (vect_enhance_data_refs_alignment): Likewise.
4477 2018-04-26 Richard Biener <rguenther@suse.de>
4480 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4481 checking of integer<->pointer conversions.
4482 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4483 sign-/zero-extending pointer types.
4484 (expand_omp_for_static_chunk): Likewise.
4486 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4487 Jean Lee <xiaoyur347@gmail.com>
4489 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4490 (TARGET_ASAN_SHADOW_OFFSET): Define.
4491 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4492 true for -fsanitize=address.
4494 2018-04-25 Mark Wielaard <mark@klomp.org>
4496 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4499 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4501 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4502 than "alu", remove explicit "memory" and "imm_disp" attributes.
4503 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4506 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4507 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4508 gen_lowpart_no_emit.
4510 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4513 * config/i386/i386.c (ix86_expand_builtin): Change memory
4514 operand to XI, extend p0 to Pmode.
4515 * config/i386/i386.md: Change unspec volatile and operand
4516 1 mode to XI, change operand 0 mode to P.
4518 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4520 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4521 GET_MODE_MASK before any checking.
4522 (nds32_can_use_bset_p): Likewise.
4523 (nds32_can_use_btgl_p): Likewise.
4525 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4527 * config/nds32/nds32-doubleword.md: New define_split pattern for
4528 illegal register number.
4530 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4532 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4534 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4536 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4538 2018-04-25 Richard Biener <rguenther@suse.de>
4540 * lto-streamer.h (LTO_major_version): Bump to 8.
4542 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4544 * BASE-VER: Set to 9.0.0.
4546 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4548 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4549 in __abskf2 and __powikf2.
4551 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4554 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4555 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4556 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4557 Use the above on operand 2. Reindent.
4558 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4560 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4563 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4565 * config/i386/i386.opt (mcet): Removed.
4566 * doc/install.texi: Remove -mcet documentation.
4567 * doc/invoke.texi: Likewise.
4569 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4572 * doc/install.texi: Remove -mcet from bootstrap-cet.
4574 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4577 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4578 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4582 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4583 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4584 containing a CONST_VECTOR.
4586 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4588 * doc/install.texi: Update newlib dependency for nvptx.
4590 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4593 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4594 instead of INTVAL when shifting x left.
4596 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4598 PR tree-optimization/85478
4599 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4600 vect_grouped_store_supported for single element vectors.
4602 2018-04-24 Richard Biener <rguenther@suse.de>
4605 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4606 load cost increase to the case of non-constant step.
4608 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4611 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4612 destination if any_malformed_asm.
4614 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4617 * expr.c (store_field): In the bitfield case, if the value comes from
4618 a function call and is returned in registers by means of a PARALLEL,
4619 do not change the mode of the temporary unless BLKmode and VOIDmode.
4621 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4623 PR rtl-optimization/85423
4624 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4625 dependencies to debug insns when the previous insn is non-debug.
4627 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4629 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4630 enums into a single definition.
4631 (fls): Fix predicates and printing.
4634 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4636 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4637 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4638 and short u6 immediate.
4639 (check_if_valid_sleep_operand): Remove.
4640 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4642 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4644 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4645 flag_always_save_lp condition.
4646 * config/nds32/nds32.opt (malways-save-lp): New option.
4648 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4650 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4651 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4652 * config/nds32/nds32.h
4653 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4654 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4656 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4658 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4659 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4661 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4662 Chung-Ju Wu <jasonwucj@gmail.com>
4664 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4666 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4667 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4669 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4671 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4673 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4675 * config/nds32/nds32-protos.h (nds32_data_alignment,
4676 nds32_local_alignment): Declare.
4677 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4678 nds32_local_alignment): New functions.
4679 (TARGET_CONSTANT_ALIGNMENT): Define.
4680 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4682 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4684 * config/nds32/nds32.c
4685 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4686 (TARGET_MODES_TIEABLE_P): Likewise.
4688 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4690 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4693 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4694 Chung-Ju Wu <jasonwucj@gmail.com>
4696 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4697 for unaligned access.
4698 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4700 * config/nds32/nds32-intrinsic.md: Likewise.
4701 * config/nds32/nds32_intrinsic.h: Likewise.
4702 * config/nds32/nds32.h (nds32_builtins): Likewise.
4703 * config/nds32/nds32.opt (munaligned-access): New option.
4704 * config/nds32/nds32.c (nds32_asm_file_start): Display
4705 flag_unaligned_access status.
4707 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4709 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4710 -mno-relax is present.
4711 * config/riscv/linux.h (LINK_SPEC): Ditto.
4713 2018-04-20 Martin Sebor <msebor@redhat.com>
4716 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4718 (gimple_fold_builtin_stxcpy_chk): Same.
4719 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4721 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4724 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4725 __powikf2 when long double is IEEE 128-bit.
4727 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4729 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4730 step to make sure stack always aligned.
4732 2018-04-20 Carl Love <cel@us.ibm.com>
4735 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4736 size check for arg0.
4738 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4739 Tom de Vries <tom@codesourcery.com>
4742 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4743 Emit insns for calls too.
4744 (nvptx_find_par): Always look for worker-level predecessor insn.
4745 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4747 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4748 (nvptx_process_pars): Propagate frames for calls.
4750 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4753 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4755 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4756 (ix86_handle_option): Don't handle OPT_mibt.
4757 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4759 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4761 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4762 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4763 (ix86_target_macros): Define __CET__ with flag_cf_protection
4764 for -fcf-protection.
4765 * config/i386/i386.c (isa2_opts): Remove -mibt.
4766 * config/i386/i386.h (TARGET_IBT): Removed.
4767 (TARGET_IBT_P): Likewise.
4768 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4769 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4770 * config/i386/i386.opt (mcet): Update help message.
4773 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4774 -mcet as an alias for -mshstk.
4776 2018-04-20 Richard Biener <rguenther@suse.de>
4779 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4780 complexity by forcing a single use of the multiply operand.
4782 2018-04-20 Martin Jambor <mjambor@suse.cz>
4785 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4786 recursion dependency to only apply to non-clones.
4788 2018-04-20 Martin Jambor <mjambor@suse.cz>
4791 * ipa-cp.c (create_specialized_node): Check that clones of
4792 self-recursive edges exist during IPA-CP.
4794 2018-04-19 Toon Moene <toon@moene.org>
4796 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4799 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4801 PR tree-optimization/85467
4802 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4803 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4804 VECTOR_CST element to type.
4806 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4809 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4810 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4811 (builtin_longjmp): Likewise.
4812 (save_stack_nonlocal): New pattern.
4813 (restore_stack_nonlocal): Likewise.
4815 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4818 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4819 Replace ASM_OUTPUT_LABEL with fprintf.
4821 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4824 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4825 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4826 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4827 define __IBT__ and __SHSTK__ for -fcf-protection.
4828 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4830 (ix86_trampoline_init): Likewise.
4831 (x86_output_mi_thunk): Likewise.
4832 (ix86_notrack_prefixed_insn_p): Likewise.
4833 (ix86_option_override_internal): Don't disallow -fcf-protection.
4834 * config/i386/i386.md (rdssp<mode>): Also enable for
4836 (incssp<mode>): Likewise.
4837 (nop_endbr): Likewise.
4838 * config/i386/i386.opt (mcet): Change help message to built-in
4842 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4843 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4844 enable CET built-in functions.
4846 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4848 * common/config/i386/i386-common.c
4849 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4850 OPTION_MASK_ISA_MOVDIRI_UNSET,
4851 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4852 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4853 * config.gcc (movdirintrin.h): New header.
4854 * config/i386/cpuid.h (bit_MOVDIRI,
4855 bit_MOVDIR64B): New bits.
4856 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4858 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4859 (VOID, PVOID, PCVOID)): New function types.
4860 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4861 __builtin_ia32_directstoreu_u64,
4862 __builtin_ia32_movdir64b): New builtins.
4863 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4864 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4866 (ix86_valid_target_attribute_inner_p): Ditto.
4867 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4868 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4869 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4870 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4871 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4872 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4873 (movdiri<mode>, movdir64b_<mode>): New.
4874 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4875 * config/i386/immintrin.h: Include movdirintrin.h.
4876 * config/i386/movdirintrin.h: New file.
4877 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4879 2018-04-19 Richard Biener <rguenther@suse.de>
4882 * cfg.c (clear_bb_flags): When loop state says we have
4883 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4885 2018-04-19 Richard Biener <rguenther@suse.de>
4887 PR tree-optimization/84737
4888 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4889 copying restrict info.
4890 (vect_setup_realignment): Use it.
4891 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4892 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4893 the first DR to all generated stores.
4894 (vectorizable_load): Likewise for loads.
4896 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4898 PR tree-optimization/85446
4899 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4900 the integral and pointer types to have the same precision.
4902 * doc/install.texi: Document --disable-cet being the default and
4905 2018-04-18 Martin Liska <mliska@suse.cz>
4907 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4910 2018-04-18 Martin Liska <mliska@suse.cz>
4913 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4916 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4917 arguments if they are comparable.
4919 2018-04-18 Martin Liska <mliska@suse.cz>
4922 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4925 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4928 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4931 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4932 ENDBR after calling __morestack.
4934 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4937 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4938 by using gcc_base_ver to generate a gcc_driver_version, and use
4939 it when generating GCC_DRIVER_NAME.
4940 * configure: Regenerate.
4942 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4945 * config.gcc: Obsolete powerpc*-*-*spe*.
4947 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4950 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4951 (stabstr_D): Change type of unum from unsigned int to
4952 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4955 2018-04-17 Jim Wilson <jimw@sifive.com>
4958 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4959 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4960 Set arg_pointer_offset after using pretend_args_size.
4962 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4964 PR rtl-optimization/85431
4965 * dse.c (record_store): Ignore zero width stores.
4968 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4969 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4970 __builtin_stack_restore rather than after it.
4971 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4972 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4973 argument instead of virtual_dynamic_stack_rtx.
4975 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4977 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4979 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4980 Add note to error message to explain internal mapping of overloaded
4981 built-in function name to non-overloaded built-in function name.
4982 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4985 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4988 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4989 where the inputs overlap with the output.
4991 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4993 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4994 (=v, v) alternative and explicit "memory" attribute.
4995 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4996 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4998 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4999 "sselog1" type instead of "sselog".
5000 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5001 "sselog". Remove explicit "memory" attribute.
5002 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5003 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5005 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5006 "sselog1" type instead of "sselog". Remove explicit "memory"
5008 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5009 use "sselog1" type instead of "sselog". Remove explicit "memory"
5011 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5012 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5014 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5015 "sselog1" type instead of "sselog". Remove explicit "memory"
5017 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5018 use "sselog1" type instead of "sselog". Remove explicit "memory"
5022 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5025 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5028 2018-04-17 Martin Jambor <mjambor@suse.cz>
5031 * ipa-cp.c (create_specialized_node): Call
5032 expand_all_artificial_thunks if necessary.
5034 2018-04-17 Martin Liska <mliska@suse.cz>
5037 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5038 in message, remote space in between '_G' and '('.
5040 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5043 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5044 avx512f_vmcmp<mode>3<round_saeonly_name>,
5045 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5046 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5047 avx512f_rndscale<mode><round_saeonly_name>,
5048 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5049 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5050 Use %<iptr>2 instead of %2 for -masm=intel.
5051 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5052 avx512f_vcvttss2usi<round_saeonly_name>,
5053 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5055 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5056 avx512f_vcvttsd2usi<round_saeonly_name>,
5057 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5058 Use %q1 instead of %1 for -masm=intel.
5059 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5060 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5061 of %3 for -masm=intel.
5062 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5064 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5066 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5068 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5069 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5071 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5073 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5074 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5075 %g1 and one with %0 and %1.
5076 (avx512er_vmrcp28<mode><round_saeonly_name>,
5077 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5079 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5080 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5081 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5082 of %0 and %{%4%} for -masm=intel.
5083 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5084 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5085 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5086 order of %0 and %{%5%}%{z%} for -masm=intel.
5088 2018-04-17 Jan Hubicka <jh@suse.cz>
5091 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5093 2018-04-17 Martin Liska <mliska@suse.cz>
5096 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5097 for target_clone error message. Make default implementation
5098 clone to be a local declaration.
5099 (separate_attrs): Add new argument and check for an empty
5101 (expand_target_clones): Handle it.
5102 (ipa_target_clone): Make redirection just for target_clones
5105 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5106 Tom de Vries <tom@codesourcery.com>
5109 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5110 tiled basic blocks without omp continue statements.
5112 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5115 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5116 vec_extract expression as having side effects to make sure it gets
5119 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5122 * config/i386/i386.c (get_builtin_code_for_version): Check
5125 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5128 * gcc/config.gcc: Support "skylake".
5129 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5131 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5132 (processor_target_table): Add "skylake".
5133 (ix86_option_override_internal): Add "skylake".
5134 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5135 PROCESSOR_CANNONLAKE.
5136 (get_builtin_code_for_version): Fix priority for
5137 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5138 PROCESSOR_SKYLAKE-AVX512.
5139 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5140 (processor_type): Add PROCESSOR_SKYLAKE.
5142 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5143 Jason Merrill <jason@redhat.com>
5146 * convert.c (convert_to_integer_1): Use direct recursion for
5147 enumeral types and types with a precision less than the number
5148 of bits in their mode.
5150 2018-04-16 Julia Koval <julia.koval@intel.com>
5153 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5154 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5156 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5159 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5160 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5161 and -mno-direct-move.
5163 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5166 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5167 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5168 Ensure negative shifts result in {0}.
5170 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5172 PR rtl-optimization/79916
5173 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5174 regs (if any) to define how to gnerate SD moves when LRA is in
5177 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5179 PR rtl-optimization/85393
5180 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5181 * except.c (expand_dw2_landing_pad_for_region): Make static.
5182 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5183 a label and unconditional jump to old_bb, rather than
5184 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5187 PR rtl-optimization/85376
5188 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5189 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5190 instead of a specific value.
5192 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5193 Bin Cheng <bin.cheng@arm.com>
5195 PR tree-optimization/82965
5196 PR tree-optimization/83991
5197 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5198 by_profile_only parameter.
5199 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5200 information if the loop was predicted to iterate too many times.
5201 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5203 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5206 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5209 2018-04-13 Martin Liska <mliska@suse.cz>
5210 Jakub Jelinek <jakub@redhat.com>
5213 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5214 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5215 * builtins.c (expand_builtin_memory_copy_args): Use
5216 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5217 handle dest_addr == pc_rtx.
5219 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5222 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5223 asked to not generate direct moves.
5224 (fix_trunc<mode>si2_stfiwx): Similar.
5225 (fix_trunc<mode>si2_internal): Similar.
5227 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5230 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5231 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5232 lookup if dest in some wider mode is known to be const0_rtx and
5233 if so, record permanent equivalence for it to be ZERO_EXTEND of
5234 the narrower mode destination.
5236 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5238 * lto-streamer-out.c (output_function): Revert 259346.
5239 * omp-expand.c (expand_oacc_for): Likewise.
5241 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5243 PR rtl-optimization/85354
5244 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5245 * sel-sched.c (sel_global_init): ... here.
5247 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5250 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5251 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5252 mode for PE-COFF targets.
5253 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5254 (i386_pe_asm_lto_end): Likewise.
5255 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5256 (TARGET_ASM_LTO_END): Likewise.
5257 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5258 (i386_pe_asm_lto_start): New function.
5259 (i386_pe_asm_lto_end): Likewise.
5261 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5262 Richard Biener <rguenther@suse.de>
5265 * lto-streamer-out.c (output_function): Fix CFG loop state before
5267 * omp-expand.c (expand_oacc_for): Handle calls to internal
5268 functions like regular functions.
5270 2018-04-12 Richard Biener <rguenther@suse.de>
5273 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5274 for the early LTO debug to properly generate references to it
5275 during DIE emission. Do not re-use that for the skeleton for
5277 (dwarf2out_early_finish): Likewise.
5279 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5282 * config/i386/sse.md
5283 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5284 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5285 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5286 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5287 and output is a reg, avoid creating invalid lowpart subreg, but
5288 instead split into a 512-bit move. Don't split if not AVX512VL,
5289 input is xmm16+ reg and output is a mem.
5290 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5291 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5292 xmm16+ reg and output is a mem.
5294 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5296 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5297 also for flag_dwarf2_cfi_asm.
5299 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5301 PR rtl-optimization/85342
5302 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5303 a bool scalar var inside of the loop instead. Don't try to update
5304 recog_data.operand after failed apply_change_group.
5306 2018-04-12 Tom de Vries <tom@codesourcery.com>
5309 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5310 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5311 array with flexible array member as array without given dimension.
5312 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5313 argument for undefined param to true.
5315 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5318 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5319 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5320 from PowerPC section.
5321 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5322 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5323 help text that is too long.
5324 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5325 help text that is too long.
5326 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5327 help text that is too long.
5329 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5331 * config/alpha/alpha.md (stack_probe_internal): Rename
5332 from "probe_stack". Update all callers.
5334 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5336 PR rtl-optimization/84566
5337 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5338 sched_macro_fuse_insns.
5340 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5343 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5344 (compute_block_dependences): ... from here.
5346 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5348 PR tree-optimization/85331
5349 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5350 from int to HOST_WIDE_INT.
5352 2018-04-11 Martin Jambor <mjambor@suse.cz>
5355 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5356 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5357 not the same as the source val.
5358 (cgraph_edge_brings_value_p): New parameter.
5359 (gather_edges_for_value): Pass destination value to
5360 cgraph_edge_brings_value_p.
5361 (perhaps_add_new_callers): Likewise.
5362 (get_info_about_necessary_edges): Likewise and exclude values brought
5363 only by self-recursive edges.
5364 (create_specialized_node): Redirect only clones of self-calling edges.
5365 (+self_recursive_pass_through_p): New function.
5366 (find_more_scalar_values_for_callers_subset): Use it.
5367 (find_aggregate_values_for_callers_subset): Likewise.
5368 (known_aggs_to_agg_replacement_list): Removed.
5369 (decide_whether_version_node): Re-calculate known constants for all
5370 remaining context clones.
5372 2018-04-11 Richard Biener <rguenther@suse.de>
5375 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5376 from early DWARF output.
5377 (dwarf2out_early_finish): Output line info unconditionally into
5378 early DWARF and add reference to it.
5380 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5383 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5384 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5385 other than V2DFmode using iptr mode attribute.
5386 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5388 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5390 PR rtl-optimization/84659
5391 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5393 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5396 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5398 (output_loc_list): Pass address of a dummy size variable even in the
5399 locview handling loop.
5400 (index_location_lists): Add comment on why skip_loc_list_entry can't
5403 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5406 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5409 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5412 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5413 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5414 and -mstring-compare-inline-limit.
5416 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5419 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5420 for stack clash protection in a register whenever we need it to be in
5423 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5425 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5426 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5428 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5431 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5433 (mlong-double-): Ditto.
5434 * config/rs6000/sysv4.opt (msdata=): Ditto.
5435 (mtls-size=): Ditto.
5437 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5439 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5440 erroneous entries for
5441 "vector int vec_ldl (int, long int *)", and
5442 "vector unsigned int vec_ldl (int, unsigned long int *)".
5443 Add comments and entries for
5444 "vector bool char vec_ldl (int, bool char *)",
5445 "vector bool short vec_ldl (int, bool short *)",
5446 "vector bool int vec_ldl (int, bool int *)",
5447 "vector bool long long vec_ldl (int, bool long long *)",
5448 "vector pixel vec_ldl (int, pixel *)",
5449 "vector long long vec_ldl (int, long long *)",
5450 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5451 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5452 type tree bool_long_long_type_node and correct definition of
5453 bool_V2DI_type_node to make reference to this new type tree.
5454 (rs6000_mangle_type): Replace erroneous reference to
5455 bool_long_type_node with bool_long_long_type_node.
5456 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5457 comments to emphasize sign distinctions for char and int types and
5458 replace RS6000_BTI_bool_long constant with
5459 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5460 use of RS6000_BTI_pixel.
5461 (bool_long_type_node): Remove this macro definition.
5462 (bool_long_long_type_node): New macro definition
5464 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5466 PR rtl-optimization/85300
5467 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5468 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5469 simplify_unary_operation fails.
5471 2018-04-10 Martin Liska <mliska@suse.cz>
5473 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5474 cgraph_edge and ipa_ref.
5476 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5480 * config/i386/sse.md
5481 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5482 computation of the VEC_MERGE selector from mask.
5483 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5484 Fix decoding of the VEC_MERGE selector into mask.
5486 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5488 PR tree-optimization/85286
5489 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5491 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5493 * final.c (final_1): Set insn_last_address as well as
5494 insn_current_address.
5496 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5499 * explow.c (emit_stack_probe): Call validize_mem on memory location
5500 before passing it to gen_probe_stack. Create address operand and
5501 legitimize it for the probe_stack_address case.
5503 2018-04-09 Jan Hubicka <jh@suse.cz>
5506 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5507 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5508 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5510 (free_lang_data): Rebuild type inheritance graph.
5512 2018-04-09 Martin Sebor <msebor@redhat.com>
5514 * invoke.texi (-finline-small-functions): Mention other optimization
5516 (-findirect-inlining, -fpartial-inlining): Same.
5517 (-finline-functions-called-once): Same.
5518 (-freorder-blocks-and-partition): Same.
5520 2018-04-09 Jan Hubicka <jh@suse.cz>
5523 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5524 jumps; choose last target that matches the criteria (i.e.
5525 no partition changes for non-crossing jumps).
5526 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5527 support for redirecting crossing jumps to non-crossing.
5529 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5531 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5532 also for naked functions.
5534 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5536 * config/arc/arc.md (add_shift): New pattern.
5537 (add_shift2): Likewise.
5538 (sub_shift): Likewise.
5539 (sub_shift_cmp0_noout): Likewise.
5540 (compare_si_ashiftsi): Likewise.
5541 (xbfu_cmp0_noout): New combine pattern.
5542 (xbfu_cmp0"): Likewise.
5543 (movsi_set_cc_insn): Place the predicable variant first.
5544 (commutative_binary_cmp0_noout): Remove clobber.
5545 (commutative_binary_cmp0): New pattern.
5546 (noncommutative_binary_cmp0): Likewise.
5547 (noncommutative_binary_cmp0_noout): Likewise.
5548 (noncommutative_binary_comparison_result_used): Removed.
5549 (rsub_cmp0): New pattern.
5550 (rsub_cmp0_noout): Likewise.
5551 (extzvsi): Changed, keep only meaningful variants.
5552 (SQH, SEZ): New iterators.
5553 (SQH_postfix): New mode attribute.
5554 (SEZ_prefix): New code attribute.
5555 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5556 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5557 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5559 (noncommutative_operator): Check the availability of barrel
5562 2018-04-09 Richard Biener <rguenther@suse.de>
5564 PR tree-optimization/85284
5565 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5566 Only use the niter constraining form of simple_iv when the exit
5569 2018-04-09 Tom de Vries <tom@codesourcery.com>
5572 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5573 (define_expand "*memory_barrier"): New define_expand.
5574 (define_insn "memory_barrier"): New insn.
5576 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5578 PR rtl-optimization/80463
5579 PR rtl-optimization/83972
5580 PR rtl-optimization/83480
5582 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5583 correct producer for the insn.
5584 (tidy_control_flow): Fixup seqnos in case of debug insns.
5586 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5588 PR rtl-optimization/83913
5590 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5591 different sched-times when merging exprs.
5593 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5595 PR rtl-optimization/83962
5597 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5598 tidy_fallthru_edge and tidy_control_flow.
5600 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5602 PR rtl-optimization/83530
5604 * sel-sched.c (force_next_insn): New global variable.
5605 (remove_insn_for_debug): When force_next_insn is true, also leave only
5606 next insn in the ready list.
5607 (sel_sched_region): When the region wasn't scheduled, make another pass
5608 over it with force_next_insn set to 1.
5610 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5612 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5614 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5615 for interrupt control.
5616 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5617 functions for interrupt control.
5618 * config/nds32/nds32-intrinsic.md: Likewise.
5619 * config/nds32/nds32_intrinsic.h: Likewise.
5620 * config/nds32/nds32.h (nds32_builtins): Likewise.
5622 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5624 * config/nds32/nds32.c (nds32_init_machine_status,
5625 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5626 strict_aligned_p field.
5627 (nds32_expand_to_rtl_hook): New function.
5628 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5629 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5631 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5632 Chung-Ju Wu <jasonwucj@gmail.com>
5634 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5635 * config/nds32/nds32-n7.md: New file.
5636 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5637 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5639 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5640 * config/nds32/nds32.md (pipeline_model): Add n7.
5641 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5642 * config/nds32/pipelines.md: Include n7 settings.
5644 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5645 Chung-Ju Wu <jasonwucj@gmail.com>
5647 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5648 * config/nds32/nds32-e8.md: New file.
5649 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5650 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5652 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5653 * config/nds32/nds32.md (pipeline_model): Add e8.
5654 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5655 * config/nds32/pipelines.md: Include e8 settings.
5657 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5658 Chung-Ju Wu <jasonwucj@gmail.com>
5660 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5661 * config/nds32/nds32-n8.md: New file.
5662 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5663 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5665 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5666 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5667 * config/nds32/nds32.md (pipeline_model): Add n8.
5668 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5669 * config/nds32/pipelines.md: Include n8 settings.
5671 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5672 Chung-Ju Wu <jasonwucj@gmail.com>
5674 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5675 * config/nds32/nds32-n9-2r1w.md: New file.
5676 * config/nds32/nds32-n9-3r2w.md: New file.
5677 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5678 nds32_register_ports): New or modify for cpu n9.
5679 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5681 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5682 * config/nds32/nds32-utils.c: New file.
5683 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5684 TARGET_MUL_SLOW): Define.
5685 * config/nds32/nds32.md (pipeline_model): New attribute.
5686 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5687 New options that support cpu n9.
5688 * config/nds32/pipelines.md: Include n9 settings.
5689 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5691 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5693 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5694 information if necessary.
5695 (output_cond_branch_compare_zero): Likewise.
5696 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5697 (nds32_target_alignment): Refine for alignment.
5698 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5699 (FUNCTION_BOUNDARY): Modify.
5700 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5702 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5704 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5706 * config/nds32/constants.md (unspec_volatile_element): Add values for
5707 TLB operation and data prefetch.
5708 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5709 functions for TLB operation and data prefetch.
5710 * config/nds32/nds32-intrinsic.md: Likewise.
5711 * config/nds32/nds32_intrinsic.h: Likewise.
5712 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5713 (nds32_print_operand): Likewise.
5714 * config/nds32/nds32.h (nds32_builtins): Likewise.
5716 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5717 Andrew Pinski <pinsika@gcc.gnu.org>
5720 * match.pd: Use constant_boolean_node of correct type instead of
5721 boolean_true_node or boolean_false_node for simplifying
5722 pointer comparisons to zero.
5724 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5726 PR tree-optimization/80021
5727 * tree.c (verify_type_variant): Make error call in verify_variant_match
5728 translatable and remove final full stop.
5730 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5732 * config/nds32/constants.md (unspec_volatile_element): Add
5733 UNSPEC_VOLATILE_EH_RETURN.
5734 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5735 nds32_output_stack_pop): Support dwarf exception handling process.
5736 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5737 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5738 exception handling process.
5739 (nds32_compute_stack_frame): Likewise.
5740 (nds32_return_addr_rtx): Likewise.
5741 (nds32_initial_elimination_offset): Likewise.
5742 (nds32_expand_prologue): Likewise.
5743 (nds32_expand_epilogue): Likewise.
5744 (nds32_dynamic_chain_address): New function.
5745 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5747 (DYNAMIC_CHAIN_ADDRESS): Define.
5748 (EH_RETURN_DATA_REGNO): Define.
5749 (EH_RETURN_STACKADJ_RTX): Define.
5750 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5751 patterns for dwarf exception handling.
5753 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5755 * config/nds32/nds32.h: Clean up obsolete macros.
5757 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5759 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5760 Add enum values for particular instructions.
5761 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5762 particular intrinsic functions.
5763 * config/nds32/nds32-intrinsic.md: Likewise.
5764 * config/nds32/nds32_intrinsic.h: Likewise.
5765 * config/nds32/nds32.h (nds32_builtins): Likewise.
5766 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5767 (btst, ave): New patterns for particular instructions.
5769 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5771 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5772 Add enum values for atomic load/store and memory sync.
5773 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5775 * config/nds32/nds32-intrinsic.md: Likewise.
5776 * config/nds32/nds32_intrinsic.h: Likewise.
5777 * config/nds32/nds32.h (nds32_builtins): Likewise.
5779 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5781 PR tree-optimization/85257
5782 * fold-const.c (native_encode_vector): If not all elts could fit
5783 and off is -1, return 0 rather than offset.
5784 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5785 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5786 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5787 adjust buffer in native_interpret_expr call.
5789 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5791 * config/nds32/constants.md (unspec_volatile_element): Add cache
5792 control enum values.
5793 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5794 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5795 * config/nds32/nds32.c (nds32_cctl_names): New.
5796 (nds32_print_operand): Handle cache control register names.
5797 * config/nds32/nds32.h (nds32_builtins): New enum values.
5798 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5800 * config/nds32/nds32.md (type): Add mmu.
5801 * config/nds32/pipelines.md (simple_insn): Add mmu.
5803 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5805 * config/nds32/nds32.md (type): Remove call.
5806 * config/nds32/pipelines.md (simple_insn): Likewise.
5808 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5810 * config/nds32/constants.md (unspec_volatile_element): Add
5811 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5812 UNSPEC_VOLATILE_FMFCFG.
5813 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5814 description for fmfcfg and fmfcsr.
5815 (bdesc_1arg): Add fmtcsr.
5816 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5817 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5818 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5819 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5820 unspec_fmfcfg): New patterns.
5821 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5822 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5823 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5824 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5825 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5826 __nds32__fmfcfg): Define.
5828 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5830 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5831 intrinsic register names.
5832 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5833 intrinsic register enum values and macros.
5835 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5837 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5838 for load/store addressing form.
5839 (nds32_print_operand_address): Likewise.
5841 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5844 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5845 based on LABEL_REF. Remove useless assertion.
5846 (pic_address_needs_scratch): Fix formatting.
5847 (sparc_legitimize_pic_address): Minor tweaks.
5848 (sparc_delegitimize_address): Adjust assertion accordingly.
5849 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5850 into symbolic_operand.
5851 (movsi_high_pic_label_ref): Likewise.
5852 (movsi_lo_sum_pic_label_ref): Likewise.
5853 (movdi_pic_label_ref): Likewise.
5854 (movdi_high_pic_label_ref): Likewise.
5855 (movdi_lo_sum_pic_label_ref): Likewise.
5857 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5859 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5860 custom LIB_SPEC setup.
5862 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5863 Kito Cheng <kito.cheng@gmail.com>
5865 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5866 * config/riscv/freebsd.h: New.
5868 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5870 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5871 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5874 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5875 Kito Cheng <kito.cheng@gmail.com>
5877 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5878 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5879 * config/nds32/nds32-protos.h (nds32_output_call,
5880 nds32_output_return): Declare.
5881 * config/nds32/nds32.md: Refine all the call and return patterns.
5883 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5886 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5887 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5889 PR rtl-optimization/84872
5890 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5891 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5894 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5896 * expr.c (copy_blkmode_to_reg): Revert 254862.
5897 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5899 2018-04-06 Richard Biener <rguenther@suse.de>
5902 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5903 after seeing a component reference with an adjacent field. Treat
5904 refs to arrays at struct end of external decls similar to
5905 refs to unconstrained commons.
5907 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5910 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5911 look through SAVE_EXPRs with non-side-effects argument. Adjust
5913 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5916 2018-04-06 Richard Biener <rguenther@suse.de>
5919 * alias.c (find_base_term): New wrapper around find_base_term
5920 unwinding CSELIB_VAL_PTR changes.
5921 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5924 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5926 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5928 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5929 constant definitions.
5930 ("nop"): lr 0,0 -> nopr r0
5931 ("nop_lr0", "nop_lr1"): New insn definitions.
5933 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5935 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5936 NDS32_V3PUSH_AVAILABLE_P macro.
5938 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5939 Chung-Ju Wu <jasonwucj@gmail.com>
5941 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5942 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5943 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5944 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5945 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5946 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5947 * config/nds32/constraints.md: New constraints and checking for hard
5948 float configuration.
5949 * config/nds32/iterators.md: New mode iterator and attribute for hard
5950 float configuration.
5951 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5953 * config/nds32/nds32-fpu.md: New file.
5954 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5955 deal with hard float code generation.
5956 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5958 (abi_type, float_reg_number): New enum type.
5959 * config/nds32/nds32-predicates.c: New predicates for hard float.
5960 * config/nds32/nds32-protos.h: Declare functions for hard float.
5961 * config/nds32/nds32.c: Implementation for hard float configuration.
5962 * config/nds32/nds32.h: Definitions for hard float configuration.
5963 * config/nds32/nds32.md: Include hard float machine description and
5964 modify patterns for hard float configuration.
5965 * config/nds32/nds32.opt: New options for hard float configuration.
5966 * config/nds32/predicates.md: New predicates for hard float
5969 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5971 * common/config/nds32/nds32-common.c
5972 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5974 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5977 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5978 CONSTRUCTOR_ELT (ctor, ...)->value.
5980 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5983 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5985 2018-04-05 Tom de Vries <tom@codesourcery.com>
5988 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5991 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5992 Kito Cheng <kito.cheng@gmail.com>
5994 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5995 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5996 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5997 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5999 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6000 Kito Cheng <kito.cheng@gmail.com>
6002 * config/nds32/constraints.md (Ufe): New memory constraint.
6003 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6004 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6005 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6007 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6008 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6010 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6012 * config/nds32/nds32.md: Use optimize_size in the condition for
6013 alu-shift instructions.
6015 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6017 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6019 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6021 * config/nds32/nds32.md (negsi2): Refine pattern.
6023 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6024 Chung-Ju Wu <jasonwucj@gmail.com>
6026 * config/nds32/iterators.md (shift_rotate): New code iterator.
6027 (shift): New code attribute.
6028 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6029 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6030 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6031 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6032 bit-wise operations.
6033 (andsi3, *andsi3): Ditto.
6034 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6035 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6036 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6037 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6038 nds32_ior_operand, nds32_xor_operand): New predicates.
6040 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6042 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6043 (addsi3, subsi3): ... this.
6045 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6047 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6049 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6051 * config/nds32/nds32.md: Adjust indention.
6053 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6055 * config/nds32/nds32.md (feature): New attribute.
6057 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6059 * config/nds32/nds32.md (subtype): New attribute.
6061 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6064 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6065 expansion to perform a bitwise AND of the argument followed by a
6066 boolean negation of the result.
6068 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6070 PR rtl-optimization/84878
6071 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6072 the basic block. Assert the use reference is not artificial and that
6073 it has an associated insn.
6075 2018-04-04 Michael Matz <matz@suse.de>
6077 * builtins.c (compute_objsize): Pass correct operand
6078 to array_at_struct_end_p.
6080 2018-04-04 Richard Biener <rguenther@suse.de>
6083 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6084 from contexts for DINFO_LEVEL_TERSE and below.
6086 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6088 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6089 resiter_operand condition.
6090 * config/nds32/nds32.md (*move<mode>): Ditto.
6092 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6093 Monk Chiang <sh.chiang04@gmail.com>
6095 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6097 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6099 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6101 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6102 Kito Cheng <kito.cheng@gmail.com>
6104 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6105 nds32_cond_code_str, output_cond_branch,
6106 output_cond_branch_compare_zero, nds32_expand_cbranch,
6107 nds32_expand_cstore, nds32_expand_movcc,
6108 nds32_output_cbranchsi4_equality_zero,
6109 nds32_output_cbranchsi4_equality_reg,
6110 nds32_output_cbranchsi4_equality_reg_or_const_int,
6111 nds32_output_cbranchsi4_greater_less_zero: New functions.
6112 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6113 nds32_expand_cstore, nds32_expand_movcc,
6114 nds32_output_cbranchsi4_equality_zero,
6115 nds32_output_cbranchsi4_equality_reg,
6116 nds32_output_cbranchsi4_equality_reg_or_const_int,
6117 nds32_output_cbranchsi4_greater_less_zero): Declare.
6118 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6119 nds32_rimm11s_operand): New predicates.
6120 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6121 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6124 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6126 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6127 * config/nds32/nds32.md: Ditto.
6128 * config/nds32/pipelines.md: Ditto.
6130 2018-04-04 Richard Biener <rguenther@suse.de>
6132 PR tree-optimization/85168
6133 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6134 propagating abnormals.
6136 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6138 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6140 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6141 Kito Cheng <kito.cheng@gmail.com>
6143 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6144 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6145 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6146 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6147 * config/nds32/nds32.md (sibcall_internal): New.
6148 (sibcall_register): Remove.
6149 (sibcall_immediate): Remove.
6150 (sibcall_value_internal): New.
6151 (sibcall_value_register): Remove.
6152 (sibcall_value_immediate): Remove.
6153 * config/nds32/predicates.md (nds32_general_register_operand): New.
6154 (nds32_call_address_operand): New.
6156 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6158 PR rtl-optimization/85167
6159 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6160 bb_defs if *split_p, instead preinitialize it to NULL.
6162 PR tree-optimization/85156
6163 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6164 evaluating the argument multiple times.
6166 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6168 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6170 (_mm_cvtpd_ps): Likewise.
6171 (_mm_cvttpd_epi32): Likewise.
6172 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6173 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6174 vector, pixel, and bool following altivec.h include.
6176 2018-04-03 Martin Sebor <msebor@redhat.com>
6178 * doc/extend.texi (Common Function Attributes): Clarify.
6179 (const attribute): Likewise.
6180 (pure attribute): Likewise.
6182 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6185 * config/i386/i386.c (ix86_expand_vector_set): Use
6186 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6188 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6190 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6191 instructions when changing rounding bits to preserve precision bits
6192 in the x87 control word.
6194 2018-04-03 Martin Liska <mliska@suse.cz>
6196 PR tree-optimization/82491
6197 * rtl.h (strip_offset_and_add): Replace += suboffset with
6198 poly_uint64 () + suboffset.
6200 2018-03-29 Martin Liska <mliska@suse.cz>
6201 Martin Jambor <mjambor@suse.cz>
6204 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6205 param_type is not an integral or pointer type.
6207 2018-04-03 Richard Biener <rguenther@suse.de>
6209 * sese.h (recompute_all_dominators): Remove.
6211 2018-04-02 Martin Sebor <msebor@redhat.com>
6213 * doc/invoke.texi (-Wrestrict): Fix typos.
6215 2018-04-02 Jim Wilson <jimw@sifive.com>
6217 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6218 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6219 (<optab>di3, <optab>si3_extend): Likewise.
6220 (<optab>si3_mask, <optab>si3_mask_1): New.
6221 (<optab>di3_mask, <optab>di3_mask_1): New.
6222 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6223 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6224 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6226 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6228 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6231 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6233 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6234 (nds32_canonicalize_comparison): New function.
6236 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6237 Kito Cheng <kito.cheng@gmail.com>
6238 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6240 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6241 * config/nds32/constants.md (unspec_volatile_element): Add
6242 UNSPEC_VOLATILE_RELAX_GROUP.
6243 * config/nds32/nds32-relax-opt.c: New file.
6244 * config/nds32/nds32-predicates.c
6245 (nds32_symbol_load_store_p): New function.
6246 * config/nds32/nds32-protos.h
6247 (nds32_symbol_load_store_p): Declare function.
6248 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6249 * config/nds32/nds32.c
6250 (nds32_register_pass): New function to register pass.
6251 (nds32_register_passes): New function to register passes.
6252 * config/nds32/nds32.md (relax_group): New pattern.
6253 * config/nds32/nds32.opt (mrelax-hint): New option.
6254 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6256 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6258 * config/nds32/t-nds32: Modify files dependency.
6260 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6262 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6263 (PROFILE_HOOK): Define its implementation.
6265 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6267 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6268 type and 32-bit size.
6270 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6273 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6274 (V_128_256): New mode iterator.
6275 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6276 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6277 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6279 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6280 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6282 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6285 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6286 NaN inputs correctly.
6288 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6291 * config/rs6000/vsx.md (??r): New mode attribute.
6292 (*vsx_mov<mode>_64bit): Use it.
6293 (*vsx_mov<mode>_32bit): Likewise.
6295 2018-03-30 Martin Sebor <msebor@redhat.com>
6297 PR tree-optimization/84818
6298 * builtins.c (check_access): Use warning_n.
6300 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6303 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6305 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6308 2018-03-30 Julia Koval <julia.koval@intel.com>
6311 * x86-tune.def (movx, partial_reg_dependency): Enable for
6314 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6317 * lra-constraints.c (process_alt_operands): Move setting
6318 this_alternative_matches below.
6320 2018-03-29 Martin Liska <mliska@suse.cz>
6323 * doc/invoke.texi: Document how LTO works with debug info.
6324 Describe auto-load support of binutils. Mention 'x86-64'
6325 as valid option value of -march option.
6327 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6329 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6332 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6333 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6334 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6337 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6340 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6341 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6342 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6343 for RS6000_BTM_POWERPC64.
6344 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6345 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6346 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6351 2018-03-28 Carl Love <cel@us.ibm.com>
6354 2017-09-27 Carl Love <cel@us.ibm.com>
6356 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6357 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6358 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6361 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6363 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6364 instead of __vector bool.
6365 (_mm_max_pu8): Likewise.
6366 (_mm_min_pi16): Likewise.
6368 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6371 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6372 (DIVWEUO): Likewise.
6374 (DIVDEUO): Likewise.
6375 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6376 DIVWEUO and DIVDEUO.
6377 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6378 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6379 (div_extend): Likewise.
6380 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6382 (__builtin_divweuo): Likewise.
6383 (__builtin_divdeo): Likewise.
6384 (__builtin_divdeuo): Likewise.
6386 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6389 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6390 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6392 PR tree-optimization/82004
6393 * gimple-match-head.c (optimize_pow_to_exp): New function.
6394 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6395 Don't fold to exp if optimize_pow_to_exp is false.
6397 2018-03-28 Martin Liska <mliska@suse.cz>
6400 * calls.c (initialize_argument_information): Fix trailing space.
6401 * common.opt: Fix typo and provide better explanation for
6402 -fsanitize-coverage option.
6403 * config/i386/i386.opt: Fix typo.
6405 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6406 Martin Liska <mliska@suse.cz>
6409 * gimplify.c (asan_poison_variable): Don't do the check for
6410 gimplify_omp_ctxp here.
6411 (gimplify_decl_expr): Do it here.
6412 (gimplify_target_expr): Likewise.
6414 2018-03-28 Martin Liska <mliska@suse.cz>
6417 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6418 chkp_type_bounds_count if MPX is not enabled.
6420 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6422 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6424 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6427 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6428 function to create the function decl for complex long double
6429 multiply and divide for -mabi=ieeelongdouble.
6430 (init_float128_ieee): Call it.
6432 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6435 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6436 -fcf-protection=branch -mibt.
6437 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6439 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6442 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6443 arm_word_relocations.
6445 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6448 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6449 extern array declarations.
6451 2018-03-27 Richard Biener <rguenther@suse.de>
6454 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6455 explicit single_use checks.
6457 2018-03-27 Richard Biener <rguenther@suse.de>
6459 PR tree-optimization/85082
6460 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6463 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6465 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6466 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6467 Turn on fasynchronous-unwind-tables and funwind-tables.
6469 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6472 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6473 (*bmi_blsr_<mode>_ccz): Ditto.
6475 2018-03-26 Tom de Vries <tom@codesourcery.com>
6477 PR tree-optimization/85063
6478 * omp-general.c (offloading_function_p): New function. Factor out
6480 * omp-offload.c (pass_omp_target_link::gate): ... here.
6481 * omp-general.h (offloading_function_p): Declare.
6482 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6483 with attribute omp declare target for offloading functions.
6485 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6487 PR tree-optimization/84005
6488 * tree-data-ref.h (get_base_for_alignment): Declare.
6489 * tree-data-ref.c (get_base_for_alignment_1): New function.
6490 (get_base_for_alignment): Likewise.
6491 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6492 get_base_for_alignment to find a suitable base object, instead
6493 of always using drb->base_address.
6495 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6498 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6499 known size by default.
6501 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6504 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6505 and non BLKmode operands.
6507 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6510 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6511 Clean up attributes.
6513 2018-03-23 Richard Biener <rguenther@suse.de>
6516 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6517 we are going to emit early debug for LTO.
6519 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6522 * function.c (match_asm_constraints_1): Don't optimize if input
6523 doesn't satisfy general_operand predicate for output's mode.
6526 * alias.c (write_dependence_p): Don't require for x_canonicalized
6527 non-VOIDmode if x has VOIDmode.
6530 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6531 just don't try to optimize it rather than assert it never happens.
6533 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6535 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6536 macro expansions for definition of ST_INTERNAL_<mode> and
6537 LD_INTERNAL_<mode> builtins.
6538 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6540 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6542 (altivec_expand_st_builtin): Likewise.
6543 (altivec_expand_builtin): Remove calls to deleted functions.
6544 (rs6000_address_for_altivec): Delete this function.
6545 * config/rs6000/vector.md: Remove expands for
6546 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6548 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6551 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6552 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6553 re-computing once computed.
6554 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6555 (arm_init_machine_status): Initialize
6556 machine->static_chain_stack_bytes.
6558 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6561 * doc/extend.texi: Add four new prototypes for vec_ld.
6562 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6563 definitions for more logical presentation.
6564 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6565 entries for V1TI variants of __builtin_altivec_ld builtin.
6566 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6567 handling of V1TI variant of LVX icode pattern.
6568 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6569 (rs6000_gimple_fold_builtin): Likewise.
6570 (altivec_init_builtins): Add code to define
6571 __builtin_altivec_lvx_v1ti function.
6573 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6576 * function.c (match_asm_constraints_1): Don't do the optimization
6577 if input isn't a REG, SUBREG, MEM or constant.
6579 2018-03-22 Tom de Vries <tom@codesourcery.com>
6581 PR tree-optimization/84956
6582 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6583 bb_has_abnormal_pred.
6585 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6588 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6589 DECL_INITIAL (decl) to decl at the end.
6590 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6593 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6595 * doc/extend.texi (__builtin_tgmath): Document when complex
6596 integer types are treated as _Complex _Float64.
6598 2018-03-21 Tom de Vries <tom@codesourcery.com>
6600 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6602 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6604 PR tree-optimization/84960
6605 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6606 if it is ENTRY block, move them into single succ of ENTRY in that case.
6608 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6610 PR tree-optimization/84811
6611 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6612 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6613 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6614 (known_subrange_p): Update accordingly. Cast each value involved
6615 in the size comparison, rather than casting the result of the
6618 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6620 PR tree-optimization/84982
6621 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6622 by flipping the least significant bit rather than all bits from
6623 bitpos to bitpos + bitsize - 1.
6625 2018-03-21 Nathan Sidwell <nathan@acm.org>
6627 * doc/extend.texi (Deprecated Features): Remove mention of
6628 long-deleted deprecations.
6630 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6633 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6634 * configure: Regenerate.
6636 2018-03-21 Tom de Vries <tom@codesourcery.com>
6638 PR tree-optimization/83126
6639 * tree-parloops.c (num_phis): New function.
6640 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6642 2018-03-21 Nathan Sidwell <nathan@acm.org>
6644 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6645 mention anon-struct/union members and trailing attributes.
6647 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6649 PR tree-optimization/84969
6650 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6651 builtin memset partitions if they set different rhs values.
6653 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6655 PR rtl-optimization/84989
6656 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6657 VEC_DUPLICATE with scalar result mode.
6659 2018-03-21 Martin Liska <mliska@suse.cz>
6662 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6663 not intended return statement.
6665 2018-03-21 Martin Liska <mliska@suse.cz>
6668 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6669 (chkp_find_bound_slots_1): Limit number of iterations.
6671 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6674 * Minor grammar fixes for x86 options.
6676 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6679 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6680 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6682 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6685 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6686 (altivec_lvx_<mode>_1op): Likewise.
6687 (altivec_stvx_<mode>_2op): Likewise.
6688 (altivec_stvx_<mode>_1op): Likewise.
6689 (altivec_lvx_<VM2:mode>): New define_expand.
6690 (altivec_stvx_<VM2:mode>): Likewise.
6691 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6692 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6693 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6694 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6695 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6696 (rs6000_gen_lvx): Likewise.
6697 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6698 (altivec_expand_stv_builtin): Likewise.
6699 (altivec_expand_builtin): Likewise.
6700 * config/rs6000/vector.md: Likewise.
6702 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6705 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6708 2018-03-20 Richard Biener <rguenther@suse.de>
6711 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6712 sign-conversions as zero, fall back to standard scalar_stmt
6715 2018-03-20 Martin Liska <mliska@suse.cz>
6718 * predict.c (rebuild_frequencies): Handle case when we have
6719 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6721 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6724 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6725 flag_section_anchors.
6726 * varasm.c (use_blocks_for_decl_p): Remove hack for
6727 dw2_force_const_mem.
6730 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6732 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6733 be created, use lowpart_subreg of operands[0] rather than operands[0]
6735 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6736 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6737 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6738 and n constraint instead of aarch64_shift_imm_di and Usd.
6739 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6740 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6742 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6745 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6746 to favor GPR over NEON registers.
6747 (<shift>di3_neon): Likewise.
6749 2018-03-20 Tom de Vries <tom@codesourcery.com>
6752 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6753 (nvptx_process_pars): Emit bar.sync asap and alap.
6755 2018-03-20 Tom de Vries <tom@codesourcery.com>
6758 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6759 seen_label if seen_label is already set.
6761 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6764 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6765 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6766 Use 1U instead of 1. Formatting fixes.
6769 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6770 instead of TREE_TYPE (s1) for the return value.
6772 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6774 PR tree-optimization/84946
6775 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6776 bitsize + bitsize in poly_uint64 rather than poly_int64.
6779 * dwarf2asm.c: Include fold-const.c.
6780 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6781 of decl rather than decl itself.
6783 PR rtl-optimization/84643
6784 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6786 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6789 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6790 calling assemble_variable.
6792 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6795 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6796 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6798 2018-03-19 Jim Wilson <jimw@sifive.com>
6801 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6802 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6803 (riscv_first_stack_step): Likewise.
6804 (riscv_option_override): Use STACK_BOUNDARY instead of
6806 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6808 (BIGGEST_ALIGNMENT): Set to 128.
6809 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6810 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6813 2018-03-19 Richard Biener <rguenther@suse.de>
6815 PR tree-optimization/84933
6816 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6817 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6819 2018-03-19 Richard Biener <rguenther@suse.de>
6821 PR tree-optimization/84859
6822 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6823 (cond_if_else_store_replacement): Perform sinking operation on
6824 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6825 Generalize what a BB with a single eligible store is.
6827 2018-03-19 Richard Biener <rguenther@suse.de>
6829 PR tree-optimization/84929
6830 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6831 chrec_is_positive against non-chrec arg.
6833 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6836 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6838 2018-03-18 Martin Liska <mliska@suse.cz>
6840 PR rtl-optimization/84635
6841 * regrename.c (build_def_use): Use matches_mode only when
6844 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6846 PR tree-optimization/84913
6847 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6848 vectorize chains of COND_EXPRs.
6850 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6852 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6854 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6856 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6858 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6860 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6862 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6863 Kito Cheng <kito.cheng@gmail.com>
6865 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6866 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6867 (nds32_adjust_reg_alloc_order): New function.
6868 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6870 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6872 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6873 nds32_print_operand, nds32_print_operand_address): Use
6874 HOST_WIDE_INT_PRINT_DEC instead.
6876 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6878 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6880 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6883 * config/i386/i386.c (initial_ix86_tune_features,
6884 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6886 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6887 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6888 rather than 1u << ix86_tune. Formatting fix.
6889 (ix86_option_override_internal): Change ix86_arch_mask from
6890 unsigned int to unsigned HOST_WIDE_INT, initialize to
6891 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6892 (ix86_function_specific_restore): Likewise.
6894 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6897 * postreload.c (reload_combine_recognize_pattern): Perform
6898 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6899 truncate_int_for_mode the result for the destination's mode.
6902 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6903 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6905 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6908 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6911 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6912 regno_allocno_class_array and sorted_pseudos.
6913 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6914 insns where regno is used.
6916 2018-03-16 Martin Liska <mliska@suse.cz>
6919 * multiple_target.c (create_dispatcher_calls): Redirect
6920 reference in the symbol table.
6922 2018-03-16 Martin Liska <mliska@suse.cz>
6925 * multiple_target.c (create_dispatcher_calls): Redirect also
6928 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6932 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6933 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6934 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6936 2018-03-16 Julia Koval <julia.koval@intel.com>
6938 * doc/invoke.texi (Skylake Server): Add CLWB.
6939 Cannonlake): Remove CLWB.
6941 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6943 PR tree-optimization/84841
6944 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6946 (FLOAT_ONE_CONST_TYPE): Define.
6947 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6948 (sort_by_operand_rank): Put entries with higher constant_type last
6949 rather than first to match comments.
6951 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6953 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6956 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6959 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6960 check and use address of target always.
6962 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6965 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6966 (indirect_thunk_bnd_needed): Likewise.
6967 (indirect_thunks_used): Likewise.
6968 (indirect_thunks_bnd_used): Likewise.
6969 (indirect_return_needed): New.
6970 (indirect_return_bnd_needed): Likewise.
6971 (output_indirect_thunk_function): Add a bool argument for
6973 (output_indirect_thunk_function): Don't generate alias for
6974 function return thunk.
6975 (ix86_code_end): Call output_indirect_thunk_function to generate
6976 function return thunks.
6977 (ix86_output_function_return): Set indirect_return_bnd_needed
6978 and indirect_return_needed instead of indirect_thunk_bnd_needed
6979 and indirect_thunk_needed.
6981 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6983 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6984 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6985 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6987 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6988 Paul Hua <paul.hua.gm@gmail.com>
6991 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6993 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6995 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6996 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6999 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7002 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7003 instead of GET_MODE_SIZE when comparing Units.
7005 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7008 * varasm.c (hash_section): Return an unchangeble hash value
7009 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7010 Return !aarch64_can_use_per_function_literal_pools_p ().
7012 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7015 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7016 rather than address of cmode as last argument to prepare_cmp_insn.
7018 2018-03-15 Julia Koval <julia.koval@intel.com>
7020 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7021 F_AVX512VNNI, F_AVX512BITALG): New.
7023 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7026 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7027 insn for floating-point loads and stores.
7029 2018-03-14 Carl Love <cel@us.ibm.com>
7031 * config/rs6000/rs6000-c.c: Add macro definitions for
7032 ALTIVEC_BUILTIN_VEC_PERMXOR.
7033 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7034 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7035 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7036 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7038 * config/doc/extend.texi: Add prototypes for vec_permxor.
7040 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7043 * diagnostic-show-locus.c (class layout_point): Convert m_line
7044 from int to linenum_type.
7045 (line_span::comparator): Use linenum "compare" function when
7046 comparing line numbers.
7047 (test_line_span): New function.
7048 (layout_range::contains_point): Convert param "row" from int to
7050 (layout_range::intersects_line_p): Likewise.
7051 (layout::will_show_line_p): Likewise.
7052 (layout::print_source_line): Likewise.
7053 (layout::should_print_annotation_line_p): Likewise.
7054 (layout::print_annotation_line): Likewise.
7055 (layout::print_leading_fixits): Likewise.
7056 (layout::annotation_line_showed_range_p): Likewise.
7057 (struct line_corrections): Likewise for field m_row.
7058 (line_corrections::line_corrections): Likewise for param "row".
7059 (layout::print_trailing_fixits): Likewise.
7060 (layout::get_state_at_point): Likewise.
7061 (layout::get_x_bound_for_row): Likewise.
7062 (layout::print_line): Likewise.
7063 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7064 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7065 * input.c (selftest::test_linenum_comparisons): New function.
7066 (selftest::input_c_tests): Call it.
7067 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7068 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7069 * selftest.h (ASSERT_GT): New macro.
7070 (ASSERT_GT_AT): New macro.
7071 (ASSERT_LT): New macro.
7072 (ASSERT_LT_AT): New macro.
7074 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7076 PR rtl-optimization/84780
7077 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7079 2018-03-14 Martin Liska <mliska@suse.cz>
7081 * tree.c (record_node_allocation_statistics): Use
7082 get_stats_node_kind.
7083 (get_stats_node_kind): New function extracted from
7084 record_node_allocation_statistics.
7085 (free_node): Use get_stats_node_kind.
7087 2018-03-14 Richard Biener <rguenther@suse.de>
7089 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7090 that the value-set of ANTIC_IN doesn't grow.
7093 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7095 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7096 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7098 2018-03-14 Julia Koval <julia.koval@intel.com>
7100 * config.gcc (icelake-client, icelake-server): New.
7102 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7103 (initial_ix86_arch_features): Ditto.
7104 (PTA_SKYLAKE): Add SGX.
7105 (PTA_ICELAKE): Remove.
7106 (PTA_ICELAKE_CLIENT): New.
7107 (PTA_ICELAKE_SERVER): New.
7108 (ix86_option_override_internal): Split up icelake on icelake client and
7110 (get_builtin_code_for_version): Ditto.
7111 (fold_builtin_cpu): Ditto.
7112 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7113 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7114 * config/i386/i386.h (processor_type): Ditto.
7115 * doc/invoke.texi: Ditto.
7117 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7120 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7121 INTEGER_CST offset, add it together with bitpos / 8 and
7122 sign extend based on POINTER_SIZE.
7126 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7129 * config/i386/constraints.md (Yc): New register constraint.
7130 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7131 Use Yc constraint for alternative 2 of operand 0. Remove
7132 preferred_for_speed attribute.
7134 2018-03-14 Richard Biener <rguenther@suse.de>
7136 PR tree-optimization/84830
7137 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7138 with the old one to avoid oscillations.
7140 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7143 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7145 (assign_by_spills): Return a flag of reload assignment failure.
7146 Do not process the reload assignment failures. Do not spill other
7147 reload pseudos if they has the same reg class. Update n if
7149 (lra_assign): Add a return arg. Set up from the result of
7150 assign_by_spills call.
7151 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7152 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7153 usage_insns if it is not NULL.
7154 (spill_hard_reg_in_range): New function.
7155 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7156 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7157 function prototypes.
7158 (lra_assign): Change prototype.
7159 * lra.c (lra): Add code to deal with fails by splitting hard reg
7162 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7164 * config/riscv/riscv.opt (mrelax): New option.
7165 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7166 "norelax" when riscv_mrelax is disabled.
7167 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7169 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7172 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7173 reassociation for int modes.
7175 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7177 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7178 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7180 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7181 * config/aarch64/aarch64-sve.md
7182 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7183 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7184 (*extend<mode><Vwide>2): Rename to...
7185 (aarch64_sve_extend<mode><Vwide>2): ...this.
7186 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7187 renaming the old pattern to...
7188 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7190 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7191 define_expand, renaming the old pattern to...
7192 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7193 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7194 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7195 account when deciding which SVE instruction the optab should use.
7196 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7198 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7200 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7201 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7202 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7203 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7204 (tlsdesc_small_<mode>): Turn a define_expand and use
7205 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7206 (tlsdesc_small_advsimd_<mode>): ...this.
7207 (tlsdesc_small_sve_<mode>): New pattern.
7209 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7211 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7212 (UNSPEC_UMUL_HIGHPART): New constants.
7213 (MUL_HIGHPART): New int iteraor.
7214 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7215 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7217 (*<su>mul<mode>3_highpart): New define_insn.
7219 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7222 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7225 2018-03-13 Martin Liska <mliska@suse.cz>
7228 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7230 (sem_item_optimizer::~sem_item_optimizer): Release it.
7231 (sem_item_optimizer::merge_classes): Register variable aliases.
7232 (sem_item_optimizer::fixup_pt_set): New function.
7233 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7234 * ipa-icf.h: Declare new variables and functions.
7236 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7239 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7240 integer_pow2p@2 and test integer_pow2p in condition.
7241 (A < 0 ? C : 0): Similarly for @1.
7244 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7245 characters starting at p contain '\0' character, don't look beyond
7249 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7250 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7253 * reg-stack.c (change_stack): Change update_end var from int to
7254 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7255 also call set_block_for_insn on the newly added insns and rescan.
7258 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7259 on the last operand.
7262 * tree.c (stabilize_reference_1): Return save_expr (e) for
7263 STATEMENT_LIST even if it doesn't have side-effects.
7265 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7267 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7269 2018-03-12 Renlin Li <renlin.li@arm.com>
7271 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7272 aarch64_output_scalar_simd_mov_immediate.
7274 2018-03-12 Martin Sebor <msebor@redhat.com>
7276 PR tree-optimization/83456
7277 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7278 for perfectly overlapping calls to memcpy.
7279 (gimple_fold_builtin_memory_chk): Same.
7280 (gimple_fold_builtin_strcpy): Handle no-warning.
7281 (gimple_fold_builtin_stxcpy_chk): Same.
7282 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7284 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7286 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7287 parameter. Use it for SFmode.
7288 (rs6000_function_arg_advance_1): Adjust.
7289 (rs6000_function_arg): Adjust.
7290 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7292 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7294 PR rtl-optimization/84169
7295 PR rtl-optimization/84780
7296 * combine.c (can_combine_p): Check for a 2-insn combination whether
7297 the destination register is used between the two insns, too.
7299 2018-03-12 Richard Biener <rguenther@suse.de>
7301 PR tree-optimization/84803
7302 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7303 for refs DR analysis didn't process.
7305 2018-03-12 Richard Biener <rguenther@suse.de>
7307 PR tree-optimization/84777
7308 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7309 force-vectorize loops ignore whether we are optimizing for size.
7311 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7313 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7314 (TARGET_MD_ASM_ADJUST): Define.
7316 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7317 Kito Cheng <kito.cheng@gmail.com>
7318 Chung-Ju Wu <jasonwucj@gmail.com>
7320 * config/nds32/nds32.c (nds32_compute_stack_frame,
7321 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7322 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7323 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7324 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7325 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7326 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7327 * config/nds32/nds32.md (prologue, epilogue): Use macro
7328 NDS32_V3PUSH_AVAILABLE_P to do checking.
7330 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7333 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7334 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7335 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7336 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7337 addition of most attributes on !orig_type_die or the attribute not
7338 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7340 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7341 Chung-Ju Wu <jasonwucj@gmail.com>
7343 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7345 * config/nds32/nds32.opt (mvh): New option.
7347 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7348 Chung-Ju Wu <jasonwucj@gmail.com>
7350 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7352 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7353 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7356 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7357 Chung-Ju Wu <jasonwucj@gmail.com>
7359 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7361 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7362 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7364 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7365 Kito Cheng <kito.cheng@gmail.com>
7366 Chung-Ju Wu <jasonwucj@gmail.com>
7368 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7369 UNSPEC_FFMISM and UNSPEC_FLMISM.
7370 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7371 for ffb, ffmism and flmism.
7372 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7373 (unspec_ffmism): Ditto.
7374 (unspec_flmism): Ditto.
7375 (nds32_expand_builtin_impl): Check if string extension is available.
7376 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7377 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7379 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7382 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7385 * lra-assigns.c (assign_by_spills): Return a flag of reload
7386 assignment failure. Do not process the reload assignment
7387 failures. Do not spill other reload pseudos if they has the same
7389 (lra_assign): Add a return arg. Set up from the result of
7390 assign_by_spills call.
7391 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7392 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7393 usage_insns if it is not NULL.
7394 (spill_hard_reg_in_range): New function.
7395 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7396 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7397 function prototypes.
7398 (lra_assign): Change prototype.
7399 * lra.c (lra): Add code to deal with fails by splitting hard reg
7402 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7405 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7407 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7410 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7411 (dw_val_node): Add val_symbolic_view.
7412 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7413 (symview_upper_bound): New.
7414 (new_line_info_table): Initialize symviews_since_reset.
7415 (dwarf2out_source_line): Count symviews_since_reset and set
7416 symview_upper_bound.
7417 (dw_val_equal_p): Handle symview.
7418 (add_AT_symview): New.
7419 (print_dw_val): Handle symview.
7420 (attr_checksum, attr_checksum_ordered): Likewise.
7421 (same_dw_val_p, size_of_die): Likewise.
7422 (value_format, output_die): Likewise.
7423 (add_high_low_attributes): Use add_AT_symview for entry_view.
7424 (dwarf2out_finish): Reset symview_upper_bound, clear
7427 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7430 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7431 Add strict argument and use it.
7432 (rs6000_split_multireg_move): Update for new strict argument.
7433 (mem_operand_gpr): Disallow all non-offsettable addresses.
7434 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7436 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7439 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7440 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7441 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7444 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7445 decl, use remap_type if we want to use the type.
7447 2018-03-09 Martin Sebor <msebor@redhat.com>
7449 PR tree-optimization/84526
7450 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7452 (builtin_access::generic_overlap): Be prepared to handle non-array
7455 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7457 PR rtl-optimization/84682
7458 * lra-constraints.c (process_address_1): Check is_address flag
7459 for address constraints.
7460 (process_alt_operands): Likewise.
7461 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7462 preprocess_constraints.
7463 * recog.h (preprocess_constraints): Add oploc parameter.
7465 * recog.c (preprocess_constraints): Test address_operand for
7466 CT_ADDRESS constraints.
7468 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7471 * lra-assigns.c (assign_by_spills): Return a flag of reload
7472 assignment failure. Do not process the reload assignment
7473 failures. Do not spill other reload pseudos if they has the same
7475 (lra_assign): Add a return arg. Set up from the result of
7476 assign_by_spills call.
7477 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7478 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7479 usage_insns if it is not NULL.
7480 (spill_hard_reg_in_range): New function.
7481 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7482 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7483 function prototypes.
7484 (lra_assign): Change prototype.
7485 * lra.c (lra): Add code to deal with fails by splitting hard reg
7488 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7491 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7492 Accept complain bool parameter. Only emit errors if it is true.
7493 (arm_parse_cpu_option_name): Likewise.
7494 (arm_target_thumb_only): Adjust callers of the above.
7495 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7496 prototype to take a default true bool parameter.
7497 (arm_parse_arch_option_name): Likewise.
7499 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7500 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7504 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7505 * configure: Regenerate.
7506 * configure.ac ("linker --version-script option"): New.
7507 ("linker soname option"): New.
7509 2018-03-09 Richard Biener <rguenther@suse.de>
7511 PR tree-optimization/84775
7512 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7513 immediate uses of predicate stmts and mark them modified.
7516 PR tree-optimization/84178
7517 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7519 (version_loop_for_if_conversion): Delay update_ssa call.
7520 (tree_if_conversion): Delay update_ssa until after predicate
7523 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7526 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7527 when the function accesses prior frames.
7529 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7532 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7533 gen_llsym, otherwise call maybe_gen_llsym.
7536 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7537 has ',' character inside of it.
7539 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7542 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7543 as clobbering CC_REGNUM.
7545 2018-03-08 Richard Biener <rguenther@suse.de>
7548 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7549 (follow_copies_to_constant): Do not follow SSA names registered
7552 2018-03-08 Richard Biener <rguenther@suse.de>
7554 PR tree-optimization/84178
7555 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7557 (version_loop_for_if_conversion): Delay update_ssa call.
7558 (tree_if_conversion): Delay update_ssa until after predicate
7561 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7563 PR tree-optimization/84178
7564 * tree-if-conv.c (release_bb_predicate): Remove the
7565 the assertion that the stmts have NULL use_ops.
7566 Discard the statements, asserting that they haven't
7567 yet been added to a BB.
7569 2018-03-08 Richard Biener <rguenther@suse.de>
7571 PR tree-optimization/84746
7572 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7573 (phi_translate): Pass in destination ANTIC_OUT set.
7574 (phi_translate_1): Likewise. For a simplified result lookup
7575 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7576 (phi_translate_set): Adjust.
7577 (do_pre_regular_insertion): Likewise.
7578 (do_pre_partial_partial_insertion): Likewise.
7580 2018-03-08 Martin Liska <mliska@suse.cz>
7582 PR gcov-profile/84735
7583 * doc/gcov.texi: Document usage of profile files.
7584 * gcov-io.h: Document changes in the format.
7586 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7590 * dwarf2out.c (struct dw_line_info_table): Update comments for
7592 (FORCE_RESET_NEXT_VIEW): New.
7593 (FORCE_RESETTING_VIEW_P): New.
7594 (RESETTING_VIEW_P): Check for -1 too.
7595 (ZERO_VIEW_P): Likewise.
7596 (new_line_info_table): Force-reset next view.
7597 (dwarf2out_begin_function): Likewise.
7598 (dwarf2out_source_line): Simplify zero_view_p initialization.
7599 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7600 view directly. Omit view when omitting .loc at line 0.
7602 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7604 PR tree-optimization/84740
7605 * tree-switch-conversion.c (process_switch): Call build_constructors
7606 only if info.phi_count is non-zero.
7608 PR tree-optimization/84739
7609 * tree-tailcall.c (find_tail_calls): Check call arguments against
7610 DECL_ARGUMENTS (current_function_decl) rather than
7611 DECL_ARGUMENTS (func) when checking for tail recursion.
7613 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7615 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7616 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7617 Volker Reichelt's entry and add entries for people that perform
7618 GCC fuzzy testing and report numerous bugs.
7620 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7623 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7624 readonly data in sdata, if that is disabled.
7625 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7626 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7627 -mreadonly-in-sdata option.
7629 2018-03-07 Martin Sebor <msebor@redhat.com>
7631 PR tree-optimization/84468
7632 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7633 basic block when looking for nul assignment.
7635 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7638 * except.h (output_function_exception_table): Adjust prototype.
7639 * except.c (output_function_exception_table): Remove FNNAME parameter
7640 and add SECTION parameter. Ouput one part of the table at a time.
7641 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7642 the first part of the exception table and emit unwind directives.
7643 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7644 (i386_pe_seh_cold_init): Likewise.
7645 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7646 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7647 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7648 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7649 * config/i386/winnt.c: Include except.h.
7650 (struct seh_frame_state): Add reg_offset, after_prologue and
7651 in_cold_section fields.
7652 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7653 (i386_pe_seh_cold_init): New function.
7654 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7655 to seh->in_cold_section.
7656 (seh_emit_push): Record the offset of the push.
7657 (seh_emit_save): Record the offet of the save.
7658 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7659 Test seh->after_prologue to disregard the epilogue.
7660 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7661 (i386_pe_end_cold_function): New function.
7663 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7666 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7667 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7670 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7672 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7673 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7676 * multiple_target.c: Include tree-inline.h and intl.h.
7677 (expand_target_clones): Diagnose and fail if node->definition and
7678 !tree_versionable_function_p (node->decl).
7680 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7682 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7684 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7685 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7686 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7688 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7691 * combine.c (try_combine): Use reg_or_subregno instead of handling
7692 just paradoxical SUBREGs and REGs.
7694 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7696 * config/arc/arc.c (arc_finalize_pic): Remove function.
7697 (arc_must_save_register): We use single base PIC register, remove
7698 checks to save/restore the PIC register.
7699 (arc_expand_prologue): Likewise.
7700 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7702 (arc_verify_short): Likewise.
7703 (arc_attr_type): Likewise.
7704 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7705 (walk_stores): Likewise.
7706 (arc_address_cost): Make it static.
7707 (arc_verify_short): Likewise.
7708 (branch_dest): Likewise.
7709 (arc_attr_type): Likewise.
7710 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7711 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7712 (arc_final_prescan_insn): Remove inserting the nops due to
7713 hardware hazards. It is done in reorg step.
7714 (insn_length_variant_t): Remove.
7715 (insn_length_parameters_t): Likewise.
7716 (arc_insn_length_parameters): Likewise.
7717 (arc_get_insn_variants): Likewise.
7718 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7720 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7723 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7726 PR tree-optimization/84687
7727 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7729 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7731 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7733 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7734 Rename to ppc_speculation_barrier.
7735 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7736 __builtin_ppc_speculation_barrier.
7738 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7741 * combine.c (combine_simplify_rtx): Don't try to simplify if
7742 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7745 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7747 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7748 to 32 bytes when compiling for POWER9.
7750 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7753 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7754 regparm >= 3 with no arg reg available also for calls with
7755 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7758 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7760 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7762 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7765 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7767 2018-03-05 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/84486
7770 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7771 When inserting a __builtin_assume_aligned call set the LHS
7772 SSA name alignment info accordingly.
7774 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7776 PR tree-optimization/84114
7777 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7778 Avoid reassociation of FLOAT_MODE addition.
7780 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7782 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7783 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7784 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7785 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7786 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7787 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7788 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7790 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7791 __builtin_ia32_wbinvd): New builtins.
7792 (SPECIAL_ARGS2): New.
7793 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7794 (SPECIAL_ARGS2): New.
7795 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7796 (ix86_valid_target_attribute_inner_p): Ditto.
7797 (ix86_init_mmx_sse_builtins): Add special_args2.
7798 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7799 TARGET_WBNOINVD_P): New.
7800 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7801 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7802 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7803 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7804 * config/i386/pconfigintrin.h: New file.
7805 * config/i386/wbnoinvdintrin.h: Ditto.
7806 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7808 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7810 2018-03-05 Richard Biener <rguenther@suse.de>
7812 PR tree-optimization/84670
7813 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7815 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7816 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7817 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7818 doesn't grow if all successors (recursively) were visited at least
7821 2018-03-05 Richard Biener <rguenther@suse.de>
7823 PR tree-optimization/84650
7824 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7825 if executed in the loop pipeline.
7827 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7829 * doc/configfiles.texi (Configuration Files): Move info about
7830 conditionalizing $target-protos.h to...
7831 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7832 differs from $target-protos.h.
7834 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7835 Chung-Ju Wu <jasonwucj@gmail.com>
7837 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7838 * config/nds32/nds32-multiple.md (setmemsi): Define.
7839 * config/nds32/nds32-memory-manipulation.c
7840 (nds32_gen_dup_4_byte_to_word_value): New.
7841 (emit_setmem_word_loop): New.
7842 (emit_setmem_byte_loop): New.
7843 (nds32_expand_setmem_loop): New.
7844 (nds32_expand_setmem_loop_v3m): New.
7845 (nds32_expand_setmem_unroll): New.
7846 (nds32_expand_setmem): New.
7848 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7849 Chung-Ju Wu <jasonwucj@gmail.com>
7851 * config/nds32/nds32-memory-manipulation.c
7852 (nds32_emit_load_store): New.
7853 (nds32_emit_post_inc_load_store): New.
7854 (nds32_emit_mem_move): New.
7855 (nds32_emit_mem_move_block): New.
7856 (nds32_expand_movmemsi_loop_unknown_size): New.
7857 (nds32_expand_movmemsi_loop_known_size): New.
7858 (nds32_expand_movmemsi_loop): New.
7859 (nds32_expand_movmemsi_unroll): New.
7860 (nds32_expand_movmemqi): Rename ...
7861 (nds32_expand_movmemsi): ... to this.
7862 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7863 (movmemsi): ... to this.
7864 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7865 (nds32_expand_movmemsi): ... to this.
7867 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7868 Monk Chiang <sh.chiang04@gmail.com>
7869 Chung-Ju Wu <jasonwucj@gmail.com>
7871 * config/nds32/nds32-protos.h
7872 (nds32_expand_load_multiple): New arguments.
7873 (nds32_expand_store_multiple): Ditto.
7874 (nds32_valid_multiple_load_store): Rename ...
7875 (nds32_valid_multiple_load_store_p): ... to this.
7876 * config/nds32/nds32-memory-manipulation.c
7877 (nds32_expand_load_multiple): Refine implementation.
7878 (nds32_expand_store_multiple): Ditto.
7879 * config/nds32/nds32-multiple.md
7880 (load_multiple): Update nds32_expand_load_multiple interface.
7881 (store_multiple): Update nds32_expand_store_multiple interface.
7882 * config/nds32/nds32-predicates.c
7883 (nds32_valid_multiple_load_store): Rename ...
7884 (nds32_valid_multiple_load_store_p): ... to this and refine
7886 * config/nds32/predicates.md
7887 (nds32_load_multiple_and_update_address_operation): New predicate.
7888 (nds32_store_multiple_and_update_address_operation): New predicate.
7890 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7891 Chung-Ju Wu <jasonwucj@gmail.com>
7893 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7894 (combo): New attribute.
7895 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7897 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7899 * config/nds32/nds32.opt: Change -mcmodel= default value.
7901 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7902 Monk Chiang <sh.chiang04@gmail.com>
7903 Chung-Ju Wu <jasonwucj@gmail.com>
7905 * config/nds32/constants.md (unspec_element): New enum.
7906 * config/nds32/constraints.md (Umw): New constraint.
7907 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7908 * config/nds32/nds32-intrinsic.md: Likewise.
7909 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7910 (nds32_valid_smw_lwm_base_p): New.
7911 (nds32_output_smw_single_word): New.
7912 (nds32_output_lmw_single_word): New.
7913 (nds32_expand_unaligned_load): New.
7914 (nds32_expand_unaligned_store): New.
7915 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7916 (nds32_output_smw_single_word): Declare.
7917 (nds32_output_lmw_single_word): Declare.
7918 (nds32_expand_unaligned_load): Declare.
7919 (nds32_expand_unaligned_store): Declare.
7920 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7921 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7922 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7923 NDS32_BUILTIN_UASTORE_DW.
7924 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7927 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7928 Kito Cheng <kito.cheng@gmail.com>
7929 Chung-Ju Wu <jasonwucj@gmail.com>
7931 * config/nds32/nds32-intrinsic.c
7932 (nds32_expand_builtin_null_ftype_reg): Delete.
7933 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7934 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7935 (nds32_read_argument): New.
7936 (nds32_legitimize_target): Ditto.
7937 (nds32_legitimize_argument): Ditto.
7938 (nds32_check_constant_argument): Ditto.
7939 (nds32_expand_unop_builtin): Ditto.
7940 (nds32_expand_unopimm_builtin): Ditto.
7941 (nds32_expand_binop_builtin): Ditto.
7942 (nds32_builtin_decl_impl): Ditto.
7943 (builtin_description): Ditto.
7944 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7945 (nds32_init_builtins_impl): Ditto.
7946 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7947 (nds32_builtin_decl): New.
7948 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7949 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7951 2018-03-02 Jeff Law <law@redhat.com>
7953 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7954 (redundant_insn, fill_simple_delay_slots): Likewise.
7955 (fill_slots_from_thread): Likewise.
7956 * resource.c (mark_referenced_resources): Likewise.
7957 (mark_set_resources, find_dead_or_set_registers): Likewise.
7959 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7961 * substring-locations.h (format_warning_va): Formatting fix for
7963 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7965 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7966 * substring-locations.c: Include intl.h.
7967 (format_warning_va): Turned into small wrapper around
7968 format_warning_n_va, renamed to ...
7969 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7970 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7972 (format_warning_at_substring_n): New function.
7973 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7974 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7975 format_warning_at_substring with just a shorter name instead of
7976 const function pointer.
7977 (fmtwarn_n): New function.
7978 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7979 appropriate, get rid of all the fmtstr temporaries, move conditionals
7980 with G_() wrapped string literals directly into fmtwarn arguments,
7981 cast dir.len to (int), formatting fixes.
7983 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7985 * doc/invoke.texi: Remove "Cilk Plus" references.
7987 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7988 Richard Biener <rguenther@suse.de>
7991 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7992 for error or warning attributes if CALL_FROM_THUNK_P is set.
7995 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7998 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7999 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8002 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8003 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8005 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8006 instead of -1U in last predictors element's probability member.
8008 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8011 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8012 arguments if they are comparable.
8014 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8016 PR tree-optimization/84634
8017 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8018 masks and masked_loop_p with a single loop_masks, making sure it's
8019 null for bb vectorization.
8021 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8023 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8024 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8025 loop->force_vectorize to check whether there is no alias.
8027 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8030 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8032 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8034 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8035 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8036 instead of a loop around prev_real_insn.
8037 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8041 * config/i386/i386.c (ix86_print_operand): Use conditional
8042 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8045 2018-03-02 Richard Biener <rguenther@suse.de>
8047 PR tree-optimization/84427
8048 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8049 (bitmap_set_subtract_values): Rewrite to handle multiple
8052 (prune_clobbered_mems): Likewise.
8053 (phi_translate): Take edge instead of pred/phiblock.
8054 (phi_translate_1): Likewise.
8055 (phi_translate_set): Likewise. Insert all translated
8056 exprs for a value into the set, keeping possibly multiple
8057 expressions per value.
8058 (compute_antic_aux): Adjust for phi_translate changes.
8059 When intersecting union the expressions and prune those
8060 not in the final value set, keeping possibly multiple
8061 expressions per value. Do not use value-insertion
8062 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8063 all expressions. Add verification that the value-sets
8064 only shrink during iteration.
8065 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8066 (do_pre_regular_insertion): Likewise.
8067 (do_pre_partial_partial_insertion): Likewise.
8069 2018-03-02 Richard Biener <rguenther@suse.de>
8072 * config/darwin.c (saved_debug_info_level): New static global.
8073 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8074 (darwin_asm_lto_end): Restore debug info generation settings.
8076 2018-03-01 Martin Liska <mliska@suse.cz>
8079 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8082 2018-03-01 Richard Biener <rguenther@suse.de>
8085 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8086 type annotation with LTO when debug was disabled at compile-time.
8088 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8090 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8092 (mips_final_postscan_insn): Likewise.
8094 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8096 PR rtl-optimization/84528
8097 * alias.c (init_alias_target): Add commentary.
8098 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8099 a unique base value if the frame pointer is not eliminated
8100 to the stack pointer.
8102 2018-03-01 Tom de Vries <tom@codesourcery.com>
8104 PR rtl-optimization/83327
8105 * lra-int.h (hard_regs_spilled_into): Declare.
8106 * lra.c (hard_regs_spilled_into): Define.
8107 (init_reg_info): Init hard_regs_spilled_into.
8108 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8109 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8110 (process_bb_lives): Handle hard_regs_spilled_into.
8111 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8112 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8114 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8116 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8117 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8118 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8119 * config/rs6000/aix72.h: New file.
8121 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8123 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8124 instead of warning_at with conditional singular and plural messages
8128 * stor-layout.c (update_alignment_for_field): For
8129 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8130 && !DECL_PACKED (field), do the alignment update, just use
8131 only desired_align instead of MAX (type_align, desired_align)
8133 (place_field): Don't do known_align < desired_align handling
8134 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8135 is non-NULL, instead do it after rli->prev_field handling and
8136 only if not within a bitfield word. For DECL_PACKED (field)
8137 use type_align of BITS_PER_UNIT.
8139 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8141 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8142 superfluous parentheses and trailing spaces.
8144 2018-02-28 Richard Biener <rguenther@suse.de>
8146 PR tree-optimization/84584
8147 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8148 SCoPs with fake exit edge.
8150 2018-02-28 Martin Liska <mliska@suse.cz>
8153 * timevar.c (timer::print): Fix format to properly print 100%
8156 2018-02-28 Richard Biener <rguenther@suse.de>
8159 * genmatch.c (capture_info::walk_match): Do not mark
8160 captured expressions without operands as expr_p given
8161 they act more like predicates and should be subject to
8162 "lost tail" side-effect preserving.
8164 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8166 PR rtl-optimization/81611
8167 * auto-inc-dec.c (attempt_change): Move dead note from
8168 mem_insn if it's the next use of regno
8169 (find_address): Take address use of reg holding
8170 non-incremented value. Add parm to limit search to the named
8172 (merge_in_block): Attempt to use a mem insn that is the next
8173 use of the original regno.
8175 2018-02-27 Martin Sebor <msebor@redhat.com>
8178 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8179 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8181 2018-02-27 Martin Sebor <msebor@redhat.com>
8183 PR translation/84207
8184 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8185 n argument to unsigned HOST_WIDE_INT.
8186 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8187 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8188 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8189 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8191 2018-02-27 Richard Biener <rguenther@suse.de>
8193 PR tree-optimization/84512
8194 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8195 Do not use the estimate returned from record_stmt_cost for
8196 the scalar iteration cost but sum properly using add_stmt_cost.
8198 2018-02-27 Richard Biener <rguenther@suse.de>
8200 PR tree-optimization/84466
8201 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8202 Adjust last change to less strictly validate use operands.
8204 2018-02-27 Martin Liska <mliska@suse.cz>
8206 PR gcov-profile/84548
8207 * gcov.c (process_file): Allow partial overlap and consider it
8208 also as group functions.
8209 (output_lines): Properly calculate range of lines for a group.
8211 2018-02-27 Martin Liska <mliska@suse.cz>
8213 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8214 'ggc' suffixes. Change first column width.
8215 (timer::print): Fix formatting of the column.
8217 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8219 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8220 preserve inline entry blocks for the sake of debug inline
8221 entry point markers alone.
8222 (remove_unused_locals): Suggest in comments a better place to
8223 force the preservation of inline entry blocks that are
8224 otherwise unused, but do not preserve them.
8226 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8228 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8230 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8233 * config/i386/constraints.md (Bs): Replace
8234 ix86_indirect_branch_register with
8235 TARGET_INDIRECT_BRANCH_REGISTER.
8237 * config/i386/i386.md (indirect_jump): Likewise.
8238 (tablejump): Likewise.
8239 (*sibcall_memory): Likewise.
8240 (*sibcall_value_memory): Likewise.
8241 Peepholes of indirect call and jump via memory: Likewise.
8242 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8243 (*sibcall_value_GOT_32): Likewise.
8244 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8245 (GOT_memory_operand): Likewise.
8246 (call_insn_operand): Likewise.
8247 (sibcall_insn_operand): Likewise.
8248 (GOT32_symbol_operand): Likewise.
8249 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8251 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8253 PR rtl-optimization/83496
8254 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8255 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8256 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8257 redundant insn, if any.
8258 (relax_delay_slots): Likewise.
8259 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8261 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8263 PR tree-optimization/83965
8264 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8265 that grouped statements are part of a reduction chain. Return
8266 true if the statement is not marked as a reduction itself but
8268 (vect_recog_dot_prod_pattern): Don't check whether the statement
8269 is part of a group here.
8270 (vect_recog_sad_pattern): Likewise.
8271 (vect_recog_widen_sum_pattern): Likewise.
8273 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8276 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8278 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8281 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8283 (ix86_output_indirect_function_return): New prototype.
8284 (ix86_split_simple_return_pop_internal): Likewise.
8285 * config/i386/i386.c (indirect_return_via_cx): New.
8286 (indirect_return_via_cx_bnd): Likewise.
8287 (indirect_thunk_name): Handle return va CX_REG.
8288 (output_indirect_thunk_function): Create alias for
8289 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8290 (ix86_output_indirect_jmp): Remove the bool argument.
8291 (ix86_output_indirect_function_return): New function.
8292 (ix86_split_simple_return_pop_internal): Likewise.
8293 * config/i386/i386.md (*indirect_jump): Don't pass false
8294 to ix86_output_indirect_jmp.
8295 (*tablejump_1): Likewise.
8296 (simple_return_pop_internal): Change it to define_insn_and_split.
8297 Call ix86_split_simple_return_pop_internal to split it for
8299 (simple_return_indirect_internal): Call
8300 ix86_output_indirect_function_return instead of
8301 ix86_output_indirect_jmp.
8303 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8306 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8307 memset and value initialization afterwards.
8309 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8311 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8313 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8316 * common/config/aarch64/aarch64-common.c
8317 (aarch_option_optimization_table[]): Switch
8318 off fomit-frame-pointer
8320 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8321 Chung-Ju Wu <jasonwucj@gmail.com>
8323 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8325 (store_multiple): Ditto.
8327 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8329 * config.gcc: Add --with-cpu support for nds32 target.
8330 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8331 * config/nds32/nds32.opt: Add -mcpu= option.
8333 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8335 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8336 isel=yes): Warn for these deprecated options.
8338 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8340 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8341 ISA_2_5_MASKS_EMBEDDED.
8343 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8345 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8346 p->max as pointers rather than using iterative_hash_expr.
8348 2018-02-23 Carl Love <cel@us.ibm.com>
8350 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8351 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8353 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8354 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8355 P8V_BUILTIN_VEC_VUNSIGNED2.
8357 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8360 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8361 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8363 * lra-constraints.c (curr_insn_transform): Set up
8364 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8366 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8367 * lra-eliminations.c (spill_pseudos): Ditto.
8368 (process_insn_for_elimination): Ditto.
8369 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8370 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8371 LRA_NON_CLOBBERED_ALT.
8373 2018-02-22 Martin Sebor <msebor@redhat.com>
8375 PR tree-optimization/84480
8376 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8377 to maybe_diag_stxncpy_trunc. Call it.
8378 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8379 from gimple_fold_builtin_strcpy. Print inlining stack.
8380 (handle_builtin_stxncpy): Print inlining stack.
8381 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8383 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8386 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8387 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8388 and -fcheck-pointer-bounds are used together.
8389 (indirect_thunk_prefix): New enum.
8390 (indirect_thunk_need_prefix): New function.
8391 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8392 "_nt" instead of "_bnd" for NOTRACK prefix.
8393 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8394 (output_indirect_thunk_function): Likewise.
8396 (ix86_code_end): Update output_indirect_thunk_function calls.
8397 (ix86_output_indirect_branch_via_reg): Replace
8398 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8399 (ix86_output_indirect_branch_via_push): Likewise.
8400 (ix86_output_function_return): Likewise.
8401 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8402 incompatible with -fcf-protection=branch and
8403 -fcheck-pointer-bounds.
8405 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8408 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8409 Change gcc_assert call to output_operand_lossage.
8411 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8413 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8415 2018-02-22 DJ Delorie <dj@redhat.com>
8416 Sebastian Perta <sebastian.perta@renesas.com>
8417 Oleg Endo <olegendo@gcc.gnu.org>
8419 * config/rx/rx.c (rx_rtx_costs): New function.
8420 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8422 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8424 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8426 2018-02-22 Martin Liska <mliska@suse.cz>
8429 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8430 Add "native" as a possible value.
8432 2018-02-22 Martin Liska <mliska@suse.cz>
8435 * config/i386/i386.c (ix86_option_override_internal):
8436 Add "native" as a possible value for -march and -mtune.
8438 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8441 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8442 to all type variants.
8444 PR tree-optimization/84503
8445 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8446 width as info->bitpos + info->bitsize - start.
8447 (merged_store_group::merge_overlapping): Simplify width computation.
8448 (check_no_overlap): New function.
8449 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8450 start + width and last_order of the group, fail if check_no_overlap
8452 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8453 to group if check_no_overlap fails.
8455 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8457 * config/rs6000/altivec.md: Delete contraint arguments to
8458 define_expand, define_split, and define_peephole2, and in
8459 define_insn_and_split if always unused.
8460 * config/rs6000/darwin.md: Ditto.
8461 * config/rs6000/dfp.md: Ditto.
8462 * config/rs6000/rs6000.md: Ditto.
8463 * config/rs6000/sync.md: Ditto.
8464 * config/rs6000/vector.md: Ditto.
8465 * config/rs6000/vsx.md: Ditto.
8467 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8469 * config/rs6000/altivec.md: Write output control strings as braced
8470 blocks instead of double-quoted strings.
8471 * config/rs6000/darwin.md: Ditto.
8472 * config/rs6000/rs6000.md: Ditto.
8473 * config/rs6000/vector.md: Ditto.
8474 * config/rs6000/vsx.md: Ditto.
8476 2018-02-21 Jason Merrill <jason@redhat.com>
8478 PR c++/84314 - ICE with templates and fastcall attribute.
8479 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8481 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8483 * ipa-cp.c (determine_versionability): Fix comment typos.
8485 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8488 * ipa-cp.c (determine_versionability): Do not version functions caling
8491 2018-02-21 Martin Liska <mliska@suse.cz>
8494 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8495 Add "native" as a possible value.
8496 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8497 the macro when native cpu detection is available.
8499 2018-02-21 Martin Liska <mliska@suse.cz>
8502 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8503 Add "native" as a possible value.
8504 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8505 when native cpu detection is available.
8507 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8508 Martin Sebor <msebor@redhat.com>
8510 PR tree-optimization/84478
8511 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8513 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8514 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8515 support which is conservatively correct, for 2 only stay conservative
8516 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8517 argument to the 2 argument get_range_strlen, adjust 6 arg
8518 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8520 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8521 (gimple_fold_builtin_strlen): Pass true as last argument to
8524 2018-02-20 Martin Sebor <msebor@redhat.com>
8527 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8528 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8529 (builtin_memref::builtin_memref): Factor out parts into
8530 set_base_and_offset and call it.
8532 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8535 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8536 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8537 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8538 search at the associated MODE_INT.
8540 2018-02-20 Jeff Law <law@redhat.com>
8543 PR tree-optimization/81592
8545 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8546 for range data rather than using global data.
8547 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8548 range data rather than using global data.
8549 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8550 pass it to children as needed.
8551 (struct directive::fmtresult): Similarly.
8552 (struct directive::set_width): Similarly.
8553 (struct directive::set_precision): Similarly.
8554 (format_integer, format_directive, parse_directive): Similarly.
8555 (format_none): Accept unnamed vr_values parameter.
8556 (format_percent, format_floating, format_character): Similarly.
8557 (format_string, format_plain): Similarly.
8558 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8559 the EVRP range analyzer for range data rather than using global data.
8560 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8561 gimple-ssa-evrp-analyze.h
8562 (class sprintf_dom_walker): Add after_dom_children member function.
8563 Add evrp_range_analyzer member.
8564 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8565 range analyzer as needed.
8566 (sprintf_dom_walker::after_dom_children): New member function.
8567 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8569 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8570 (evrp_range_analyzer::pop_to_marker): Likewise.
8572 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8574 PR tree-optimization/84419
8575 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8576 with the required type if its current type is compatible but
8579 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8582 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8583 after vectorization.
8585 2018-02-20 Martin Liska <mliska@suse.cz>
8588 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8589 possible values if we don't have a hint.
8591 2018-02-20 Martin Liska <mliska@suse.cz>
8595 * final.c (shorten_branches): Build align_tab array with one
8597 * opts.c (finish_options): Add alignment option limit check.
8598 (MAX_CODE_ALIGN): Likewise.
8599 (MAX_CODE_ALIGN_VALUE): Likewise.
8600 * doc/invoke.texi: Document maximum allowed option value for
8601 all -falign-* options.
8603 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8606 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8607 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8608 * var-tracking.c (emit_note_insn_var_location): Remove all references
8609 to NOTE_INSN_CALL_ARG_LOCATION.
8610 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8611 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8612 Use copy_rtx_if_shared.
8613 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8614 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8615 (dwarf2out_var_location): Remove handling of
8616 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8618 * final.c (final_scan_insn): Remove all references to
8619 NOTE_INSN_CALL_ARG_LOCATION.
8620 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8621 before dumping final insns.
8622 * except.c (emit_note_eh_region_end): Remove all references to
8623 NOTE_INSN_CALL_ARG_LOCATION.
8624 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8625 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8626 * config/arc/arc.c (hwloop_optimize): Likewise.
8627 * config/arm/arm.c (create_fix_barrier): Likewise.
8628 * config/s390/s390.c (s390_chunkify_start): Likewise.
8629 * config/sh/sh.c (find_barrier): Likewise.
8630 * config/i386/i386.c (rest_of_insert_endbranch,
8631 ix86_seh_fixup_eh_fallthru): Likewise.
8632 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8633 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8634 * config/frv/frv.c (frv_function_prologue): Likewise.
8635 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8637 (note_outside_basic_block_p): Remove all references to
8638 NOTE_INSN_CALL_ARG_LOCATION.
8639 * gengtype.c (adjust_field_rtx_def): Likewise.
8640 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8642 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8643 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8646 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8649 PR tree-optimization/84452
8650 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8651 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8654 2018-02-19 Martin Liska <mliska@suse.cz>
8657 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8659 2018-02-19 Martin Liska <mliska@suse.cz>
8660 Richard Sandiford <richard.sandiford@linaro.org>
8662 PR tree-optimization/82491
8663 * gimple-fold.c (get_base_constructor): Make earlier bail out
8666 2018-02-19 Carl Love <cel@us.ibm.com>
8668 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8669 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8671 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8672 P8V_BUILTIN_VEC_NEG.
8674 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8676 * config/rl78/rl78.md (movdf): New define expand.
8678 2018-02-19 Martin Liska <mliska@suse.cz>
8681 * doc/invoke.texi: Fix typo.
8682 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8684 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8686 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8687 handle rs6000_single_float and rs6000_double_float specially for
8690 2018-02-16 Jeff Law <law@redhat.com>
8692 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8695 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8698 * ipa-inline.c (inline_small_functions): Fix a typo.
8700 2018-02-16 Nathan Sidwell <nathan@acm.org>
8702 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8704 2018-02-16 Carl Love <cel@us.ibm.com>
8706 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8707 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8708 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8709 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8710 expansion to P8V_BUILTIN_VEC_FLOAT2.
8712 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8714 PR rtl-optimization/70023
8715 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8716 src_regno into account.
8718 2018-02-16 Carl Love <cel@us.ibm.com>
8720 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8721 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8722 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8723 * config/rs6000/rs6000.c: Remove case statements for
8724 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8725 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8726 and P9V_BUILTIN_VEC_VINSERT4B.
8727 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8728 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8729 * config/rs6000/vsx.md:
8730 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8733 2018-02-16 Carl Love <cel@us.ibm.com>
8735 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8737 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8739 * config/rs6000/rs6000-c.c: Add the definitions for
8740 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8741 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8742 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8743 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8744 definition for insert4b and define insn *insert3b_internal.
8745 * doc/extend.texi: Add documentation for vec_extract4b.
8747 2018-02-16 Nathan Sidwell <nathan@acm.org>
8749 * doc/extend.texi (Backwards Compatibility): Mention friend
8750 injection. Note for-scope is deprecated.
8751 * doc/invoke.texi (-ffriend-injection): Deprecate.
8753 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8755 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8756 that moved to I2, also allow destinations that are a paradoxical
8757 subreg (instead of a normal reg).
8759 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8762 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8765 2018-02-16 Richard Biener <rguenther@suse.de>
8767 PR tree-optimization/84037
8768 PR tree-optimization/84016
8770 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8771 Adjust vec_construct for the fact we need additional higher latency
8772 128bit inserts for AVX256 and AVX512 vector builds.
8773 (ix86_add_stmt_cost): Scale vector construction cost for
8776 2018-02-16 Richard Biener <rguenther@suse.de>
8778 PR tree-optimization/84417
8779 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8780 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8781 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8783 2018-02-16 Martin Liska <mliska@suse.cz>
8786 * internal-fn.def (ASAN_CHECK): Set proper flags.
8787 (ASAN_MARK): Likewise.
8789 2018-02-16 Julia Koval <julia.koval@intel.com>
8791 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8792 from PTA_CANNONLAKE.
8794 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8797 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8798 Use ++iter rather than iter++ for std::list iterators.
8799 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8800 defer deleting them until all nodes in the forest are processed. Do
8801 free even leaf nodes. Change to_process into auto_vec.
8804 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8805 * vec.h (vec_default_construct): Use memset instead of placement new
8806 if BROKEN_VALUE_INITIALIZATION is defined.
8807 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8808 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8811 PR rtl-optimization/83723
8812 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8813 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8814 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8816 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8818 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8820 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8822 PR rtl-optimization/81443
8823 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8824 from inner REGs to paradoxical SUBREGs.
8826 2018-02-16 Richard Biener <rguenther@suse.de>
8828 PR tree-optimization/84399
8829 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8830 For operands we can analyze at their definition make sure we can
8831 analyze them at each use as well.
8833 2018-02-16 Richard Biener <rguenther@suse.de>
8835 PR tree-optimization/84190
8836 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8837 volatile accesses if the decl isn't volatile.
8839 2018-02-15 Jason Merrill <jason@redhat.com>
8841 PR c++/84314 - ICE with templates and fastcall attribute.
8842 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8843 TYPE_CANONICAL on an existing type.
8845 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8847 PR tree-optimization/84383
8848 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8849 dstoff nor call operand_equal_p if dstbase is NULL.
8851 PR tree-optimization/84334
8852 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8853 also a CONSTANT_CLASS_P, punt.
8855 2018-02-14 Jim Wilson <jimw@sifive.com>
8857 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8858 first SMALL_OPERAND check. New local min_second_step. Move assert
8859 to where locals are set. Add TARGET_RVC support.
8860 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8862 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8864 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8866 2018-02-14 Martin Sebor <msebor@redhat.com>
8868 PR tree-optimization/83698
8869 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8870 arrays constrain the offset range to their bounds.
8871 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8872 (builtin_access::overlap): Avoid setting the size of overlap if it's
8874 (maybe_diag_overlap): Also consider arrays when deciding what values
8875 of offsets to include in diagnostics.
8877 2018-02-14 Martin Sebor <msebor@redhat.com>
8880 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8881 that correspond to the kind of a declaration.
8883 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8886 * config/pa/pa.md: Load address of PIC label using the linkage table
8887 if the label is nonlocal.
8889 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8891 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8892 warning message if user requests -maltivec=be.
8893 * doc/invoke.texi: Document deprecation of -maltivec=be.
8895 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8898 * config/rs6000/rs6000-c.c: Update definitions for
8899 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8900 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8902 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8905 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8906 add _get_ssp intrinsics. Remove argument from
8907 __builtin_ia32_rdssp[d|q].
8908 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8909 * config/i386/i386-builtin.def: Remove argument from
8910 __builtin_ia32_rdssp[d|q].
8911 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8912 ix86_expand_special_args_builtin for _rdssp[d|q].
8913 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8914 Clear register before usage.
8915 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8916 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8918 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8920 PR tree-optimization/84357
8921 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8922 operand 1 of an ARRAY_REF too.
8924 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8927 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8928 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8930 (set_of_reg): New struct.
8931 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8932 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8933 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8935 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8936 Split into bitclr, bitset, bitinvert patterns if appropriate.
8937 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8938 use rx_fuse_in_memory_bitop.
8939 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8940 to named insn, correct maximum insn length.
8942 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8945 * machmode.def: Define a complex mode for PARTIAL_INT.
8946 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8948 * doc/rtl.texi: Document CSPImode.
8949 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8951 (msp430_hard_regno_nregs_with_padding): Likewise.
8953 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8956 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8958 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8960 PR rtl-optimization/84169
8961 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8962 we generated a parallel as new i3 and we split that to new i2 and i3
8963 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8964 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8965 those to i2, not i1. Partially rewrite this scan code.
8967 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8970 * stor-layout.c (place_field): For variable length fields, adjust
8971 offset_align afterwards not just based on the field's alignment,
8972 but also on the size.
8975 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8976 of exps and logs in the use_exp2 case.
8978 2018-02-13 Jeff Law <law@redhat.com>
8980 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8983 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8986 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8990 * common.opt (gas-loc-support, gas-locview-support): New.
8991 (ginline-points, ginternal-reset-location-views): New.
8992 * doc/invoke.texi: Document them. Use @itemx where intended.
8993 (gvariable-location-views): Adjust.
8994 * target.def (reset_location_view): New.
8995 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8996 (TARGET_RESET_LOCATION_VIEW): New.
8997 * doc/tm.texi: Rebuilt.
8998 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8999 (dwarf2out_default_as_locview_support): New.
9000 (output_asm_line_debug_info): Use option variables.
9001 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9002 (output_loc_list): Likewise.
9003 (add_high_low_attributes): Check option variables.
9004 Don't output entry view attribute in strict mode.
9005 (gen_inlined_subroutine_die): Check option variables.
9006 (dwarf2out_inline_entry): Likewise.
9007 (init_sections_and_labels): Likewise.
9008 (dwarf2out_early_finish): Likewise.
9009 (maybe_reset_location_view): New, from...
9010 (dwarf2out_var_location): ... here. Call it.
9011 * debug.h (dwarf2out_default_as_loc_support): Declare.
9012 (dwarf2out_default_as_locview_support): Declare.
9013 * hooks.c (hook_int_rtx_insn_0): New.
9014 * hooks.h (hook_int_rtx_insn_0): Declare.
9015 * toplev.c (process_options): Take -gas-loc-support and
9016 -gas-locview-support from dwarf2out. Enable
9017 -gvariable-location-views by default only with locview
9018 assembler support. Enable -ginternal-reset-location-views by
9019 default only if the target defines the corresponding hook.
9020 Enable -ginline-points by default if location views are
9021 enabled; force it disabled if statement frontiers are
9023 * tree-inline.c (expand_call_inline): Check option variables.
9024 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9026 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9028 PR tree-optimization/84321
9029 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9030 handling. Also check whether the anti-range contains any values
9031 that satisfy the mask; switch to a VR_RANGE if not.
9033 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9036 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9038 2018-02-13 Martin Jambor <mjambor@suse.cz>
9041 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9042 of call statements, also set location of a load to a temporary.
9044 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9046 * config/rl78/rl78.c (add_vector_labels): New function.
9047 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9048 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9049 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9050 which checks that no arguments are passed.
9051 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9052 * doc/extend.texi: Documentation for the new attribute.
9054 2018-02-13 Andreas Schwab <schwab@suse.de>
9056 * config/riscv/linux.h (CPP_SPEC): Define.
9058 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9061 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9062 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9063 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9064 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9065 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9066 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9067 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9068 temporarily for AES and PCLMUL builtins.
9070 PR tree-optimization/84339
9071 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9072 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9076 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9077 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9078 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9080 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9082 * omp-simd-clone.h: New file.
9083 * omp-simd-clone.c: Include omp-simd-clone.h.
9084 (expand_simd_clones): No longer static.
9085 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9086 cgraph.h and omp-simd-clone.h.
9087 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9088 (vect_recog_widen_shift_pattern): Formatting fix.
9089 (vect_pattern_recog_1): Don't check optab for calls.
9092 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9093 operands[2] into a REG before using gen_lowpart on it.
9095 2018-02-12 Jeff Law <law@redhat.com>
9098 * config/sh/sh.c (find_barrier): Consider a sibling call
9101 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9102 successfully back substituting a reg.
9104 2018-02-12 Richard Biener <rguenther@suse.de>
9106 PR tree-optimization/84037
9107 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9108 parameter, move visited init to caller.
9109 (vect_slp_analyze_operations): Separate cost from validity
9110 check, initialize visited once for all instances.
9111 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9113 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9115 (vect_model_promotion_demotion_cost): Likewise.
9116 (vectorizable_bswap): Guard cost modeling with !slp_node
9117 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9119 (vectorizable_call): Likewise.
9120 (vectorizable_conversion): Likewise.
9121 (vectorizable_assignment): Likewise.
9122 (vectorizable_shift): Likewise.
9123 (vectorizable_operation): Likewise.
9124 (vectorizable_store): Likewise.
9125 (vectorizable_load): Likewise.
9126 (vectorizable_condition): Likewise.
9127 (vectorizable_comparison): Likewise.
9129 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9132 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9133 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9134 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9136 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9139 * params.def (inline-min-speedup): Increase from 8 to 15.
9140 (max-inline-insns-auto): Decrease from 40 to 30.
9141 * ipa-split.c (consider_split): Add some buffer for function to
9142 be considered inlining candidate.
9143 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9146 2018-02-12 Richard Biener <rguenther@suse.de>
9148 PR tree-optimization/84037
9149 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9150 matched stmts if we cannot swap the non-matched ones.
9152 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9154 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9155 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9156 _mm_maskz_scalef_round_ss): New intrinsics.
9157 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9158 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9159 __builtin_ia32_scalefss_round): Remove.
9160 (__builtin_ia32_scalefsd_mask_round,
9161 __builtin_ia32_scalefss_mask_round): New intrinsics.
9162 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9163 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9164 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9165 "<round_constraint>")): Changed to ...
9166 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9167 "<round_scalar_constraint>")): ... this.
9168 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9169 %0, %1, %2<round_op3>}"): Changed to ...
9170 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9171 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9172 %2<round_scalar_mask_op3>}"): ... this.
9173 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9175 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9177 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9178 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9179 (_mm_maskz_sqrt_round_ss): New intrinsics.
9180 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9181 (__builtin_ia32_sqrtsd_mask_round)
9182 (__builtin_ia32_sqrtss_mask_round): New builtins.
9183 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9184 (__builtin_ia32_sqrtss_round): Remove.
9185 (__builtin_ia32_sqrtsd_mask_round)
9186 (__builtin_ia32_sqrtss_mask_round): New builtins.
9187 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9188 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9189 ((match_operand:VF_128 1 "vector_operand"
9190 "xBm,<round_constraint>")): Changed to ...
9191 ((match_operand:VF_128 1 "vector_operand"
9192 "xBm,<round_scalar_constraint>")): ... this.
9193 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9194 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9195 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9196 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9197 %<iptr>1<round_scalar_mask_op3>}): ... this.
9198 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9199 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9201 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9204 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9205 Cast vec_cmpeq result to correct type.
9206 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9207 Cast vec_cmpgt result to correct type.
9209 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9211 * final.c (final_scan_insn_1): Renamed from...
9212 (final_scan_insn): ... this. New wrapper, to recover
9213 seen from the outermost call in recursive ones.
9214 * config/sparc/sparc.c (output_return): Drop seen from call.
9215 (output_sibcall): Likewise.
9216 * config/visium/visium.c (output_branch): Likewise.
9218 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9220 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9223 2018-02-10 Alan Modra <amodra@gmail.com>
9226 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9227 Specify LR as an input.
9229 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9232 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9233 remove_member_access_dummy_vars): New functions.
9234 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9235 lower_omp_1, execute_lower_omp): Use them.
9237 PR rtl-optimization/84308
9238 * shrink-wrap.c (spread_components): Release todo vector.
9240 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9242 PR rtl-optimization/57193
9243 * ira-color.c (struct allocno_color_data): Add member
9244 conflict_allocno_hard_prefs.
9245 (update_conflict_allocno_hard_prefs): New.
9246 (bucket_allocno_compare_func): Add a preference based on
9247 conflict_allocno_hard_prefs.
9248 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9249 (color_allocnos): Remove a dead code. Initiate
9250 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9252 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9255 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9256 constraint from =wa to wa. Avoid a subreg on the output operand,
9257 instead use a pseudo and subreg it in a move.
9258 (p9_xxbrd_<mode>): Changed to ...
9259 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9260 (p9_xxbrd_v2df): New expander.
9261 (p9_xxbrw_<mode>): Changed to ...
9262 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9263 (p9_xxbrw_v4sf): New expander.
9265 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9267 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9269 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9272 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9273 multiply in 32-bit mode.
9274 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9275 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9278 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9280 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9281 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9282 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9283 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9285 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9288 * dwarf2out.c (is_trivial_indirect_ref): New function.
9289 (dwarf2out_late_global_decl): Do not generate a location
9290 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9291 and that are not defined in the current unit.
9293 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9295 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9296 instead of a libcall for UNORDERED.
9298 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9301 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9302 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9304 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9307 * config/s390/s390.c (s390_set_current_function): Invoke
9308 s390_indirect_branch_settings also if fndecl didn't change.
9310 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9312 * config/rs6000/rs6000.md (blockage): Set length to zero.
9314 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9316 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9318 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9321 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9322 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9326 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9327 PARALLEL incoming that failed vt_get_decl_and_offset check.
9330 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9331 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9333 (get_variable_section): For decls in named .bss* sections pass true as
9334 second argument to bss_initializer_p.
9336 2018-02-09 Marek Polacek <polacek@redhat.com>
9337 Jakub Jelinek <jakub@redhat.com>
9340 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9341 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9342 Sync some changes from cxx_fold_indirect_ref.
9344 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9346 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9348 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9349 (BLOCK_INLINE_ENTRY_LABEL): New.
9350 (dwarf2out_var_location): Disregard inline entry markers.
9351 (inline_entry_data): New struct.
9352 (inline_entry_data_hasher): New hashtable type.
9353 (inline_entry_data_hasher::hash): New.
9354 (inline_entry_data_hasher::equal): New.
9355 (inline_entry_data_table): New variable.
9356 (add_high_low_attributes): Add DW_AT_entry_pc and
9357 DW_AT_GNU_entry_view attributes if a pending entry is found
9358 in inline_entry_data_table. Add old entry_pc attribute only
9359 if debug nonbinding markers are disabled.
9360 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9361 markers are enabled.
9362 (block_within_block_p, dwarf2out_inline_entry): New.
9363 (dwarf2out_finish): Check that no entries remained in
9364 inline_entry_data_table.
9365 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9366 (final_scan_insn, notice_source_line): Likewise.
9367 (rest_of_clean_state): Skip inline entry markers.
9368 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9370 * gimple.c (gimple_build_debug_inline_entry): New.
9371 * gimple.h (enum gimple_debug_subcode): Add
9372 GIMPLE_DEBUG_INLINE_ENTRY.
9373 (gimple_build_debug_inline_entry): Declare.
9374 (gimple_debug_inline_entry_p): New.
9375 (gimple_debug_nonbind_marker_p): Adjust.
9376 * insn-notes.def (INLINE_ENTRY): New.
9377 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9378 inline entry marker notes.
9379 (print_insn): Likewise.
9380 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9381 (INSN_DEBUG_MARKER_KIND): Likewise.
9382 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9383 * tree-inline.c (expand_call_inline): Build and insert
9384 debug_inline_entry stmt.
9385 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9386 inline entry blocks early, if nonbind markers are enabled.
9387 (dump_scope_block): Dump fragment info.
9388 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9389 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9390 (gimple_build_debug_inline_entry): New.
9391 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9392 Enable/disable inline entry points too.
9393 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9394 (DEBUG_INSN): Describe inline entry markers.
9396 * common.opt (gvariable-location-views): New.
9397 (gvariable-location-views=incompat5): New.
9398 * config.in: Rebuilt.
9399 * configure: Rebuilt.
9400 * configure.ac: Test assembler for view support.
9401 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9402 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9403 * dwarf2out.c (var_loc_view): New typedef.
9404 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9405 (dwarf2out_locviews_in_attribute): New.
9406 (dwarf2out_locviews_in_loclist): New.
9407 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9408 (enum dw_line_info_opcode): Add LI_adv_address.
9409 (struct dw_line_info_table): Add view.
9410 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9411 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9412 (zero_view_p): New variable.
9413 (ZERO_VIEW_P): New macro.
9414 (output_asm_line_debug_info): New.
9415 (struct var_loc_node): Add view.
9416 (add_AT_view_list, AT_loc_list): New.
9417 (add_var_loc_to_decl): Add view param. Test it against last.
9418 (new_loc_list): Add view params. Record them.
9419 (AT_loc_list_ptr): Handle loc and view lists.
9420 (view_list_to_loc_list_val_node): New.
9421 (print_dw_val): Handle dw_val_class_view_list.
9422 (size_of_die): Likewise.
9423 (value_format): Likewise.
9424 (loc_list_has_views): New.
9425 (gen_llsym): Set vl_symbol too.
9426 (maybe_gen_llsym, skip_loc_list_entry): New.
9427 (dwarf2out_maybe_output_loclist_view_pair): New.
9428 (output_loc_list): Output view list or entries too.
9429 (output_view_list_offset): New.
9430 (output_die): Handle dw_val_class_view_list.
9431 (output_dwarf_version): New.
9432 (output_compilation_unit_header): Use it.
9433 (output_skeleton_debug_sections): Likewise.
9434 (output_rnglists, output_line_info): Likewise.
9435 (output_pubnames, output_aranges): Update version comments.
9436 (output_one_line_info_table): Output view numbers in asm comments.
9437 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9438 Call maybe_gen_llsym.
9439 (loc_list_from_tree_1): Adjust.
9440 (add_AT_location_description): Create view list attribute if
9441 needed, check it's absent otherwise.
9442 (convert_cfa_to_fb_loc_list): Adjust.
9443 (maybe_emit_file): Call output_asm_line_debug_info for test.
9444 (dwarf2out_var_location): Reset views as needed. Precompute
9445 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9446 attribute. Set view.
9447 (new_line_info_table): Reset next view.
9448 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9449 (dwarf2out_source_line): Likewise. Output view resets and labels to
9450 the assembler, or select appropriate line info opcodes.
9451 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9452 (optimize_string_length): Catch it. Adjust.
9453 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9454 dw_val_class_view_list, and remove it if no longer needed.
9455 (hash_loc_list): Hash view numbers.
9456 (loc_list_hasher::equal): Compare them.
9457 (optimize_location_lists): Check whether a view list symbol is
9458 needed, and whether the locview attribute is present, and
9459 whether they match. Remove the locview attribute if no longer
9461 (index_location_lists): Call skip_loc_list_entry for test.
9462 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9463 Use output_dwarf_version.
9464 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9465 (struct dw_val_node): Add val_view_list.
9466 * final.c (SEEN_NEXT_VIEW): New.
9467 (set_next_view_needed): New.
9468 (clear_next_view_needed): New.
9469 (maybe_output_next_view): New.
9470 (final_start_function): Rename to...
9471 (final_start_function_1): ... this. Take pointer to FIRST,
9472 add SEEN parameter. Emit param bindings in the initial view.
9473 (final_start_function): Reintroduce SEEN-less interface.
9474 (final): Rename to...
9475 (final_1): ... this. Take SEEN parameter. Output final pending
9476 next view at the end.
9477 (final): Reintroduce seen-less interface.
9478 (final_scan_insn): Output pending next view before switching
9479 sections or ending a block. Mark the next view as needed when
9480 outputting variable locations. Notify debug backend of section
9481 changes, and of location view changes.
9482 (rest_of_handle_final): Adjust.
9483 * toplev.c (process_options): Autodetect value for debug variable
9484 location views option. Warn on incompat5 without -gdwarf-5.
9485 * doc/invoke.texi (gvariable-location-views): New.
9486 (gvariable-location-views=incompat5): New.
9487 (gno-variable-location-views): New.
9489 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9491 PR tree-optimization/84136
9492 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9493 that the result of find_edge is non-NULL.
9495 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9498 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9499 storing integer register in SImode. Fix cost of 256 and 512
9500 byte aligned SSE register store.
9502 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9504 * config/i386/i386.c (ix86_multiplication_cost): Fix
9505 multiplication cost for TARGET_AVX512DQ.
9507 2018-02-08 Marek Polacek <polacek@redhat.com>
9509 PR tree-optimization/84238
9510 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9513 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9515 PR tree-optimization/84265
9516 * tree-vect-stmts.c (vectorizable_store): Don't treat
9517 VMAT_CONTIGUOUS accesses as grouped.
9518 (vectorizable_load): Likewise.
9520 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9522 PR tree-optimization/81635
9523 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9524 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9525 (test_round_for_mask): New functions.
9526 (wide_int_cc_tests): Call test_round_for_mask.
9527 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9528 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9529 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9530 range returned by get_range_info.
9532 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9535 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9536 * symtab.c: Include builtins.h
9537 (symtab_node::output_to_lto_symbol_table_p): Move here
9538 from lto-streamer-out.c:output_symbol_p.
9539 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9540 (output_symbol_p): Move all logic to symtab.c
9541 (produce_symtab): Update.
9543 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9545 * config/s390/s390-opts.h (enum indirect_branch): Define.
9546 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9547 (s390_indirect_branch_via_thunk)
9548 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9549 (enum s390_indirect_branch_type): Define.
9550 * config/s390/s390.c (struct s390_frame_layout, struct
9551 machine_function): Remove.
9552 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9553 (indirect_branch_table_label_no, indirect_branch_table_name):
9555 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9556 (enum s390_indirect_branch_option): Define.
9557 (s390_return_addr_from_memory): New function.
9558 (s390_handle_string_attribute): New function.
9559 (s390_attribute_table): Add new attribute handler.
9560 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9561 (s390_indirect_branch_via_thunk): New function.
9562 (s390_indirect_branch_via_inline_thunk): New function.
9563 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9564 sibling call optimization for non z10 compiles.
9565 (s390_emit_call): Force indirect branch target to be a single
9566 register. Add r1 clobber for non-z10 compiles.
9567 (s390_emit_epilogue): Emit return jump via return_use expander.
9568 (s390_reorg): Handle JUMP_INSNs as execute targets.
9569 (s390_option_override_internal): Perform validity checks for the
9570 new command line options.
9571 (s390_indirect_branch_attrvalue): New function.
9572 (s390_indirect_branch_settings): New function.
9573 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9574 (s390_output_indirect_thunk_function): New function.
9575 (s390_code_end): Implement target hook.
9576 (s390_case_values_threshold): Implement target hook.
9577 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9579 * config/s390/s390.h (struct s390_frame_layout)
9580 (struct machine_function): Move here from s390.c.
9581 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9582 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9583 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9584 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9585 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9586 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9587 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9588 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9589 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9590 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9591 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9592 (mnemonic attribute): Add values which aren't recognized
9594 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9595 pattern for branch conversion. Fix mnemonic attribute.
9596 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9597 indirect branch via thunk if requested.
9598 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9599 ("*indirect_jump"): Disable for branch conversion using out of
9601 ("indirect_jump_via_thunk<mode>_z10")
9602 ("indirect_jump_via_thunk<mode>")
9603 ("indirect_jump_via_inlinethunk<mode>_z10")
9604 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9605 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9606 ("casesi_jump_via_inlinethunk<mode>_z10")
9607 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9608 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9609 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9610 ("*indirect2_jump"): Disable for branch conversion.
9611 ("casesi_jump"): Turn into expander and expand patterns for branch
9613 ("return_use"): New expander.
9614 ("*return"): Emit return via thunk and rename it to ...
9615 ("*return<mode>"): ... this one.
9616 * config/s390/s390.opt: Add new options and and enum for the
9619 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9621 * lra-constraints.c (match_reload): Unconditionally use
9622 gen_lowpart_SUBREG, rather than selecting between that
9623 and equivalent gen_rtx_SUBREG code.
9625 2018-02-08 Richard Biener <rguenther@suse.de>
9627 PR tree-optimization/84233
9628 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9629 changed flag instead of boguously re-using phi_inserted.
9631 2018-02-08 Martin Jambor <mjambor@suse.cz>
9633 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9634 static local variables.
9636 2018-02-08 Richard Biener <rguenther@suse.de>
9638 PR tree-optimization/84278
9639 * tree-vect-stmts.c (vectorizable_store): When looking for
9640 smaller vector types to perform grouped strided loads/stores
9641 make sure the mode is supported by the target.
9642 (vectorizable_load): Likewise.
9644 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9646 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9647 Increase LDP/STP opportunities by adding adjacent callee-saves.
9649 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9651 PR rtl-optimization/84068
9652 PR rtl-optimization/83459
9653 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9655 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9657 PR tree-optimization/84224
9658 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9659 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9662 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9665 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9666 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9667 count, remove one USE.
9669 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9671 * doc/install.texi (Configuration): Document the
9672 --with-long-double-format={ibm,ieee} PowerPC configuration
9676 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9677 Convert from define_expand to be define_insn_and_split. Rework
9678 float/double/_Float128 conversions to QI/HI/SImode to work with
9679 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9680 conversions to QI/HImode types did a store and then a load to
9681 truncate the value. For conversions to VSX registers, don't split
9682 the insn, instead emit the code directly. Use the code iterator
9683 any_fix to combine signed and unsigned conversions.
9684 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9685 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9686 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9687 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9688 (fix_<mode>di2_hw): Likewise.
9689 (fixuns_<mode>di2_hw): Likewise.
9690 (fix_<mode>si2_hw): Likewise.
9691 (fixuns_<mode>si2_hw): Likewise.
9692 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9693 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9694 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9695 fix<uns>_trunc<SFDF:mode>si2_p8.
9696 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9697 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9698 (fix<uns>_<mode>_mem): Likewise.
9699 (fctiw<u>z_<mode>_mem): Likewise.
9700 (fix<uns>_<mode>_mem): Likewise.
9701 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9702 the register allocator from doing a direct move to the GPRs to do
9703 a store, and instead use the ISA 3.0 store byte/half-word from
9704 vector register instruction. For IEEE 128-bit floating point,
9705 also optimize stores of 32-bit ints.
9706 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9708 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9710 * genextract.c (push_pathstr_operand): New function to support
9712 (walk_rtx): Call push_pathstr_operand.
9713 (print_path): Support [a-zA-Z].
9715 2018-02-07 Richard Biener <rguenther@suse.de>
9717 PR tree-optimization/84037
9718 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9719 (cse_and_gimplify_to_preheader): Declare.
9720 (vect_get_place_in_interleaving_chain): Likewise.
9721 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9723 (_loop_vec_info::~_loop_vec_info): Delete it.
9724 (cse_and_gimplify_to_preheader): New function.
9725 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9726 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9727 (vectorizable_load): Likewise. For grouped stores always base
9728 the IV on the first element.
9729 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9730 condition before gimplifying.
9732 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9734 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9735 *DIV_EXPR and *MOD_EXPR.
9737 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9740 * config/i386/i386.c (ix86_option_override_internal): Mask out
9741 the CF_SET bit when checking -fcf-protection.
9743 2018-02-07 Tom de Vries <tom@codesourcery.com>
9746 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9749 2018-02-07 Richard Biener <rguenther@suse.de>
9751 PR tree-optimization/84204
9752 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9755 PR tree-optimization/84205
9756 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9757 special-case isl_ast_op_zdiv_r.
9759 PR tree-optimization/84223
9760 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9761 Only add conditions from within the region.
9762 (gather_bbs::after_dom_children): Adjust.
9764 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9767 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9768 * config/avr/avr.md: Only post-reload split REG-REG moves if
9769 either register is GENERAL_REG_P.
9771 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9773 PR tree-optimization/84235
9774 * tree-ssa-scopedtables.c
9775 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9776 if the subtraction is performed in floating point type where NaNs are
9777 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9778 build 1. Formatting fix.
9780 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9783 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9784 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9785 and skip it regardless of bb boundaries. Use CALL_P macro,
9786 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9787 unnecessarily, formatting fix.
9789 2018-02-06 Michael Collison <michael.collison@arm.com>
9791 * config/arm/thumb2.md:
9792 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9793 (*thumb_mov_notscc): Ditto.
9795 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9798 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9799 unsigned_fix, not "s".
9801 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9803 * configure.ac (gcc_fn_eh_frame_ro): New function.
9804 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9805 correct .eh_frame permissions.
9806 * configure: Regenerate.
9808 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9810 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9813 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9815 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9816 Display warning message for -mno-speculate-indirect-jumps.
9818 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9820 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9822 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9824 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9826 PR tree-optimization/84225
9827 * tree-eh.c (find_trapping_overflow): Only call
9828 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9830 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9833 * config/i386/i386.c: Reimplement the check of possible options
9834 -mibt/-mshstk conbination. Change error messages.
9835 * doc/invoke.texi: Fix a typo: remove extra '='.
9837 2018-02-06 Marek Polacek <polacek@redhat.com>
9839 PR tree-optimization/84228
9840 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9842 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9845 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9846 emitted arch directives.
9847 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9848 __ARM_FEATURE_COPROC before changing architectures.
9850 2018-02-06 Richard Biener <rguenther@suse.de>
9852 * config/i386/i386.c (print_reg): Fix typo.
9853 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9855 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9857 * configure: Regenerate.
9859 2018-02-05 Martin Sebor <msebor@redhat.com>
9861 PR tree-optimization/83369
9862 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9865 2018-02-05 Martin Liska <mliska@suse.cz>
9867 * doc/invoke.texi: Cherry-pick upstream r323995.
9869 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9871 * ira.c (ira_init_register_move_cost): Adjust comment.
9873 2018-02-05 Martin Liska <mliska@suse.cz>
9875 PR gcov-profile/84137
9876 * doc/gcov.texi: Fix typo in documentation.
9878 2018-02-05 Martin Liska <mliska@suse.cz>
9880 PR gcov-profile/83879
9881 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9882 using dlopen functionality.
9884 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9886 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9887 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9888 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9889 _mm_maskz_range_round_ss): New intrinsics.
9890 (__builtin_ia32_rangesd128_round)
9891 (__builtin_ia32_rangess128_round): Remove.
9892 (__builtin_ia32_rangesd128_mask_round,
9893 __builtin_ia32_rangess128_mask_round): New builtins.
9894 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9895 __builtin_ia32_rangess128_round): Remove.
9896 (__builtin_ia32_rangesd128_mask_round,
9897 __builtin_ia32_rangess128_mask_round): New builtins.
9898 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9899 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9900 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9901 "<round_saeonly_constraint>")): Changed to ...
9902 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9903 "<round_saeonly_scalar_constraint>")): ... this.
9904 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9905 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9906 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9907 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9908 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9910 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9912 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9914 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9915 Remove all values except native, 8540 and 8548.
9917 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9919 * config/i386/i386.c (ix86_output_function_return): Pass
9920 INVALID_REGNUM, instead of -1, as invalid register number to
9921 indirect_thunk_name and output_indirect_thunk.
9923 2018-02-02 Julia Koval <julia.koval@intel.com>
9925 * config.gcc: Add -march=icelake.
9926 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9927 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9928 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9929 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9930 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9931 (processor_target_table): Add icelake.
9932 (ix86_option_override_internal): Handle new PTAs.
9933 (get_builtin_code_for_version): Handle icelake.
9934 (M_INTEL_COREI7_ICELAKE): New.
9935 (fold_builtin_cpu): Handle icelake.
9936 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9937 * doc/invoke.texi: Add -march=icelake.
9939 2018-02-02 Julia Koval <julia.koval@intel.com>
9941 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9942 to wide_int_bitmask.
9943 * wide-int-bitmask.h: New.
9945 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9948 * config/i386/i386.md: Replace Pmode with word_mode in
9949 builtin_setjmp_setup and builtin_longjmp to support x32.
9951 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9955 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9957 (rs6000_supported_cpu_names): New static variable.
9958 (linux_cpu_translation_table): Likewise.
9959 (elf_platform) <cpu>: Define new static variable and use it.
9960 Translate kernel AT_PLATFORM name to canonical name if needed.
9961 Error if platform name is unknown.
9963 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9966 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9968 2018-02-01 Jeff Law <law@redhat.com>
9971 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9972 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9973 the scratch if RELEASE_VIA_POP is false.
9974 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9975 If we have to save a temporary register, decrement SIZE appropriately.
9976 Pass new arguments to release_scratch_register_on_entry.
9977 (ix86_adjust_stack_and_probe): Likewise.
9978 (ix86_emit_probe_stack_range): Pass new arguments to
9979 release_scratch_register_on_entry.
9981 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9983 PR rtl-optimization/84157
9984 * combine.c (change_zero_ext): Use REG_P predicate in
9985 front of HARD_REGISTER_P predicate.
9987 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9989 * config/avr/avr.c (avr_option_override): Move disabling of
9990 -fdelete-null-pointer-checks to...
9991 * common/config/avr/avr-common.c (avr_option_optimization_table):
9994 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9996 PR tree-optimization/81635
9997 * tree-data-ref.c (split_constant_offset_1): For types that
9998 wrap on overflow, try to use range info to prove that wrapping
10001 2018-02-01 Renlin Li <renlin.li@arm.com>
10004 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10005 TAILCALL_ADDR_REGS.
10006 (aarch64_register_move_cost): Likewise.
10007 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10008 TAILCALL_ADDR_REGS.
10009 (REG_CLASS_NAMES): Likewise.
10010 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10011 TAILCALL_ADDR_REGS. Remove IP registers.
10012 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10014 2018-02-01 Richard Biener <rguenther@suse.de>
10016 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10017 for specifying RPO order and allow NULL for that.
10018 * domwalk.c (dom_walker::dom_walker): Likewise.
10019 (dom_walker::walk): Handle NULL RPO order.
10020 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10022 (rewrite_update_dom_walker): Likewise.
10023 (mark_def_dom_walker): Likewise.
10025 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10027 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10028 (aarch64_maybe_expand_sve_subreg_move): Declare.
10029 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10030 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10032 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10033 that are semantically a reverse operation.
10034 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10035 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10036 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10038 (aarch64_can_change_mode_class): For big-endian, forbid changes
10039 between two SVE modes if they have different element sizes.
10041 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10043 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10044 the TImode handling for big-endian targets.
10046 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10048 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10049 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10051 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10052 Remove BSWAP handing for big-endian targets and use the form of
10053 LD1RQ appropariate for the mode.
10055 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10057 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10058 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10059 duplicated element.
10061 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10064 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10065 check for operands that need to go through aarch64_sve_reload_be.
10067 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10069 PR tree-optimization/81661
10070 PR tree-optimization/84117
10071 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10072 * tree-eh.c: Include gimplify.h.
10073 (find_trapping_overflow, replace_trapping_overflow,
10074 rewrite_to_non_trapping_overflow): New functions.
10075 * tree-vect-loop.c: Include tree-eh.h.
10076 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10077 * tree-data-ref.c: Include tree-eh.h.
10078 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10080 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10082 PR rtl-optimization/84123
10083 * combine.c (change_zero_ext): Check if hard register satisfies
10084 can_change_dest_mode before calling gen_lowpart_SUBREG.
10086 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10089 * ira.c (ira_init_register_move_cost): Remove assert.
10091 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10093 PR rtl-optimization/84071
10094 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10095 * doc/tm.texi: Regenerate.
10097 2018-01-31 Richard Biener <rguenther@suse.de>
10099 PR tree-optimization/84132
10100 * tree-data-ref.c (analyze_miv_subscript): Properly
10101 check whether evolution_function_is_affine_multivariate_p
10102 before calling gcd_of_steps_may_divide_p.
10104 2018-01-31 Julia Koval <julia.koval@intel.com>
10107 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10108 * config/i386/i386.md (rdpid_rex64) New.
10109 (rdpid): Make 32bit only.
10111 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10114 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10115 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10117 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10120 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10122 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10124 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10126 PR rtl-optimization/84071
10127 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10128 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10130 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10132 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10133 (arc_attribute_table): Add 'aux' attribute.
10134 (arc_in_small_data_p): Consider aux like variables.
10135 (arc_is_aux_reg_p): New function.
10136 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10137 (arc_get_aux_arg): New function.
10138 (prepare_move_operands): Handle aux-register access.
10139 (arc_handle_aux_attribute): New function.
10140 * doc/extend.texi (ARC Variable attributes): Add subsection.
10142 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10144 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10145 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10146 (arc_attribute_table): Add 'uncached' attribute.
10147 (arc_print_operand): Print '.di' flag for uncached memory
10149 (arc_in_small_data_p): Do not consider for small data the uncached
10151 (arc_is_uncached_mem_p): New function.
10152 * config/arc/predicates.md (compact_store_memory_operand): Check
10153 for uncached memory accesses.
10154 (nonvol_nonimm_operand): Likewise.
10155 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10157 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10160 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10161 falign-loops=): Add Optimization flag.
10163 2018-01-30 Jeff Law <law@redhat.com>
10166 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10167 INT_REGISTERS_SAVED. Check it prior to calling
10168 get_scratch_register_on_entry.
10169 (ix86_adjust_stack_and_probe): Similarly.
10170 (ix86_emit_probe_stack_range): Similarly.
10171 (ix86_expand_prologue): Corresponding changes.
10173 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10176 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10177 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10179 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10182 * lra-constraints.c (curr_insn_transform): Process AND in the
10185 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10187 PR rtl-optimization/83986
10188 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10189 dependence against last_pending_memory_flush in addition to
10190 pending_jump_insns.
10192 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10194 PR tree-optimization/81611
10195 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10198 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10201 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10204 2018-01-30 Richard Biener <rguenther@suse.de>
10205 Jakub Jelinek <jakub@redhat.com>
10207 PR tree-optimization/84111
10208 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10209 inner loops added during recursion, as they don't have up-to-date
10212 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10215 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10216 (can_inline_edge_by_limits_p): ... here.
10217 (can_early_inline_edge_p, check_callers,
10218 update_caller_keys, update_callee_keys, recursive_inlining,
10219 add_new_edges_to_heap, speculation_useful_p,
10220 inline_small_functions,
10221 inline_small_functions, flatten_function,
10222 inline_to_all_callers_1): Update.
10224 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10226 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10229 2018-01-30 Richard Biener <rguenther@suse.de>
10231 PR tree-optimization/83008
10232 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10233 invariant and constant vector uses in stmts when they need
10234 more than one stmt.
10236 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10239 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10240 * configure: Regenerate.
10242 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10244 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10246 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10247 Use gen_rtx_REG rather than gen_lowpart.
10249 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10251 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10252 rather than 0 when creating partial subregs.
10254 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10256 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10259 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10262 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10263 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10264 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10265 flags. This restores the settings used before the 2017-07-24.
10266 Turning off pre increment/decrement/modify allows IVOPTS to
10267 optimize DF/SF loops where the index is an int.
10269 2018-01-29 Richard Biener <rguenther@suse.de>
10270 Kelvin Nilsen <kelvin@gcc.gnu.org>
10273 * tree-vect-stmts.c (vectorizable_call): Don't call
10274 targetm.vectorize_builtin_md_vectorized_function if callee is
10277 2018-01-22 Carl Love <cel@us.ibm.com>
10279 * doc/extend.tex: Fix typo in second arg in
10280 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10282 2018-01-29 Richard Biener <rguenther@suse.de>
10284 PR tree-optimization/84086
10285 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10286 (flush_ssaname_freelist): When SSA names were released reset
10287 the SCEV hash table.
10289 2018-01-29 Richard Biener <rguenther@suse.de>
10291 PR tree-optimization/84057
10292 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10293 removed paths when removing edges.
10295 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10297 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10298 -mfunction-return=@var{choice}.
10300 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10302 PR diagnostic/84034
10303 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10304 Handle CR like TAB.
10305 (layout::print_source_line): Likewise.
10306 (test_get_line_width_without_trailing_whitespace): Add test cases.
10308 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10310 PR middle-end/84040
10311 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10314 2018-01-26 Jim Wilson <jimw@sifive.com>
10316 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10318 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10321 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10323 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10324 and CMP + SUB-immediate -> SUBS.
10326 2018-01-26 Martin Sebor <msebor@redhat.com>
10328 PR tree-optimization/83896
10329 * tree-ssa-strlen.c (get_string_len): Rename...
10330 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10331 Avoid assuming length is constant.
10332 (handle_char_store): Use HOST_WIDE_INT for string length.
10334 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10337 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10338 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10340 2018-01-26 Richard Biener <rguenther@suse.de>
10342 PR rtl-optimization/84003
10343 * dse.c (record_store): Only record redundant stores when
10344 the earlier store aliases at least all accesses the later one does.
10346 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10348 PR rtl-optimization/83985
10349 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10350 REG_CFA_RESTORE insns.
10351 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10352 REG_CFA_RESTORE insns here.
10355 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10356 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10358 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10360 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10361 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10362 (arc_init): Likewise.
10363 (arc_override_options): Likewise.
10364 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10366 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10368 * config/arc/arc.h (TARGET_DBNZ): Define.
10369 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10370 properly set the tune attribute.
10371 (dbnz): Use TARGET_DBNZ guard.
10372 * config/arc/arc.opt (mtune): Add core3 option.
10374 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10376 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10377 recognize new pic like addresses.
10378 (arc_delegitimize_address): Clean up.
10380 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10382 * config/arc/arc-arches.def: Option mrf16 valid for all
10384 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10385 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10386 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10387 * config/arc/arc-tables.opt: Regenerate.
10388 * config/arc/arc.c (arc_conditional_register_usage): Handle
10389 reduced register file case.
10390 (arc_file_start): Set must have build attributes.
10391 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10392 mrf16 option value.
10393 * config/arc/arc.opt (mrf16): Add new option.
10394 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10395 * config/arc/genmultilib.awk: Handle new mrf16 option.
10396 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10397 * config/arc/t-multilib: Regenerate.
10398 * doc/invoke.texi (ARC Options): Document mrf16 option.
10400 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10402 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10403 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10404 (arc_attribute_table): Add 'secure_call' attribute.
10405 (arc_print_operand): Print secure call operand.
10406 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10408 (arc_is_secure_call_p): New function. * config/arc/arc.md
10409 (call_i): Add support for sjli instruction.
10410 (call_value_i): Likewise.
10411 * config/arc/constraints.md (Csc): New constraint.
10413 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10414 John Eric Martin <John.Martin@emmicro-us.com>
10416 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10417 * config/arc/arc.c (_arc_jli_section): New struct.
10418 (arc_jli_section): New type.
10419 (rc_jli_sections): New static variable.
10420 (arc_handle_jli_attribute): New function.
10421 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10422 (arc_file_end): New function.
10423 (TARGET_ASM_FILE_END): Define.
10424 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10425 (arc_add_jli_section): New function.
10426 (jli_call_scan): Likewise.
10427 (arc_reorg): Call jli_call_scan.
10428 (arc_output_addsi): Remove 'S' from printing asm operand.
10429 (arc_is_jli_call_p): New function.
10430 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10432 (movhi_insn): Likewise.
10433 (movsi_insn): Likewise.
10434 (movsi_set_cc_insn): Likewise.
10435 (loadqi_update): Likewise.
10436 (load_zeroextendqisi_update): Likewise.
10437 (load_signextendqisi_update): Likewise.
10438 (loadhi_update): Likewise.
10439 (load_zeroextendhisi_update): Likewise.
10440 (load_signextendhisi_update): Likewise.
10441 (loadsi_update): Likewise.
10442 (loadsf_update): Likewise.
10443 (movsicc_insn): Likewise.
10444 (bset_insn): Likewise.
10445 (bxor_insn): Likewise.
10446 (bclr_insn): Likewise.
10447 (bmsk_insn): Likewise.
10448 (bicsi3_insn): Likewise.
10449 (cmpsi_cc_c_insn): Likewise.
10450 (movsi_ne): Likewise.
10451 (movsi_cond_exec): Likewise.
10452 (clrsbsi2): Likewise.
10453 (norm_f): Likewise.
10463 (call_i): Remove 'S' asm letter, add jli instruction.
10464 (call_value_i): Likewise.
10465 * config/arc/arc.op (mjli-always): New option.
10466 * config/arc/constraints.md (Cji): New constraint.
10467 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10469 (subsf3_fpx): Likewise.
10470 (mulsf3_fpx): Likewise.
10471 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10473 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10474 function attrbutes.
10475 * doc/invoke.texi (ARC): Document mjli-always option.
10477 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10479 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10480 avoid addition with 0 and use incw and decw where possible.
10482 2018-01-26 Richard Biener <rguenther@suse.de>
10484 PR tree-optimization/81082
10485 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10486 association if it requires casting to unsigned.
10487 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10488 from fold_plusminus_mult_expr to catch important cases late when
10489 range info is available.
10491 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10493 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10494 * configure.ac (hidden_linkonce): New test.
10495 * configure: Regenerate.
10496 * config.in: Regenerate.
10498 2018-01-26 Julia Koval <julia.koval@intel.com>
10500 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10501 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10502 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10503 _mm_mask_bitshuffle_epi64_mask): Fix type.
10504 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10505 USI_FTYPE_V4DI_V4DI_USI): Remove.
10506 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10507 __builtin_ia32_vpshufbitqmb256_mask,
10508 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10509 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10510 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10512 2018-01-26 Alan Modra <amodra@gmail.com>
10515 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10516 UNSPEC_VBPERMQ. Sort other unspecs.
10518 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10520 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10522 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10524 PR middle-end/83055
10525 * predict.c (drop_profile): Do not push/pop cfun; update also
10527 (handle_missing_profiles): Fix logic looking for zero profiles.
10529 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10531 PR middle-end/83977
10532 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10533 on functions with #pragma omp declare simd or functions with simd
10535 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10536 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10537 Remove trailing \n from warning_at calls.
10539 2018-01-25 Tom de Vries <tom@codesourcery.com>
10542 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10543 for neutered workers.
10545 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10548 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10549 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10551 2018-01-24 Jeff Law <law@redhat.com>
10554 * i386.c (get_probe_interval): Move to earlier point.
10555 (ix86_compute_frame_layout): If -fstack-clash-protection and
10556 the frame is larger than the probe interval, then use pushes
10557 to save registers rather than reg->mem moves.
10558 (ix86_expand_prologue): Remove conditional for int_registers_saved
10561 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10564 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10565 min/max for never referenced object.
10567 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10569 PR middle-end/83977
10570 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10572 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10573 attributes from DECL_ATTRIBUTES (decl) without affecting
10574 DECL_ATTRIBUTES (current_function_decl).
10575 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10576 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10578 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10580 PR tree-optimization/83979
10581 * fold-const.c (fold_comparison): Use constant_boolean_node
10582 instead of boolean_{true,false}_node.
10584 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10586 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10589 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10591 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10592 Simplify the clause that sets the length attribute.
10593 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10594 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10595 clause that sets the length attribute.
10596 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10598 2018-01-24 Tom de Vries <tom@codesourcery.com>
10601 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10602 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10603 Add strict parameter.
10604 (prevent_branch_around_nothing): Insert dummy insn between branch to
10605 label and label with no ptx insn inbetween.
10606 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10608 2018-01-24 Tom de Vries <tom@codesourcery.com>
10611 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10612 for neutered threads in warp.
10613 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10615 2018-01-24 Richard Biener <rguenther@suse.de>
10617 PR tree-optimization/83176
10618 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10621 2018-01-24 Richard Biener <rguenther@suse.de>
10623 PR tree-optimization/82819
10624 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10625 code generating pluses that are no-ops in the target precision.
10627 2018-01-24 Richard Biener <rguenther@suse.de>
10629 PR middle-end/84000
10630 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10632 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10634 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10635 to merge probabilities.
10636 * predict.c (probably_never_executed): Also mark as cold functions
10637 with global 0 profile and guessed local profile.
10638 * profile-count.c (profile_probability::combine_with_count): New
10640 * profile-count.h (profile_probability::operator*,
10641 profile_probability::operator*=, profile_probability::operator/,
10642 profile_probability::operator/=): Reduce precision to adjusted
10643 and set value to guessed on contradictory divisions.
10644 (profile_probability::combine_with_freq): Remove.
10645 (profile_probability::combine_wiht_count): Declare.
10646 (profile_count::force_nonzero):: Set to adjusted.
10647 (profile_count::probability_in):: Set quality to adjusted.
10648 * tree-ssa-tail-merge.c (replace_block_by): Use
10649 combine_with_count.
10651 2018-01-23 Andrew Waterman <andrew@sifive.com>
10652 Jim Wilson <jimw@sifive.com>
10654 * config/riscv/riscv.c (riscv_stack_boundary): New.
10655 (riscv_option_override): Set riscv_stack_boundary. Handle
10656 riscv_preferred_stack_boundary_arg.
10657 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10658 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10659 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10660 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10661 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10662 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10664 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10667 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10668 of struct ix86_frame.
10669 (ix86_expand_epilogue): Likewise. Add a local variable for
10670 the reg_save_offset field in struct ix86_frame.
10672 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10674 PR tree-optimization/82604
10675 * tree-loop-distribution.c (enum partition_kind): New enum item
10676 PKIND_PARTIAL_MEMSET.
10677 (partition_builtin_p): Support above new enum item.
10678 (generate_code_for_partition): Ditto.
10679 (compute_access_range): Differentiate cases that equality can be
10680 proven at all loops, the innermost loops or no loops.
10681 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10682 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10683 (finalize_partitions, distribute_loop): Don't fuse partition of
10684 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10685 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10686 parloop is enabled.
10688 2018-01-23 Martin Liska <mliska@suse.cz>
10690 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10691 order to ignore the predictor.
10692 (PRED_POLYMORPHIC_CALL): Likewise.
10693 (PRED_RECURSIVE_CALL): Likewise.
10695 2018-01-23 Martin Liska <mliska@suse.cz>
10697 * tree-profile.c (tree_profiling): Print function header to
10698 aware reader which function we are working on.
10699 * value-prof.c (gimple_find_values_to_profile): Do not print
10700 not interesting value histograms.
10702 2018-01-23 Martin Liska <mliska@suse.cz>
10704 * profile-count.h (enum profile_quality): Add
10705 profile_uninitialized as the first value. Do not number values
10706 as they are zero based.
10707 (profile_count::verify): Update sanity check.
10708 (profile_probability::verify): Likewise.
10710 2018-01-23 Nathan Sidwell <nathan@acm.org>
10712 * doc/invoke.texi (ffor-scope): Deprecate.
10714 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10716 PR tree-optimization/83510
10717 * domwalk.c (set_all_edges_as_executable): New function.
10718 (dom_walker::dom_walker): Convert bool param
10719 "skip_unreachable_blocks" to enum reachability. Move setup of
10720 edge flags to set_all_edges_as_executable and only do it when
10721 reachability is REACHABLE_BLOCKS.
10722 * domwalk.h (enum dom_walker::reachability): New enum.
10723 (dom_walker::dom_walker): Convert bool param
10724 "skip_unreachable_blocks" to enum reachability.
10725 (set_all_edges_as_executable): New decl.
10726 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10727 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10729 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10730 but converting true to REACHABLE_BLOCKS.
10731 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10733 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10734 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10735 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10737 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10738 if check_all_array_refs will be called.
10740 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10742 * tree.c (selftest::test_location_wrappers): Add more test
10745 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10747 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10748 (selftest::test_bit_in_range): Likewise.
10750 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10753 * doc/sourcebuild.texi (vect_float): Say that the selector
10754 only describes the situation when -funsafe-math-optimizations is on.
10755 (vect_float_strict): Document.
10757 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10759 PR tree-optimization/83965
10760 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10761 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10762 instead of checking only for a reduction.
10763 (vect_recog_widen_sum_pattern): Likewise.
10765 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10767 * predict.c (probably_never_executed): Only use precise profile info.
10768 (compute_function_frequency): Skip after inlining hack since we now
10769 have quality checking.
10771 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10773 * profile-count.h (profile_probability::very_unlikely,
10774 profile_probability::unlikely, profile_probability::even): Set
10775 precision to guessed.
10777 2018-01-23 Richard Biener <rguenther@suse.de>
10779 PR tree-optimization/83963
10780 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10781 Properly terminate dominator walk when crossing the exit edge not
10782 when visiting its source block.
10784 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10787 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10788 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10790 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10792 PR tree-optimization/83957
10793 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10794 semicolon after for body surrounded by braces.
10796 PR tree-optimization/83081
10797 * profile-count.h (profile_probability::split): New method.
10798 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10799 Use profile_probability::split.
10800 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10801 when splitting a single conditional jump into 2.
10803 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10805 PR tree-optimization/69452
10806 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10811 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10812 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10813 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10815 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10817 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10819 * config/rl78/rl78.md (movdi): New define_expand.
10820 * config/rl78/rl78.c (rl78_split_movdi): New function.
10822 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10825 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10827 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10828 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10829 128-bit to produce an UNSPEC move to get the double word with the
10830 signbit and then a shift directly to do signbit.
10831 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10832 implementation with a new version that just does either a direct
10833 move or a regular move. Move memory interface to separate insns.
10834 Move insns so they are next to the expander.
10835 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10836 with signbit move. Split big and little endian case.
10837 (signbit<mode>2_dm_mem_le): Likewise.
10838 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10839 (signbit<mode>2_dm2): Likewise.
10841 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10843 * config/rl78/rl78.md (anddi3): New define_expand.
10845 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10847 * config/rl78/rl78.md (umindi3): New define_expand.
10849 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10851 * config/rl78/rl78.md (smindi3): New define_expand.
10853 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10855 * config/rl78/rl78.md (smaxdi3): New define_expand.
10857 2018-01-22 Carl Love <cel@us.ibm.com>
10859 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10860 LVX_V1TI): Add macro expansion.
10861 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10862 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10863 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10864 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10865 Change check to determine if the instruction is a byte reversing
10866 entry. Fix typo in comment.
10867 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10868 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10869 Add def_builtin calls for new builtins.
10870 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10871 Add define_insn expansion.
10873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10875 * config/rl78/rl78.md (umaxdi3): New define_expand.
10877 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10879 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10880 for non-QImode registers.
10882 2018-01-22 Richard Biener <rguenther@suse.de>
10884 PR tree-optimization/83963
10885 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10886 including the loop exit block.
10887 (scop_detection::merge_sese): Likewise.
10888 (scop_detection::add_scop): Do it here instead.
10890 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10892 * doc/sourcebuild.texi (arm_softfloat): Document.
10894 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10897 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10898 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10899 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10901 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10902 David Edelsohn <dje.gcc@gmail.com>
10905 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10906 Change "crset eq" to "crset 2".
10907 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10908 (*call_indirect_aix<mode>_nospec): Likewise.
10909 (*call_value_indirect_aix<mode>_nospec): Likewise.
10910 (*call_indirect_elfv2<mode>_nospec): Likewise.
10911 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10912 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10913 change assembly output from . to $.
10914 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10915 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10916 (*tablejump<mode>_internal1_nospec): Likewise.
10918 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10921 * config/sh/sh_optimize_sett_clrt.cc:
10922 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10924 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10926 PR tree-optimization/83940
10927 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10928 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10929 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10930 use it to pass back the definition type.
10931 (vect_check_store_rhs): Likewise rhs_dt_out.
10932 (vect_build_gather_load_calls): Add a mask_dt argument and use
10933 it instead of a call to vect_is_simple_use.
10934 (vectorizable_store): Update calls to vect_check_load_store_mask
10935 and vect_check_store_rhs. Use the dt returned by the latter instead
10936 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10937 instead of calls to vect_is_simple_use. Pass the scalar rather
10938 than the vector operand to vect_is_simple_use when handling
10939 second and subsequent copies of an rhs value.
10940 (vectorizable_load): Update calls to vect_check_load_store_mask
10941 and vect_build_gather_load_calls. Use the cached mask_dt and
10942 gs_info.offset_dt instead of calls to vect_is_simple_use.
10944 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10946 PR middle-end/83945
10947 * tree-emutls.c: Include gimplify.h.
10948 (lower_emutls_2): New function.
10949 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10950 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10951 it before further processing.
10954 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10955 UINTVAL (trueop1) instead of INTVAL (op1).
10957 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10961 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10962 INCOMING_FRAME_SP_OFFSET if not defined.
10963 (scan_trace): Add ENTRY argument. If true and
10964 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10965 emit a note to adjust the CFA offset.
10966 (create_cfi_notes): Adjust scan_trace callers.
10967 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10968 INCOMING_FRAME_SP_OFFSET in the CIE.
10969 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10970 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10972 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10973 * doc/tm.texi: Regenerated.
10975 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10977 PR rtl-optimization/83147
10978 * lra-constraints.c (remove_inheritance_pseudos): Use
10979 lra_substitute_pseudo_within_insn.
10981 2018-01-19 Tom de Vries <tom@codesourcery.com>
10982 Cesar Philippidis <cesar@codesourcery.com>
10985 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10987 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10990 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10991 spaces for function labels.
10993 2018-01-19 Martin Liska <mliska@suse.cz>
10995 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10996 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10997 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10998 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10999 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11000 (PRED_CONST_RETURN): Change from 69 to 65.
11001 (PRED_NULL_RETURN): Change from 91 to 71.
11002 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11003 (PRED_LOOP_GUARD): Change from 66 to 73.
11005 2018-01-19 Martin Liska <mliska@suse.cz>
11007 * predict.c (predict_insn_def): Add new assert.
11008 (struct branch_predictor): Change type to signed integer.
11009 (test_prediction_value_range): Amend test to cover
11010 PROB_UNINITIALIZED.
11011 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11012 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11013 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11014 (PRED_LOOP_IV_COMPARE): Likewise.
11015 * predict.h (PROB_UNINITIALIZED): Define new constant.
11017 2018-01-19 Martin Liska <mliska@suse.cz>
11019 * predict.c (dump_prediction): Add new format for
11020 analyze_brprob.py script which is enabled with -details
11022 * profile-count.h (precise_p): New function.
11024 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11026 PR tree-optimization/83922
11027 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11028 there are no statements that need masking.
11029 (vect_active_double_reduction_p): New function.
11030 (vect_analyze_loop_operations): Use it when handling phis that
11031 are not in the loop header.
11033 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11035 PR tree-optimization/83914
11036 * tree-vect-loop.c (vectorizable_induction): Don't convert
11037 init_expr or apply the peeling adjustment for inductions
11038 that are nested within the vectorized loop.
11040 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11042 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11045 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11049 * function.h (gimplify_parameters): Add gimple_seq * argument.
11050 * function.c: Include gimple.h and options.h.
11051 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11052 for the added local temporaries if needed.
11053 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11054 if there are any parameter cleanups, wrap whole body into a
11055 try/finally with the cleanups.
11057 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11060 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11061 Use GET_MODE_CLASS for scalar floating point.
11063 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11067 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11068 Fix call of call_cgraph_insertion_hooks.
11070 2018-01-18 Martin Sebor <msebor@redhat.com>
11072 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11074 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11077 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11080 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11083 * common.opt: (-ffile-prefix-map): New option.
11084 * opts.c (common_handle_option): Defer it.
11085 * opts-global.c (handle_common_deferred_options): Handle it.
11086 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11087 * file-prefix-map.h: New file.
11088 (remap_debug_filename, add_debug_prefix_map): ...here.
11089 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11090 * final.c (debug_prefix_map, add_debug_prefix_map
11091 remap_debug_filename): Move to...
11092 * file-prefix-map.c: New file.
11093 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11094 generalize, get rid of alloca(), use strrchr() instead of strchr().
11095 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11096 Implement in terms of add_prefix_map().
11097 (remap_macro_filename, remap_debug_filename): Implement in term of
11099 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11100 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11101 * dbxout.c: Include file-prefix-map.h.
11102 * varasm.c: Likewise.
11103 * vmsdbgout.c: Likewise.
11104 * xcoffout.c: Likewise.
11105 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11106 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11107 * doc/invoke.texi (-ffile-prefix-map): Document.
11108 (-fdebug-prefix-map): Update description.
11110 2018-01-18 Martin Liska <mliska@suse.cz>
11112 * config/i386/i386.c (indirect_thunk_name): Document that also
11114 (output_indirect_thunk): Document why both instructions
11115 (pause and lfence) are generated.
11117 2018-01-18 Richard Biener <rguenther@suse.de>
11119 PR tree-optimization/83887
11120 * graphite-scop-detection.c
11121 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11122 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11123 (scop_detection::merge_sese): Re-implement with a flood-fill
11124 algorithm that properly finds a SESE region if it exists.
11126 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11129 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11130 pointer_diff optimizations use view_convert instead of convert.
11132 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11134 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11135 Generate different code for -mno-speculate-indirect-jumps.
11136 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11137 (*call_indirect_aix<mode>): Disable for
11138 -mno-speculate-indirect-jumps.
11139 (*call_indirect_aix<mode>_nospec): New define_insn.
11140 (*call_value_indirect_aix<mode>): Disable for
11141 -mno-speculate-indirect-jumps.
11142 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11143 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11144 -mno-speculate-indirect-jumps.
11145 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11147 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11149 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11150 long double type, set the flags for noting the default long double
11151 type, even if we don't pass or return a long double type.
11153 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11156 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11159 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11161 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11162 support for merge[hl].
11163 (fold_mergehl_helper): New helper function.
11164 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11165 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11166 (altivec_vmrglw_direct): Add xxmrglw insn.
11168 2018-01-17 Andrew Waterman <andrew@sifive.com>
11170 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11171 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11173 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11176 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11177 call the lto_location_cache before reading the
11178 DECL_SOURCE_LOCATION of the types.
11180 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11181 Richard Sandiford <richard.sandiford@linaro.org>
11183 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11184 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11185 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11186 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11187 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11189 * config/aarch64/constraints.md (aarch64_movti_operand):
11191 * config/aarch64/predicates.md (Uti): Add new constraint.
11193 2018-01-17 Carl Love <cel@us.ibm.com>
11195 * config/rs6000/vsx.md (define_expand xl_len_r,
11196 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11197 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11199 (define_expand, define_insn): Move the shift left from the
11200 define_insn to the define_expand for lxvl and stxvl instructions.
11201 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11202 and XL_LEN_R definitions to PURE.
11204 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11206 * config/i386/i386.c (indirect_thunk_name): Declare regno
11207 as unsigned int. Compare regno with INVALID_REGNUM.
11208 (output_indirect_thunk): Ditto.
11209 (output_indirect_thunk_function): Ditto.
11210 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11211 in the call to output_indirect_thunk_function.
11213 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11215 PR middle-end/83884
11216 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11217 rather than the size of inner_type to determine the stack slot size
11218 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11220 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11223 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11226 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11228 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11229 endian Linux systems to optionally enable multilibs for selecting
11230 the long double type if the user configured an explicit type.
11231 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11232 have no long double multilibs if not defined.
11233 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11234 warn if the user used -mabi={ieee,ibm}longdouble and we built
11235 multilibs for long double.
11236 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11237 appropriate multilib option.
11238 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11240 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11241 for building long double multilibs.
11242 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11244 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11246 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11249 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11251 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11254 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11257 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11260 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11262 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11263 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11265 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11267 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11268 different rtl trees depending on TARGET_64BIT.
11269 (rs6000_gen_lvx): Likewise.
11271 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11273 * config/visium/visium.md (nop): Tweak comment.
11274 (hazard_nop): Likewise.
11276 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11278 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11279 -mspeculate-indirect-jumps.
11280 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11281 for -mno-speculate-indirect-jumps.
11282 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11283 (*call_value_indirect_elfv2<mode>): Disable for
11284 -mno-speculate-indirect-jumps.
11285 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11286 (indirect_jump): Emit different RTL for
11287 -mno-speculate-indirect-jumps.
11288 (*indirect_jump<mode>): Disable for
11289 -mno-speculate-indirect-jumps.
11290 (*indirect_jump<mode>_nospec): New define_insn.
11291 (tablejump): Emit different RTL for
11292 -mno-speculate-indirect-jumps.
11293 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11294 (tablejumpsi_nospec): New define_expand.
11295 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11296 (tablejumpdi_nospec): New define_expand.
11297 (*tablejump<mode>_internal1): Disable for
11298 -mno-speculate-indirect-jumps.
11299 (*tablejump<mode>_internal1_nospec): New define_insn.
11300 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11303 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11305 * caller-save.c (insert_save): Drop unnecessary parameter. All
11308 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11309 Richard Biener <rguenth@suse.de>
11312 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11313 return early, inline manually is_gimple_sizepos. Make sure if we
11314 call gimplify_expr we don't end up with a gimple constant.
11315 * tree.c (variably_modified_type_p): Don't return true for
11316 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11317 * gimplify.h (is_gimple_sizepos): Remove.
11319 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11321 PR tree-optimization/83857
11322 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11323 vectorizable_live_operation for pure SLP statements.
11324 (vectorizable_live_operation): Handle PHIs.
11326 2018-01-16 Richard Biener <rguenther@suse.de>
11328 PR tree-optimization/83867
11329 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11330 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11332 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11335 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11336 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11337 If off is not INTEGER_CST, issue a may not be aligned warning
11338 rather than isn't aligned. Use isn%'t rather than isn't.
11339 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11341 <case MULT_EXPR>: Improve the case when bottom and one of the
11342 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11343 operand, in that case check if the other operand is multiple of
11344 bottom divided by the INTEGER_CST operand.
11346 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11349 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11350 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11351 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11352 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11353 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11354 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11355 (pa_function_arg_size): New function.
11357 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11359 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11360 in a separate statement.
11362 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11364 PR tree-optimization/83847
11365 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11366 group gathers and scatters.
11368 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11370 PR rtl-optimization/86620
11371 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11373 PR rtl-optimization/83213
11374 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11375 to last if both are JUMP_INSNs.
11377 PR tree-optimization/83843
11378 * gimple-ssa-store-merging.c
11379 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11380 store_immediate_info for bswap/nop orig_stores.
11382 2018-01-15 Andrew Waterman <andrew@sifive.com>
11384 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11386 <UDIV>: Increase cost if !TARGET_DIV.
11388 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11390 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11391 (define_attr "cr_logical_3op"): New.
11392 (cceq_ior_compare): Adjust.
11393 (cceq_ior_compare_complement): Adjust.
11394 (*cceq_rev_compare): Adjust.
11395 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11396 (is_cracked_insn): Adjust.
11397 (insn_must_be_first_in_group): Adjust.
11398 * config/rs6000/40x.md: Adjust.
11399 * config/rs6000/440.md: Adjust.
11400 * config/rs6000/476.md: Adjust.
11401 * config/rs6000/601.md: Adjust.
11402 * config/rs6000/603.md: Adjust.
11403 * config/rs6000/6xx.md: Adjust.
11404 * config/rs6000/7450.md: Adjust.
11405 * config/rs6000/7xx.md: Adjust.
11406 * config/rs6000/8540.md: Adjust.
11407 * config/rs6000/cell.md: Adjust.
11408 * config/rs6000/e300c2c3.md: Adjust.
11409 * config/rs6000/e500mc.md: Adjust.
11410 * config/rs6000/e500mc64.md: Adjust.
11411 * config/rs6000/e5500.md: Adjust.
11412 * config/rs6000/e6500.md: Adjust.
11413 * config/rs6000/mpc.md: Adjust.
11414 * config/rs6000/power4.md: Adjust.
11415 * config/rs6000/power5.md: Adjust.
11416 * config/rs6000/power6.md: Adjust.
11417 * config/rs6000/power7.md: Adjust.
11418 * config/rs6000/power8.md: Adjust.
11419 * config/rs6000/power9.md: Adjust.
11420 * config/rs6000/rs64.md: Adjust.
11421 * config/rs6000/titan.md: Adjust.
11423 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11425 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11426 ix86_indirect_branch_register logic.
11428 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11430 * config/i386/constraints.md (Bs): Update
11431 ix86_indirect_branch_register check. Don't check
11432 ix86_indirect_branch_register with GOT_memory_operand.
11434 * config/i386/predicates.md (GOT_memory_operand): Don't check
11435 ix86_indirect_branch_register here.
11436 (GOT32_symbol_operand): Likewise.
11438 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11440 * config/i386/predicates.md (constant_call_address_operand):
11441 Rewrite ix86_indirect_branch_register logic.
11442 (sibcall_insn_operand): Likewise.
11444 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11446 * config/i386/constraints.md (Bs): Replace
11447 ix86_indirect_branch_thunk_register with
11448 ix86_indirect_branch_register.
11450 * config/i386/i386.md (indirect_jump): Likewise.
11451 (tablejump): Likewise.
11452 (*sibcall_memory): Likewise.
11453 (*sibcall_value_memory): Likewise.
11454 Peepholes of indirect call and jump via memory: Likewise.
11455 * config/i386/i386.opt: Likewise.
11456 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11457 (GOT_memory_operand): Likewise.
11458 (call_insn_operand): Likewise.
11459 (sibcall_insn_operand): Likewise.
11460 (GOT32_symbol_operand): Likewise.
11462 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11464 PR middle-end/83837
11465 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11466 type rather than type addr's type points to.
11467 (expand_omp_atomic_mutex): Likewise.
11468 (expand_omp_atomic): Likewise.
11470 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11473 * config/i386/i386.c (output_indirect_thunk_function): Use
11474 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11475 for __x86_return_thunk.
11477 2018-01-15 Richard Biener <rguenther@suse.de>
11479 PR middle-end/83850
11480 * expmed.c (extract_bit_field_1): Fix typo.
11482 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11485 * config/arm/iterators.md (VF): New mode iterator.
11486 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11487 Remove integer-related logic from pattern.
11488 (neon_vabd<mode>_3): Likewise.
11490 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11492 PR middle-end/82694
11493 * common.opt (fstrict-overflow): No longer an alias.
11494 (fwrapv-pointer): New option.
11495 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11496 also for pointer types based on flag_wrapv_pointer.
11497 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11498 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11499 opts->x_flag_wrapv got set.
11500 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11501 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11502 POINTER_TYPE_OVERFLOW_UNDEFINED.
11503 * match.pd: Likewise in address comparison pattern.
11504 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11506 2018-01-15 Richard Biener <rguenther@suse.de>
11509 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11510 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11511 Reset type names to their identifier if their TYPE_DECL doesn't
11512 have linkage (and thus is used for ODR and devirt).
11513 (save_debug_info_for_decl): Remove.
11514 (save_debug_info_for_type): Likewise.
11515 (add_tree_to_fld_list): Adjust.
11516 * tree-pretty-print.c (dump_generic_node): Make dumping of
11517 type names more robust.
11519 2018-01-15 Richard Biener <rguenther@suse.de>
11521 * BASE-VER: Bump to 8.0.1.
11523 2018-01-14 Martin Sebor <msebor@redhat.com>
11526 * builtins.c (check_access): Avoid warning when the no-warning bit
11529 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11531 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11532 * ira-color (allocno_hard_regs_compare): Likewise.
11534 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11537 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11538 Use .pushsection/.popsection.
11540 2018-01-14 Martin Sebor <msebor@redhat.com>
11543 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11545 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11547 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11548 entry from extra_headers.
11549 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11550 extra_headers, make the list bitwise identical to the i?86-*-* one.
11552 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11554 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11555 -mcmodel=large with -mindirect-branch=thunk,
11556 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11557 -mfunction-return=thunk-extern.
11558 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11559 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11560 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11562 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11564 * config/i386/i386.c (print_reg): Print the name of the full
11565 integer register without '%'.
11566 (ix86_print_operand): Handle 'V'.
11567 * doc/extend.texi: Document 'V' modifier.
11569 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11571 * config/i386/constraints.md (Bs): Disallow memory operand for
11572 -mindirect-branch-register.
11574 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11575 (GOT_memory_operand): Likewise.
11576 (call_insn_operand): Likewise.
11577 (sibcall_insn_operand): Likewise.
11578 (GOT32_symbol_operand): Likewise.
11579 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11580 for -mindirect-branch-register.
11581 (tablejump): Likewise.
11582 (*sibcall_memory): Likewise.
11583 (*sibcall_value_memory): Likewise.
11584 Disallow peepholes of indirect call and jump via memory for
11585 -mindirect-branch-register.
11586 (*call_pop): Replace m with Bw.
11587 (*call_value_pop): Likewise.
11588 (*sibcall_pop_memory): Replace m with Bs.
11589 * config/i386/i386.opt (mindirect-branch-register): New option.
11590 * doc/invoke.texi: Document -mindirect-branch-register option.
11592 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11594 * config/i386/i386-protos.h (ix86_output_function_return): New.
11595 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11596 set function_return_type.
11597 (indirect_thunk_name): Add ret_p to indicate thunk for function
11599 (output_indirect_thunk_function): Pass false to
11600 indirect_thunk_name.
11601 (ix86_output_indirect_branch_via_reg): Likewise.
11602 (ix86_output_indirect_branch_via_push): Likewise.
11603 (output_indirect_thunk_function): Create alias for function
11604 return thunk if regno < 0.
11605 (ix86_output_function_return): New function.
11606 (ix86_handle_fndecl_attribute): Handle function_return.
11607 (ix86_attribute_table): Add function_return.
11608 * config/i386/i386.h (machine_function): Add
11609 function_return_type.
11610 * config/i386/i386.md (simple_return_internal): Use
11611 ix86_output_function_return.
11612 (simple_return_internal_long): Likewise.
11613 * config/i386/i386.opt (mfunction-return=): New option.
11614 (indirect_branch): Mention -mfunction-return=.
11615 * doc/extend.texi: Document function_return function attribute.
11616 * doc/invoke.texi: Document -mfunction-return= option.
11618 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11620 * config/i386/i386-opts.h (indirect_branch): New.
11621 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11622 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11623 with local indirect jump when converting indirect call and jump.
11624 (ix86_set_indirect_branch_type): New.
11625 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11626 (indirectlabelno): New.
11627 (indirect_thunk_needed): Likewise.
11628 (indirect_thunk_bnd_needed): Likewise.
11629 (indirect_thunks_used): Likewise.
11630 (indirect_thunks_bnd_used): Likewise.
11631 (INDIRECT_LABEL): Likewise.
11632 (indirect_thunk_name): Likewise.
11633 (output_indirect_thunk): Likewise.
11634 (output_indirect_thunk_function): Likewise.
11635 (ix86_output_indirect_branch_via_reg): Likewise.
11636 (ix86_output_indirect_branch_via_push): Likewise.
11637 (ix86_output_indirect_branch): Likewise.
11638 (ix86_output_indirect_jmp): Likewise.
11639 (ix86_code_end): Call output_indirect_thunk_function if needed.
11640 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11642 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11643 (ix86_attribute_table): Add indirect_branch.
11644 * config/i386/i386.h (machine_function): Add indirect_branch_type
11645 and has_local_indirect_jump.
11646 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11648 (tablejump): Likewise.
11649 (*indirect_jump): Use ix86_output_indirect_jmp.
11650 (*tablejump_1): Likewise.
11651 (simple_return_indirect_internal): Likewise.
11652 * config/i386/i386.opt (mindirect-branch=): New option.
11653 (indirect_branch): New.
11656 (thunk-inline): Likewise.
11657 (thunk-extern): Likewise.
11658 * doc/extend.texi: Document indirect_branch function attribute.
11659 * doc/invoke.texi: Document -mindirect-branch= option.
11661 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11664 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11666 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11668 * ipa-inline.c (want_inline_small_function_p): Return false if
11669 inlining has already failed with CIF_FINAL_ERROR.
11670 (update_caller_keys): Call want_inline_small_function_p before
11672 (update_callee_keys): Likewise.
11674 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11676 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11678 (rs6000_quadword_masked_address_p): Likewise.
11679 (quad_aligned_load_p): Likewise.
11680 (quad_aligned_store_p): Likewise.
11681 (const_load_sequence_p): Add comment to describe the outer-most loop.
11682 (mimic_memory_attributes_and_flags): New function.
11683 (rs6000_gen_stvx): Likewise.
11684 (replace_swapped_aligned_store): Likewise.
11685 (rs6000_gen_lvx): Likewise.
11686 (replace_swapped_aligned_load): Likewise.
11687 (replace_swapped_load_constant): Capitalize argument name in
11688 comment describing this function.
11689 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11690 and stores that access quad-word aligned addresses and replace
11691 with stvx or lvx instructions when appropriate.
11692 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11693 New function prototype.
11694 (rs6000_quadword_masked_address_p): Likewise.
11695 (rs6000_gen_lvx): Likewise.
11696 (rs6000_gen_stvx): Likewise.
11697 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11698 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11699 when memory address is aligned.
11700 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11701 this split to select lvx instruction when memory address is aligned.
11702 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11703 instruction when memory address is aligned.
11704 (*vsx_le_perm_load_v16qi): Likewise.
11705 (four unnamed splitters): Modify to select the stvx instruction
11706 when memory is aligned.
11708 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11710 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11711 which appears in the queue multiple times.
11713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11714 Alan Hayward <alan.hayward@arm.com>
11715 David Sherwood <david.sherwood@arm.com>
11717 * tree-vectorizer.h (vec_lower_bound): New structure.
11718 (_loop_vec_info): Add check_nonzero and lower_bounds.
11719 (LOOP_VINFO_CHECK_NONZERO): New macro.
11720 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11721 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11722 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11723 fields. Make seg_len the distance travelled, not including the
11725 (dr_direction_indicator): Declare.
11726 (dr_zero_step_indicator): Likewise.
11727 (dr_known_forward_stride_p): Likewise.
11728 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11730 (runtime_alias_check_p): Allow runtime alias checks with
11732 (operator ==): Compare access_size and align.
11733 (prune_runtime_alias_test_list): Rework for new distinction between
11734 the access_size and seg_len.
11735 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11737 (get_segment_min_max): New function.
11738 (create_intersect_range_checks): Use it.
11739 (dr_step_indicator): New function.
11740 (dr_direction_indicator): Likewise.
11741 (dr_zero_step_indicator): Likewise.
11742 (dr_known_forward_stride_p): Likewise.
11743 * tree-loop-distribution.c (data_ref_segment_size): Return
11744 DR_STEP * (niters - 1).
11745 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11747 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11748 (vect_preserves_scalar_order_p): New function, split out from...
11749 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11750 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11751 (vect_vfa_access_size): New function.
11752 (vect_vfa_align): Likewise.
11753 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11754 (dump_lower_bound): New function.
11755 (vect_check_lower_bound): Likewise.
11756 (vect_small_gap_p): Likewise.
11757 (vectorizable_with_step_bound_p): Likewise.
11758 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11759 depencies if the vectorization factor is 1. Convert the checks
11760 for nonzero steps into checks on the bounds of DR_STEP. Try using
11761 a bunds check for variable steps if the minimum required step is
11762 relatively small. Update calls to the dr_with_seg_len
11763 constructor and to vect_compile_time_alias.
11764 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11766 (vect_loop_versioning): Call it.
11767 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11769 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11772 Alan Hayward <alan.hayward@arm.com>
11773 David Sherwood <david.sherwood@arm.com>
11775 * doc/sourcebuild.texi (vect_scatter_store): Document.
11776 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11778 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11780 * genopinit.c (main): Add supports_vec_scatter_store and
11781 supports_vec_scatter_store_cached to target_optabs.
11782 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11783 IFN_MASK_SCATTER_STORE.
11784 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11786 * internal-fn.h (internal_store_fn_p): Declare.
11787 (internal_fn_stored_value_index): Likewise.
11788 * internal-fn.c (scatter_store_direct): New macro.
11789 (expand_scatter_store_optab_fn): New function.
11790 (direct_scatter_store_optab_supported_p): New macro.
11791 (internal_store_fn_p): New function.
11792 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11793 IFN_MASK_SCATTER_STORE.
11794 (internal_fn_mask_index): Likewise.
11795 (internal_fn_stored_value_index): New function.
11796 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11797 for scatter stores.
11798 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11799 * optabs-query.c (supports_vec_scatter_store_p): New function.
11800 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11801 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11802 true for scatter stores.
11803 (vect_gather_scatter_fn_p): Handle scatter stores too.
11804 (vect_check_gather_scatter): Consider using scatter stores if
11805 supports_vec_scatter_store_p.
11806 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11807 scatter stores too.
11808 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11809 internal_fn_stored_value_index.
11810 (check_load_store_masking): Handle scatter stores too.
11811 (vect_get_store_rhs): Make public.
11812 (vectorizable_call): Use internal_store_fn_p.
11813 (vectorizable_store): Handle scatter store internal functions.
11814 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11815 when deciding whether the end of the group has been reached.
11816 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11817 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11818 (mask_scatter_store<mode>): New insns.
11820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11821 Alan Hayward <alan.hayward@arm.com>
11822 David Sherwood <david.sherwood@arm.com>
11824 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11825 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11826 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11828 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11829 Use vect_truncate_gather_scatter_offset if we can't treat the
11830 operation as a normal gather load or scatter store.
11831 (get_group_load_store_type): Take the gather_scatter_info
11832 as argument. Try using a gather load or scatter store for
11833 single-element groups.
11834 (get_load_store_type): Update calls to get_group_load_store_type
11835 and vect_use_strided_gather_scatters_p.
11837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11838 Alan Hayward <alan.hayward@arm.com>
11839 David Sherwood <david.sherwood@arm.com>
11841 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11842 optional tree argument.
11843 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11845 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11846 but continue to use the current value as a fallback.
11847 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11848 to compare the updates.
11849 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11850 (get_load_store_type): Use it when handling a strided access.
11851 (vect_get_strided_load_store_ops): New function.
11852 (vect_get_data_ptr_increment): Likewise.
11853 (vectorizable_load): Handle strided gather loads. Always pass
11854 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11857 Alan Hayward <alan.hayward@arm.com>
11858 David Sherwood <david.sherwood@arm.com>
11860 * doc/md.texi (gather_load@var{m}): Document.
11861 (mask_gather_load@var{m}): Likewise.
11862 * genopinit.c (main): Add supports_vec_gather_load and
11863 supports_vec_gather_load_cached to target_optabs.
11864 * optabs-tree.c (init_tree_optimization_optabs): Use
11865 ggc_cleared_alloc to allocate target_optabs.
11866 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11867 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11869 * internal-fn.h (internal_load_fn_p): Declare.
11870 (internal_gather_scatter_fn_p): Likewise.
11871 (internal_fn_mask_index): Likewise.
11872 (internal_gather_scatter_fn_supported_p): Likewise.
11873 * internal-fn.c (gather_load_direct): New macro.
11874 (expand_gather_load_optab_fn): New function.
11875 (direct_gather_load_optab_supported_p): New macro.
11876 (direct_internal_fn_optab): New function.
11877 (internal_load_fn_p): Likewise.
11878 (internal_gather_scatter_fn_p): Likewise.
11879 (internal_fn_mask_index): Likewise.
11880 (internal_gather_scatter_fn_supported_p): Likewise.
11881 * optabs-query.c (supports_at_least_one_mode_p): New function.
11882 (supports_vec_gather_load_p): Likewise.
11883 * optabs-query.h (supports_vec_gather_load_p): Declare.
11884 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11885 and memory_type field.
11886 (NUM_PATTERNS): Bump to 15.
11887 * tree-vect-data-refs.c: Include internal-fn.h.
11888 (vect_gather_scatter_fn_p): New function.
11889 (vect_describe_gather_scatter_call): Likewise.
11890 (vect_check_gather_scatter): Try using internal functions for
11891 gather loads. Recognize existing calls to a gather load function.
11892 (vect_analyze_data_refs): Consider using gather loads if
11893 supports_vec_gather_load_p.
11894 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11895 (vect_get_gather_scatter_offset_type): Likewise.
11896 (vect_convert_mask_for_vectype): Likewise.
11897 (vect_add_conversion_to_patterm): Likewise.
11898 (vect_try_gather_scatter_pattern): Likewise.
11899 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11900 (vect_vect_recog_func_ptrs): Add it.
11901 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11902 internal_fn_mask_index and internal_gather_scatter_fn_p.
11903 (check_load_store_masking): Take the gather_scatter_info as an
11904 argument and handle gather loads.
11905 (vect_get_gather_scatter_ops): New function.
11906 (vectorizable_call): Check internal_load_fn_p.
11907 (vectorizable_load): Likewise. Handle gather load internal
11909 (vectorizable_store): Update call to check_load_store_masking.
11910 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11911 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11912 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11913 (aarch64_gather_scale_operand_d): New predicates.
11914 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11915 (mask_gather_load<mode>): New insns.
11917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * optabs.def (fold_left_plus_optab): New optab.
11922 * doc/md.texi (fold_left_plus_@var{m}): Document.
11923 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11924 * internal-fn.c (fold_left_direct): Define.
11925 (expand_fold_left_optab_fn): Likewise.
11926 (direct_fold_left_optab_supported_p): Likewise.
11927 * fold-const-call.c (fold_const_fold_left): New function.
11928 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11929 * tree-parloops.c (valid_reduction_p): New function.
11930 (gather_scalar_reductions): Use it.
11931 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11932 (vect_finish_replace_stmt): Declare.
11933 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11934 (needs_fold_left_reduction_p): New function, split out from...
11935 (vect_is_simple_reduction): ...here. Accept reductions that
11936 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11937 (vect_force_simple_reduction): Also store the reduction type in
11938 the assignment's STMT_VINFO_REDUC_TYPE.
11939 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11940 (merge_with_identity): New function.
11941 (vect_expand_fold_left): Likewise.
11942 (vectorize_fold_left_reduction): Likewise.
11943 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11944 scalar phi in place for it. Check for target support and reject
11945 cases that would reassociate the operation. Defer the transform
11946 phase to vectorize_fold_left_reduction.
11947 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11948 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11949 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11953 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11954 call to ifc_temp_var.
11956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11957 Alan Hayward <alan.hayward@arm.com>
11958 David Sherwood <david.sherwood@arm.com>
11960 * target.def (legitimize_address_displacement): Take the original
11961 offset as a poly_int.
11962 * targhooks.h (default_legitimize_address_displacement): Update
11964 * targhooks.c (default_legitimize_address_displacement): Likewise.
11965 * doc/tm.texi: Regenerate.
11966 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11967 as an argument, moving assert of ad->disp == ad->disp_term to...
11968 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11969 Try calling targetm.legitimize_address_displacement before expanding
11970 the address rather than afterwards, and adjust for the new interface.
11971 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11972 Match the new hook interface. Handle SVE addresses.
11973 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11974 new hook interface.
11976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11978 * Makefile.in (OBJS): Add early-remat.o.
11979 * target.def (select_early_remat_modes): New hook.
11980 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11981 * doc/tm.texi: Regenerate.
11982 * targhooks.h (default_select_early_remat_modes): Declare.
11983 * targhooks.c (default_select_early_remat_modes): New function.
11984 * timevar.def (TV_EARLY_REMAT): New timevar.
11985 * passes.def (pass_early_remat): New pass.
11986 * tree-pass.h (make_pass_early_remat): Declare.
11987 * early-remat.c: New file.
11988 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11990 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11993 Alan Hayward <alan.hayward@arm.com>
11994 David Sherwood <david.sherwood@arm.com>
11996 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11997 vfm1 with a bound_epilog parameter.
11998 (vect_do_peeling): Update calls accordingly, and move the prologue
11999 call earlier in the function. Treat the base bound_epilog as 0 for
12000 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12001 this base when peeling for gaps.
12002 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12003 with fully-masked loops.
12004 (vect_estimate_min_profitable_iters): Handle the single peeled
12005 iteration in that case.
12007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12008 Alan Hayward <alan.hayward@arm.com>
12009 David Sherwood <david.sherwood@arm.com>
12011 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12012 single-element interleaving even if the size is not a power of 2.
12013 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12014 accesses for single-element interleaving if the group size is
12017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12018 Alan Hayward <alan.hayward@arm.com>
12019 David Sherwood <david.sherwood@arm.com>
12021 * doc/md.texi (fold_extract_last_@var{m}): Document.
12022 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12023 * optabs.def (fold_extract_last_optab): New optab.
12024 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12025 * internal-fn.c (fold_extract_direct): New macro.
12026 (expand_fold_extract_optab_fn): Likewise.
12027 (direct_fold_extract_optab_supported_p): Likewise.
12028 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12029 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12030 EXTRACT_LAST_REDUCTION.
12031 (get_initial_def_for_reduction): Do not create an initial vector
12032 for EXTRACT_LAST_REDUCTION reductions.
12033 (vectorizable_reduction): Leave the scalar phi in place for
12034 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12035 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12036 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12037 transform phase to vectorizable_condition.
12038 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12040 (vect_finish_stmt_generation): ...here.
12041 (vect_finish_replace_stmt): New function.
12042 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12043 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12045 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12048 Alan Hayward <alan.hayward@arm.com>
12049 David Sherwood <david.sherwood@arm.com>
12051 * doc/md.texi (extract_last_@var{m}): Document.
12052 * optabs.def (extract_last_optab): New optab.
12053 * internal-fn.def (EXTRACT_LAST): New internal function.
12054 * internal-fn.c (cond_unary_direct): New macro.
12055 (expand_cond_unary_optab_fn): Likewise.
12056 (direct_cond_unary_optab_supported_p): Likewise.
12057 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12058 loops using EXTRACT_LAST.
12059 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12060 (extract_last_<mode>): ...this optab.
12061 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
12068 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12069 * doc/tm.texi: Regenerate.
12070 * targhooks.h (default_empty_mask_is_expensive): Declare.
12071 * targhooks.c (default_empty_mask_is_expensive): New function.
12072 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12073 if the target says that empty masks are expensive.
12074 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12076 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12079 Alan Hayward <alan.hayward@arm.com>
12080 David Sherwood <david.sherwood@arm.com>
12082 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12083 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12084 (vect_use_loop_mask_for_alignment_p): New function.
12085 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12086 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12087 niters_skip argument. Make sure that the first niters_skip elements
12088 of the first iteration are inactive.
12089 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12090 Update call to vect_set_loop_masks_directly.
12091 (get_misalign_in_elems): New function, split out from...
12092 (vect_gen_prolog_loop_niters): ...here.
12093 (vect_update_init_of_dr): Take a code argument that specifies whether
12094 the adjustment should be added or subtracted.
12095 (vect_update_init_of_drs): Likewise.
12096 (vect_prepare_for_masked_peels): New function.
12097 (vect_do_peeling): Skip prologue peeling if we're using a mask
12098 instead. Update call to vect_update_inits_of_drs.
12099 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12101 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12102 alignment. Do not include the number of peeled iterations in
12103 the minimum threshold in that case.
12104 (vectorizable_induction): Adjust the start value down by
12105 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12106 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12107 Take the number of skipped iterations into account when calculating
12109 * tree-vect-stmts.c (vect_gen_while_not): New function.
12111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12112 Alan Hayward <alan.hayward@arm.com>
12113 David Sherwood <david.sherwood@arm.com>
12115 * doc/sourcebuild.texi (vect_fully_masked): Document.
12116 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12117 default value to 0.
12118 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12120 (vect_analyze_loop_2): ...here. Don't check the vectorization
12121 factor against the number of loop iterations if the loop is
12124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12129 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12130 (dump_groups): Update accordingly.
12131 (iv_use::mem_type): New member variable.
12132 (address_p): New function.
12133 (record_use): Add a mem_type argument and initialize the new
12135 (record_group_use): Add a mem_type argument. Use address_p.
12136 Remove obsolete null checks of base_object. Update call to record_use.
12137 (find_interesting_uses_op): Update call to record_group_use.
12138 (find_interesting_uses_cond): Likewise.
12139 (find_interesting_uses_address): Likewise.
12140 (get_mem_type_for_internal_fn): New function.
12141 (find_address_like_use): Likewise.
12142 (find_interesting_uses_stmt): Try find_address_like_use before
12143 calling find_interesting_uses_op.
12144 (addr_offset_valid_p): Use the iv mem_type field as the type
12145 of the addressed memory.
12146 (add_autoinc_candidates): Likewise.
12147 (get_address_cost): Likewise.
12148 (split_small_address_groups_p): Use address_p.
12149 (split_address_groups): Likewise.
12150 (add_iv_candidate_for_use): Likewise.
12151 (autoinc_possible_for_pair): Likewise.
12152 (rewrite_groups): Likewise.
12153 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12154 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12155 (get_alias_ptr_type_for_ptr_address): New function.
12156 (rewrite_use_address): Rewrite address uses in calls that were
12157 identified by find_address_like_use.
12159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12160 Alan Hayward <alan.hayward@arm.com>
12161 David Sherwood <david.sherwood@arm.com>
12163 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12165 * gimple-expr.h (is_gimple_addressable: Likewise.
12166 * gimple-expr.c (is_gimple_address): Likewise.
12167 * internal-fn.c (expand_call_mem_ref): New function.
12168 (expand_mask_load_optab_fn): Use it.
12169 (expand_mask_store_optab_fn): Likewise.
12171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12172 Alan Hayward <alan.hayward@arm.com>
12173 David Sherwood <david.sherwood@arm.com>
12175 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12176 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12177 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12178 (cond_umax@var{mode}): Document.
12179 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12180 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12181 (cond_umin_optab, cond_umax_optab): New optabs.
12182 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12183 (COND_IOR, COND_XOR): New internal functions.
12184 * internal-fn.h (get_conditional_internal_fn): Declare.
12185 * internal-fn.c (cond_binary_direct): New macro.
12186 (expand_cond_binary_optab_fn): Likewise.
12187 (direct_cond_binary_optab_supported_p): Likewise.
12188 (get_conditional_internal_fn): New function.
12189 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12190 Cope with reduction statements that are vectorized as calls rather
12192 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12193 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12194 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12195 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12196 (UNSPEC_COND_EOR): New unspecs.
12197 (optab): Add mappings for them.
12198 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12199 (sve_int_op, sve_fp_op): New int attributes.
12201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12202 Alan Hayward <alan.hayward@arm.com>
12203 David Sherwood <david.sherwood@arm.com>
12205 * optabs.def (while_ult_optab): New optab.
12206 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12207 * internal-fn.def (WHILE_ULT): New internal function.
12208 * internal-fn.h (direct_internal_fn_supported_p): New override
12209 that takes two types as argument.
12210 * internal-fn.c (while_direct): New macro.
12211 (expand_while_optab_fn): New function.
12212 (convert_optab_supported_p): Likewise.
12213 (direct_while_optab_supported_p): New macro.
12214 * wide-int.h (wi::udiv_ceil): New function.
12215 * tree-vectorizer.h (rgroup_masks): New structure.
12216 (vec_loop_masks): New typedef.
12217 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12218 and fully_masked_p.
12219 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12220 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12221 (vect_max_vf): New function.
12222 (slpeel_make_loop_iterate_ntimes): Delete.
12223 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12224 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12225 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12226 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12227 internal-fn.h, stor-layout.h and optabs-query.h.
12228 (vect_set_loop_mask): New function.
12229 (add_preheader_seq): Likewise.
12230 (add_header_seq): Likewise.
12231 (interleave_supported_p): Likewise.
12232 (vect_maybe_permute_loop_masks): Likewise.
12233 (vect_set_loop_masks_directly): Likewise.
12234 (vect_set_loop_condition_masked): Likewise.
12235 (vect_set_loop_condition_unmasked): New function, split out from
12236 slpeel_make_loop_iterate_ntimes.
12237 (slpeel_make_loop_iterate_ntimes): Rename to..
12238 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12239 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12240 (vect_do_peeling): Update call accordingly.
12241 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12243 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12244 mask_compare_type, can_fully_mask_p and fully_masked_p.
12245 (release_vec_loop_masks): New function.
12246 (_loop_vec_info): Use it to free the loop masks.
12247 (can_produce_all_loop_masks_p): New function.
12248 (vect_get_max_nscalars_per_iter): Likewise.
12249 (vect_verify_full_masking): Likewise.
12250 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12251 retries, and free the mask rgroups before retrying. Check loop-wide
12252 reasons for disallowing fully-masked loops. Make the final decision
12253 about whether use a fully-masked loop or not.
12254 (vect_estimate_min_profitable_iters): Do not assume that peeling
12255 for the number of iterations will be needed for fully-masked loops.
12256 (vectorizable_reduction): Disable fully-masked loops.
12257 (vectorizable_live_operation): Likewise.
12258 (vect_halve_mask_nunits): New function.
12259 (vect_double_mask_nunits): Likewise.
12260 (vect_record_loop_mask): Likewise.
12261 (vect_get_loop_mask): Likewise.
12262 (vect_transform_loop): Handle the case in which the final loop
12263 iteration might handle a partial vector. Call vect_set_loop_condition
12264 instead of slpeel_make_loop_iterate_ntimes.
12265 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12266 (check_load_store_masking): New function.
12267 (prepare_load_store_mask): Likewise.
12268 (vectorizable_store): Handle fully-masked loops.
12269 (vectorizable_load): Likewise.
12270 (supportable_widening_operation): Use vect_halve_mask_nunits for
12272 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12273 (vect_gen_while): New function.
12274 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12275 (aarch64_uqdec<mode>): New insn.
12277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12278 Alan Hayward <alan.hayward@arm.com>
12279 David Sherwood <david.sherwood@arm.com>
12281 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12282 (reduc_xor_scal_optab): New optabs.
12283 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12284 (reduc_xor_scal_@var{m}): Document.
12285 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12286 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12287 internal functions.
12288 * fold-const-call.c (fold_const_call): Handle them.
12289 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12290 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12291 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12292 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12293 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12294 (UNSPEC_XORV): New unspecs.
12295 (optab): Add entries for them.
12296 (BITWISEV): New int iterator.
12297 (bit_reduc_op): New int attributes.
12299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12300 Alan Hayward <alan.hayward@arm.com>
12301 David Sherwood <david.sherwood@arm.com>
12303 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12304 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12305 * optabs.def (vec_shl_insert_optab): New optab.
12306 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12307 (duplicate_and_interleave): Likewise.
12308 * tree-vect-loop.c: Include internal-fn.h.
12309 (neutral_op_for_slp_reduction): New function, split out from
12310 get_initial_defs_for_reduction.
12311 (get_initial_def_for_reduction): Handle option 2 for variable-length
12312 vectors by loading the neutral value into a vector and then shifting
12313 the initial value into element 0.
12314 (get_initial_defs_for_reduction): Replace the code argument with
12315 the neutral value calculated by neutral_op_for_slp_reduction.
12316 Use gimple_build_vector for constant-length vectors.
12317 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12318 but the first group_size elements have a neutral value.
12319 Use duplicate_and_interleave otherwise.
12320 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12321 Update call to get_initial_defs_for_reduction. Handle SLP
12322 reductions for variable-length vectors by creating one vector
12323 result for each scalar result, with the elements associated
12324 with other scalar results stubbed out with the neutral value.
12325 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12326 Require IFN_VEC_SHL_INSERT for double reductions on
12327 variable-length vectors, or SLP reductions that have
12328 a neutral value. Require can_duplicate_and_interleave_p
12329 support for variable-length unchained SLP reductions if there
12330 is no neutral value, such as for MIN/MAX reductions. Also require
12331 the number of vector elements to be a multiple of the number of
12332 SLP statements when doing variable-length unchained SLP reductions.
12333 Update call to vect_create_epilog_for_reduction.
12334 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12335 and remove initial values.
12336 (duplicate_and_interleave): Make public.
12337 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12338 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12341 Alan Hayward <alan.hayward@arm.com>
12342 David Sherwood <david.sherwood@arm.com>
12344 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12345 (can_duplicate_and_interleave_p): New function.
12346 (vect_get_and_check_slp_defs): Take the vector of statements
12347 rather than just the current one. Remove excess parentheses.
12348 Restriction rejectinon of vect_constant_def and vect_external_def
12349 for variable-length vectors to boolean types, or types for which
12350 can_duplicate_and_interleave_p is false.
12351 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12352 (duplicate_and_interleave): New function.
12353 (vect_get_constant_vectors): Use gimple_build_vector for
12354 constant-length vectors and suitable variable-length constant
12355 vectors. Use duplicate_and_interleave for other variable-length
12356 vectors. Don't defer the update when inserting new statements.
12358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12359 Alan Hayward <alan.hayward@arm.com>
12360 David Sherwood <david.sherwood@arm.com>
12362 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12363 min_profitable_iters doesn't go negative.
12365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12366 Alan Hayward <alan.hayward@arm.com>
12367 David Sherwood <david.sherwood@arm.com>
12369 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12370 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12371 * optabs.def (vec_mask_load_lanes_optab): New optab.
12372 (vec_mask_store_lanes_optab): Likewise.
12373 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12374 (MASK_STORE_LANES): Likewise.
12375 * internal-fn.c (mask_load_lanes_direct): New macro.
12376 (mask_store_lanes_direct): Likewise.
12377 (expand_mask_load_optab_fn): Handle masked operations.
12378 (expand_mask_load_lanes_optab_fn): New macro.
12379 (expand_mask_store_optab_fn): Handle masked operations.
12380 (expand_mask_store_lanes_optab_fn): New macro.
12381 (direct_mask_load_lanes_optab_supported_p): Likewise.
12382 (direct_mask_store_lanes_optab_supported_p): Likewise.
12383 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12385 (vect_load_lanes_supported): Likewise.
12386 * tree-vect-data-refs.c (strip_conversion): New function.
12387 (can_group_stmts_p): Likewise.
12388 (vect_analyze_data_ref_accesses): Use it instead of checking
12389 for a pair of assignments.
12390 (vect_store_lanes_supported): Take a masked_p parameter.
12391 (vect_load_lanes_supported): Likewise.
12392 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12393 vect_store_lanes_supported and vect_load_lanes_supported.
12394 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12395 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12396 parameter. Don't allow gaps for masked accesses.
12397 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12398 and vect_load_lanes_supported.
12399 (get_load_store_type): Take a masked_p parameter and update
12400 call to get_group_load_store_type.
12401 (vectorizable_store): Update call to get_load_store_type.
12402 Handle IFN_MASK_STORE_LANES.
12403 (vectorizable_load): Update call to get_load_store_type.
12404 Handle IFN_MASK_LOAD_LANES.
12406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12407 Alan Hayward <alan.hayward@arm.com>
12408 David Sherwood <david.sherwood@arm.com>
12410 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12412 * config/aarch64/aarch64-protos.h
12413 (aarch64_sve_struct_memory_operand_p): Declare.
12414 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12415 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12416 (VPRED, vpred): Handle SVE structure modes.
12417 * config/aarch64/constraints.md (Utx): New constraint.
12418 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12419 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12420 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12421 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12422 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12423 structure modes. Split into pieces after RA.
12424 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12425 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12427 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12428 SVE structure modes.
12429 (aarch64_classify_address): Likewise.
12430 (sizetochar): Move earlier in file.
12431 (aarch64_print_operand): Handle SVE register lists.
12432 (aarch64_array_mode): New function.
12433 (aarch64_sve_struct_memory_operand_p): Likewise.
12434 (TARGET_ARRAY_MODE): Redefine.
12436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12437 Alan Hayward <alan.hayward@arm.com>
12438 David Sherwood <david.sherwood@arm.com>
12440 * target.def (array_mode): New target hook.
12441 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12442 * doc/tm.texi: Regenerate.
12443 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12444 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12445 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12446 targetm.array_mode.
12447 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12451 Alan Hayward <alan.hayward@arm.com>
12452 David Sherwood <david.sherwood@arm.com>
12454 * fold-const.c (fold_binary_loc): Check the argument types
12455 rather than the result type when testing for a vector operation.
12457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12459 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12460 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12467 (sve): Document new AArch64 extension.
12468 * doc/md.texi (w): Extend the description of the AArch64
12469 constraint to include SVE vectors.
12470 (Upl, Upa): Document new AArch64 predicate constraints.
12471 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12473 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12474 (msve-vector-bits=): New option.
12475 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12476 SVE when these are disabled.
12477 (sve): New extension.
12478 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12479 modes. Adjust their number of units based on aarch64_sve_vg.
12480 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12481 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12482 aarch64_addr_query_type.
12483 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12484 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12485 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12486 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12487 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12488 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12489 (aarch64_simd_imm_zero_p): Delete.
12490 (aarch64_check_zero_based_sve_index_immediate): Declare.
12491 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12492 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12493 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12494 (aarch64_sve_float_mul_immediate_p): Likewise.
12495 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12496 rather than an rtx.
12497 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12498 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12499 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12500 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12501 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12502 (aarch64_regmode_natural_size): Likewise.
12503 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12504 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12506 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12507 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12508 for VG and the SVE predicate registers.
12509 (V_ALIASES): Add a "z"-prefixed alias.
12510 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12511 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12512 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12513 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12514 (REG_CLASS_NAMES): Add entries for them.
12515 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12516 and the predicate registers.
12517 (aarch64_sve_vg): Declare.
12518 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12519 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12520 (REGMODE_NATURAL_SIZE): Define.
12521 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12523 * config/aarch64/aarch64.c: Include cfgrtl.h.
12524 (simd_immediate_info): Add a constructor for series vectors,
12525 and an associated step field.
12526 (aarch64_sve_vg): New variable.
12527 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12528 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12529 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12530 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12531 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12532 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12533 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12534 (aarch64_get_mask_mode): New functions.
12535 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12536 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12537 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12538 predicate modes and predicate registers. Explicitly restrict
12539 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12540 to store a vector mode if it is recognized by
12541 aarch64_classify_vector_mode.
12542 (aarch64_regmode_natural_size): New function.
12543 (aarch64_hard_regno_caller_save_mode): Return the original mode
12545 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12546 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12547 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12548 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12550 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12551 does not overlap dest if the function is frame-related. Handle
12553 (aarch64_split_add_offset): New function.
12554 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12555 them aarch64_add_offset.
12556 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12557 and update call to aarch64_sub_sp.
12558 (aarch64_add_cfa_expression): New function.
12559 (aarch64_expand_prologue): Pass extra temporary registers to the
12560 functions above. Handle the case in which we need to emit new
12561 DW_CFA_expressions for registers that were originally saved
12562 relative to the stack pointer, but now have to be expressed
12563 relative to the frame pointer.
12564 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12566 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12567 IP0 and IP1 values for SVE frames.
12568 (aarch64_expand_vec_series): New function.
12569 (aarch64_expand_sve_widened_duplicate): Likewise.
12570 (aarch64_expand_sve_const_vector): Likewise.
12571 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12572 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12573 into the register, rather than emitting a SET directly.
12574 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12575 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12576 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12577 (offset_9bit_signed_scaled_p): New functions.
12578 (aarch64_replicate_bitmask_imm): New function.
12579 (aarch64_bitmask_imm): Use it.
12580 (aarch64_cannot_force_const_mem): Reject expressions involving
12581 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12582 (aarch64_classify_index): Handle SVE indices, by requiring
12583 a plain register index with a scale that matches the element size.
12584 (aarch64_classify_address): Handle SVE addresses. Assert that
12585 the mode of the address is VOIDmode or an integer mode.
12586 Update call to aarch64_classify_symbol.
12587 (aarch64_classify_symbolic_expression): Update call to
12588 aarch64_classify_symbol.
12589 (aarch64_const_vec_all_in_range_p): New function.
12590 (aarch64_print_vector_float_operand): Likewise.
12591 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12592 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12593 and the FP immediates 1.0 and 0.5.
12594 (aarch64_print_address_internal): Handle SVE addresses.
12595 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12596 (aarch64_regno_regclass): Handle predicate registers.
12597 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12599 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12600 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12601 (aarch64_convert_sve_vector_bits): New function.
12602 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12603 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12604 rather than an rtx.
12605 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12606 Handle SVE vector and predicate modes. Accept VL-based constants
12607 that need only one temporary register, and VL offsets that require
12608 no temporary registers.
12609 (aarch64_conditional_register_usage): Mark the predicate registers
12610 as fixed if SVE isn't available.
12611 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12612 Return true for SVE vector and predicate modes.
12613 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12614 rather than an unsigned int. Handle SVE modes.
12615 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12617 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12619 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12620 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12621 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12622 (aarch64_sve_float_mul_immediate_p): New functions.
12623 (aarch64_sve_valid_immediate): New function.
12624 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12625 Explicitly reject structure modes. Check for INDEX constants.
12626 Handle PTRUE and PFALSE constants.
12627 (aarch64_check_zero_based_sve_index_immediate): New function.
12628 (aarch64_simd_imm_zero_p): Delete.
12629 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12630 vector modes. Accept constants in the range of CNT[BHWD].
12631 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12632 ask for an Advanced SIMD mode.
12633 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12634 (aarch64_simd_vector_alignment): Handle SVE predicates.
12635 (aarch64_vectorize_preferred_vector_alignment): New function.
12636 (aarch64_simd_vector_alignment_reachable): Use it instead of
12638 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12639 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12641 (MAX_VECT_LEN): Delete.
12642 (expand_vec_perm_d): Add a vec_flags field.
12643 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12644 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12645 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12647 (aarch64_evpc_rev): Rename to...
12648 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12649 (aarch64_evpc_rev_global): New function.
12650 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12651 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12653 (aarch64_evpc_sve_tbl): New function.
12654 (aarch64_expand_vec_perm_const_1): Update after rename of
12655 aarch64_evpc_rev. Handle SVE permutes too, trying
12656 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12657 than aarch64_evpc_tbl.
12658 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12659 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12660 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12661 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12662 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12663 (aarch64_expand_sve_vcond): New functions.
12664 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12665 of aarch64_vector_mode_p.
12666 (aarch64_dwarf_poly_indeterminate_value): New function.
12667 (aarch64_compute_pressure_classes): Likewise.
12668 (aarch64_can_change_mode_class): Likewise.
12669 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12670 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12671 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12672 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12673 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12674 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12675 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12676 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12678 (Dn, Dl, Dr): Accept const as well as const_vector.
12679 (Dz): Likewise. Compare against CONST0_RTX.
12680 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12681 of "vector" where appropriate.
12682 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12683 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12684 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12685 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12686 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12687 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12688 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12689 (v_int_equiv): Extend to SVE modes.
12690 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12692 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12693 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12694 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12695 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12696 (SVE_COND_FP_CMP): New int iterators.
12697 (perm_hilo): Handle the new unpack unspecs.
12698 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12700 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12701 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12702 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12703 (aarch64_equality_operator, aarch64_constant_vector_operand)
12704 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12705 (aarch64_sve_nonimmediate_operand): Likewise.
12706 (aarch64_sve_general_operand): Likewise.
12707 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12708 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12709 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12710 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12711 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12712 (aarch64_sve_float_arith_immediate): Likewise.
12713 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12714 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12715 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12716 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12717 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12718 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12719 (aarch64_sve_float_arith_operand): Likewise.
12720 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12721 (aarch64_sve_float_mul_operand): Likewise.
12722 (aarch64_sve_vec_perm_operand): Likewise.
12723 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12724 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12725 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12726 as well as const_vector.
12727 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12728 in file. Use CONST0_RTX and CONSTM1_RTX.
12729 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12730 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12731 Use aarch64_simd_imm_zero.
12732 * config/aarch64/aarch64-sve.md: New file.
12733 * config/aarch64/aarch64.md: Include it.
12734 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12735 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12736 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12737 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12738 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12739 (sve): New attribute.
12740 (enabled): Disable instructions with the sve attribute unless
12742 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12743 aarch64_expand_mov_immediate.
12744 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12745 CNT[BHSD] immediates.
12746 (movti): Split CONST_POLY_INT moves into two halves.
12747 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12748 Split additions that need a temporary here if the destination
12749 is the stack pointer.
12750 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12751 (*add<mode>3_poly_1): New instruction.
12752 (set_clobber_cc): New expander.
12754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12756 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12757 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12758 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12759 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12760 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12761 Change innermode from fixed_mode_size to machine_mode.
12762 (simplify_subreg): Update call accordingly. Handle a constant-sized
12763 subreg of a variable-length CONST_VECTOR.
12765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12770 (add_offset_to_base): New function, split out from...
12771 (create_mem_ref): ...here. When handling a scale other than 1,
12772 check first whether the address is valid without the offset.
12773 Add it into the base if so, leaving the index and scale as-is.
12775 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12778 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12779 fold_for_warn before checking if arg2 is INTEGER_CST.
12781 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12783 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12784 (store_multiple_operation): Delete.
12785 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12786 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12787 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12788 guarded by TARGET_STRING.
12789 (rs6000_output_load_multiple): Delete.
12790 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12791 OPTION_MASK_STRING / TARGET_STRING handling.
12792 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12793 (const rs6000_opt_masks) <"string">: Change mask to 0.
12794 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12795 (MASK_STRING): Delete.
12796 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12798 (load_multiple): Delete.
12805 (store_multiple): Delete.
12812 (movmemsi_8reg): Delete.
12813 (corresponding unnamed define_insn): Delete.
12814 (movmemsi_6reg): Delete.
12815 (corresponding unnamed define_insn): Delete.
12816 (movmemsi_4reg): Delete.
12817 (corresponding unnamed define_insn): Delete.
12818 (movmemsi_2reg): Delete.
12819 (corresponding unnamed define_insn): Delete.
12820 (movmemsi_1reg): Delete.
12821 (corresponding unnamed define_insn): Delete.
12822 * config/rs6000/rs6000.opt (mno-string): New.
12823 (mstring): Replace by deprecation warning stub.
12824 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12826 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12828 * regrename.c (regrename_do_replace): If replacing the same
12829 reg multiple times, try to reuse last created gen_raw_REG.
12832 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12833 main to workaround a bug in GDB.
12835 2018-01-12 Tom de Vries <tom@codesourcery.com>
12838 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12840 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12842 PR rtl-optimization/80481
12843 * ira-color.c (get_cap_member): New function.
12844 (allocnos_conflict_by_live_ranges_p): Use it.
12845 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12846 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12848 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12851 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12852 (*saddl_se_1): Ditto.
12853 (*ssubsi_1): Ditto.
12854 (*ssubl_se_1): Ditto.
12856 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12858 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12859 rather than wi::to_widest for DR_INITs.
12860 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12861 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12862 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12864 (vect_analyze_group_access_1): Note that here.
12866 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12868 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12869 polynomial type sizes.
12871 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12873 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12874 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12875 (gimple_add_tmp_var): Likewise.
12877 2018-01-12 Martin Liska <mliska@suse.cz>
12879 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12880 (gimple_alloc_sizes): Likewise.
12881 (dump_gimple_statistics): Use PRIu64 in printf format.
12882 * gimple.h: Change uint64_t to int.
12884 2018-01-12 Martin Liska <mliska@suse.cz>
12886 * tree-core.h: Use uint64_t instead of int.
12887 * tree.c (tree_node_counts): Likewise.
12888 (tree_node_sizes): Likewise.
12889 (dump_tree_statistics): Use PRIu64 in printf format.
12891 2018-01-12 Martin Liska <mliska@suse.cz>
12893 * Makefile.in: As qsort_chk is implemented in vec.c, add
12894 vec.o to linkage of gencfn-macros.
12895 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12896 passing the info to record_node_allocation_statistics.
12897 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12899 * ggc-common.c (struct ggc_usage): Add operator== and use
12900 it in operator< and compare function.
12901 * mem-stats.h (struct mem_usage): Likewise.
12902 * vec.c (struct vec_usage): Remove operator< and compare
12903 function. Can be simply inherited.
12905 2018-01-12 Martin Jambor <mjambor@suse.cz>
12908 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12909 * tree-ssa-math-opts.c: Include domwalk.h.
12910 (convert_mult_to_fma_1): New function.
12911 (fma_transformation_info): New type.
12912 (fma_deferring_state): Likewise.
12913 (cancel_fma_deferring): New function.
12914 (result_of_phi): Likewise.
12915 (last_fma_candidate_feeds_initial_phi): Likewise.
12916 (convert_mult_to_fma): Added deferring logic, split actual
12917 transformation to convert_mult_to_fma_1.
12918 (math_opts_dom_walker): New type.
12919 (math_opts_dom_walker::after_dom_children): New method, body moved
12920 here from pass_optimize_widening_mul::execute, added deferring logic
12922 (pass_optimize_widening_mul::execute): Moved most of code to
12923 math_opts_dom_walker::after_dom_children.
12924 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12925 * config/i386/i386.c (ix86_option_override_internal): Added
12926 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12928 2018-01-12 Richard Biener <rguenther@suse.de>
12931 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12932 inline instance vars.
12934 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12937 * config/rx/rx.c (rx_is_restricted_memory_address):
12938 Handle SUBREG case.
12940 2018-01-12 Richard Biener <rguenther@suse.de>
12942 PR tree-optimization/80846
12943 * target.def (split_reduction): New target hook.
12944 * targhooks.c (default_split_reduction): New function.
12945 * targhooks.h (default_split_reduction): Declare.
12946 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12947 target requests first reduce vectors by combining low and high
12949 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12950 (get_vectype_for_scalar_type_and_size): Export.
12951 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12952 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12953 * doc/tm.texi: Regenerate.
12954 * config/i386/i386.c (ix86_split_reduction): Implement
12955 TARGET_VECTORIZE_SPLIT_REDUCTION.
12957 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12960 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12961 in PIC mode except for TARGET_VXWORKS_RTP.
12962 * config/sparc/sparc.c: Include cfgrtl.h.
12963 (TARGET_INIT_PIC_REG): Define.
12964 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12965 (sparc_pic_register_p): New predicate.
12966 (sparc_legitimate_address_p): Use it.
12967 (sparc_legitimize_pic_address): Likewise.
12968 (sparc_delegitimize_address): Likewise.
12969 (sparc_mode_dependent_address_p): Likewise.
12970 (gen_load_pcrel_sym): Remove 4th parameter.
12971 (load_got_register): Adjust call to above. Remove obsolete stuff.
12972 (sparc_expand_prologue): Do not call load_got_register here.
12973 (sparc_flat_expand_prologue): Likewise.
12974 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12975 (sparc_use_pseudo_pic_reg): New function.
12976 (sparc_init_pic_reg): Likewise.
12977 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12978 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12980 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12982 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12983 Add item for branch_cost.
12985 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12987 PR rtl-optimization/83565
12988 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12989 not extend the result to a larger mode for rotate operations.
12990 (num_sign_bit_copies1): Likewise.
12992 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12995 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12997 Use values-Xc.o for -pedantic.
12998 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13000 2018-01-12 Martin Liska <mliska@suse.cz>
13003 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13005 (possible_polymorphic_call_targets): Use it.
13006 (ipa_devirt): Likewise.
13008 2018-01-12 Martin Liska <mliska@suse.cz>
13010 * profile-count.h (enum profile_quality): Use 0 as invalid
13011 enum value of profile_quality.
13013 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13015 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13016 -mext-string options.
13018 2018-01-12 Richard Biener <rguenther@suse.de>
13020 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13021 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13022 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13024 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13026 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13028 * configure.ac (--with-long-double-format): Add support for the
13029 configuration option to change the default long double format on
13031 * config.gcc (powerpc*-linux*-*): Likewise.
13032 * configure: Regenerate.
13033 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13034 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13035 used without modification.
13037 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13039 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13040 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13041 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13042 MISC_BUILTIN_SPEC_BARRIER.
13043 (rs6000_init_builtins): Likewise.
13044 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13046 (speculation_barrier): New define_insn.
13047 * doc/extend.texi: Document __builtin_speculation_barrier.
13049 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13052 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13053 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13054 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13056 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13057 integral modes instead of "ss" and "sd".
13058 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13059 vectors with 32-bit and 64-bit elements.
13060 (vecdupssescalarmodesuffix): New mode attribute.
13061 (vec_dup<mode>): Use it.
13063 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13066 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13067 frame if argument is passed on stack.
13069 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13072 * ree.c (combine_reaching_defs): Optimize also
13073 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13074 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13076 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13078 PR middle-end/83189
13079 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13081 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13083 PR middle-end/83718
13084 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13085 after they are computed.
13087 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13089 PR tree-optimization/83695
13090 * gimple-loop-linterchange.cc
13091 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13092 reset cached scev information after interchange.
13093 (pass_linterchange::execute): Remove call to scev_reset_htab.
13095 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13097 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13098 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13099 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13100 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13101 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13102 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13103 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13104 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13105 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13106 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13107 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13108 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13109 (V_lane_reg): Likewise.
13110 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13112 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13113 (vfmal_lane_low<mode>_intrinsic,
13114 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13115 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13116 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13117 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13118 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13119 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13121 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13123 * config/arm/arm-cpus.in (fp16fml): New feature.
13124 (ALL_SIMD): Add fp16fml.
13125 (armv8.2-a): Add fp16fml as an option.
13126 (armv8.3-a): Likewise.
13127 (armv8.4-a): Add fp16fml as part of fp16.
13128 * config/arm/arm.h (TARGET_FP16FML): Define.
13129 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13131 * config/arm/arm-modes.def (V2HF): Define.
13132 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13133 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13134 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13135 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13136 vfmsl_low, vfmsl_high): New set of builtins.
13137 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13138 (vfml_op): New code attribute.
13139 (VFMLHALVES): New int iterator.
13140 (VFML, VFMLSEL): New mode attributes.
13141 (V_reg): Define mapping for V2HF.
13142 (V_hi, V_lo): New mode attributes.
13143 (VF_constraint): Likewise.
13144 (vfml_half, vfml_half_selector): New int attributes.
13145 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13147 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13148 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13150 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13151 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13152 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13153 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13155 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13156 Document new effective target and option set.
13158 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13160 * config/arm/arm-cpus.in (armv8_4): New feature.
13161 (ARMv8_4a): New fgroup.
13162 (armv8.4-a): New arch.
13163 * config/arm/arm-tables.opt: Regenerate.
13164 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13165 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13166 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13167 Add matching rules for -march=armv8.4-a and extensions.
13168 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13170 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13173 * config/rx/rx.md (BW): New mode attribute.
13174 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13176 2018-01-11 Richard Biener <rguenther@suse.de>
13178 PR tree-optimization/83435
13179 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13180 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13181 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13183 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13184 Alan Hayward <alan.hayward@arm.com>
13185 David Sherwood <david.sherwood@arm.com>
13187 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13189 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13190 (aarch64_print_address_internal): Use it to check for a zero offset.
13192 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13193 Alan Hayward <alan.hayward@arm.com>
13194 David Sherwood <david.sherwood@arm.com>
13196 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13197 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13198 Return a poly_int64 rather than a HOST_WIDE_INT.
13199 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13200 rather than a HOST_WIDE_INT.
13201 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13202 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13203 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13204 final_offset from HOST_WIDE_INT to poly_int64.
13205 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13206 to_constant when getting the number of units in an Advanced SIMD
13208 (aarch64_builtin_vectorized_function): Check for a constant number
13210 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13212 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13213 attribute instead of GET_MODE_NUNITS.
13214 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13215 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13216 GET_MODE_SIZE for fixed-size registers.
13217 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13218 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13219 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13220 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13221 (aarch64_print_operand, aarch64_print_address_internal)
13222 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13223 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13224 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13225 Handle polynomial GET_MODE_SIZE.
13226 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13227 wider than SImode without modification.
13228 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13229 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13230 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13231 passing and returning SVE modes.
13232 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13233 rather than GEN_INT.
13234 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13235 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13236 (aarch64_allocate_and_probe_stack_space): Likewise.
13237 (aarch64_layout_frame): Cope with polynomial offsets.
13238 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13239 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13240 polynomial offsets.
13241 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13242 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13243 poly_int64 rather than a HOST_WIDE_INT.
13244 (aarch64_get_separate_components, aarch64_process_components)
13245 (aarch64_expand_prologue, aarch64_expand_epilogue)
13246 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13247 (aarch64_anchor_offset): New function, split out from...
13248 (aarch64_legitimize_address): ...here.
13249 (aarch64_builtin_vectorization_cost): Handle polynomial
13250 TYPE_VECTOR_SUBPARTS.
13251 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13253 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13254 number of elements from the PARALLEL rather than the mode.
13255 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13256 rather than GET_MODE_BITSIZE.
13257 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13258 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13259 (aarch64_expand_vec_perm_const_1): Handle polynomial
13260 d->perm.length () and d->perm elements.
13261 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13262 Apply to_constant to d->perm elements.
13263 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13264 polynomial CONST_VECTOR_NUNITS.
13265 (aarch64_move_pointer): Take amount as a poly_int64 rather
13267 (aarch64_progress_pointer): Avoid temporary variable.
13268 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13269 the mode attribute instead of GET_MODE.
13271 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13272 Alan Hayward <alan.hayward@arm.com>
13273 David Sherwood <david.sherwood@arm.com>
13275 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13276 x exists before using it.
13277 (aarch64_add_constant_internal): Rename to...
13278 (aarch64_add_offset_1): ...this. Replace regnum with separate
13279 src and dest rtxes. Handle the case in which they're different,
13280 including when the offset is zero. Replace scratchreg with an rtx.
13281 Use 2 additions if there is no spare register into which we can
13282 move a 16-bit constant.
13283 (aarch64_add_constant): Delete.
13284 (aarch64_add_offset): Replace reg with separate src and dest
13285 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13286 Use aarch64_add_offset_1.
13287 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13288 an rtx rather than an int. Take the delta as a poly_int64
13289 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13290 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13291 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13292 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13293 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13294 and aarch64_add_sp.
13295 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13296 aarch64_add_constant.
13298 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13300 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13301 Use scalar_float_mode.
13303 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13305 * config/aarch64/aarch64-simd.md
13306 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13307 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13308 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13309 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13310 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13311 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13312 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13313 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13314 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13315 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13317 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13320 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13321 targ_options->x_arm_arch_string is non NULL.
13323 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13325 * config/aarch64/aarch64.h
13326 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13328 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13331 * expmed.c (emit_store_flag_force): Swap if const op0
13332 and change VOIDmode to mode of op0.
13334 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13336 PR rtl-optimization/83761
13337 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13338 than bytes to mode_for_size.
13340 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13342 PR middle-end/83189
13343 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13344 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13347 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13349 PR middle-end/83575
13350 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13351 when in layout mode.
13352 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13353 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13356 2018-01-10 Michael Collison <michael.collison@arm.com>
13358 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13359 * config/aarch64/aarch64-option-extension.def: Add
13360 AARCH64_OPT_EXTENSION of 'fp16fml'.
13361 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13362 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13363 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13364 * config/aarch64/constraints.md (Ui7): New constraint.
13365 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13366 (VFMLA_SEL_W): Ditto.
13369 (VFMLA16_LOW): New int iterator.
13370 (VFMLA16_HIGH): Ditto.
13371 (UNSPEC_FMLAL): New unspec.
13372 (UNSPEC_FMLSL): Ditto.
13373 (UNSPEC_FMLAL2): Ditto.
13374 (UNSPEC_FMLSL2): Ditto.
13375 (f16mac): New code attribute.
13376 * config/aarch64/aarch64-simd-builtins.def
13377 (aarch64_fmlal_lowv2sf): Ditto.
13378 (aarch64_fmlsl_lowv2sf): Ditto.
13379 (aarch64_fmlalq_lowv4sf): Ditto.
13380 (aarch64_fmlslq_lowv4sf): Ditto.
13381 (aarch64_fmlal_highv2sf): Ditto.
13382 (aarch64_fmlsl_highv2sf): Ditto.
13383 (aarch64_fmlalq_highv4sf): Ditto.
13384 (aarch64_fmlslq_highv4sf): Ditto.
13385 (aarch64_fmlal_lane_lowv2sf): Ditto.
13386 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13387 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13388 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13389 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13390 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13391 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13392 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13393 (aarch64_fmlal_lane_highv2sf): Ditto.
13394 (aarch64_fmlsl_lane_highv2sf): Ditto.
13395 (aarch64_fmlal_laneq_highv2sf): Ditto.
13396 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13397 (aarch64_fmlalq_lane_highv4sf): Ditto.
13398 (aarch64_fmlsl_lane_highv4sf): Ditto.
13399 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13400 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13401 * config/aarch64/aarch64-simd.md:
13402 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13403 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13404 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13405 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13406 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13407 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13408 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13409 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13410 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13411 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13412 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13413 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13414 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13415 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13416 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13417 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13418 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13419 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13420 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13421 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13422 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13423 (vfmlsl_low_u32): Ditto.
13424 (vfmlalq_low_u32): Ditto.
13425 (vfmlslq_low_u32): Ditto.
13426 (vfmlal_high_u32): Ditto.
13427 (vfmlsl_high_u32): Ditto.
13428 (vfmlalq_high_u32): Ditto.
13429 (vfmlslq_high_u32): Ditto.
13430 (vfmlal_lane_low_u32): Ditto.
13431 (vfmlsl_lane_low_u32): Ditto.
13432 (vfmlal_laneq_low_u32): Ditto.
13433 (vfmlsl_laneq_low_u32): Ditto.
13434 (vfmlalq_lane_low_u32): Ditto.
13435 (vfmlslq_lane_low_u32): Ditto.
13436 (vfmlalq_laneq_low_u32): Ditto.
13437 (vfmlslq_laneq_low_u32): Ditto.
13438 (vfmlal_lane_high_u32): Ditto.
13439 (vfmlsl_lane_high_u32): Ditto.
13440 (vfmlal_laneq_high_u32): Ditto.
13441 (vfmlsl_laneq_high_u32): Ditto.
13442 (vfmlalq_lane_high_u32): Ditto.
13443 (vfmlslq_lane_high_u32): Ditto.
13444 (vfmlalq_laneq_high_u32): Ditto.
13445 (vfmlslq_laneq_high_u32): Ditto.
13446 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13447 (AARCH64_FL_FOR_ARCH8_4): New.
13448 (AARCH64_ISA_F16FML): New ISA flag.
13449 (TARGET_F16FML): New feature flag for fp16fml.
13450 (doc/invoke.texi): Document new fp16fml option.
13452 2018-01-10 Michael Collison <michael.collison@arm.com>
13454 * config/aarch64/aarch64-builtins.c:
13455 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13456 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13457 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13458 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13459 (AARCH64_ISA_SHA3): New ISA flag.
13460 (TARGET_SHA3): New feature flag for sha3.
13461 * config/aarch64/iterators.md (sha512_op): New int attribute.
13462 (CRYPTO_SHA512): New int iterator.
13463 (UNSPEC_SHA512H): New unspec.
13464 (UNSPEC_SHA512H2): Ditto.
13465 (UNSPEC_SHA512SU0): Ditto.
13466 (UNSPEC_SHA512SU1): Ditto.
13467 * config/aarch64/aarch64-simd-builtins.def
13468 (aarch64_crypto_sha512hqv2di): New builtin.
13469 (aarch64_crypto_sha512h2qv2di): Ditto.
13470 (aarch64_crypto_sha512su0qv2di): Ditto.
13471 (aarch64_crypto_sha512su1qv2di): Ditto.
13472 (aarch64_eor3qv8hi): Ditto.
13473 (aarch64_rax1qv2di): Ditto.
13474 (aarch64_xarqv2di): Ditto.
13475 (aarch64_bcaxqv8hi): Ditto.
13476 * config/aarch64/aarch64-simd.md:
13477 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13478 (aarch64_crypto_sha512su0qv2di): Ditto.
13479 (aarch64_crypto_sha512su1qv2di): Ditto.
13480 (aarch64_eor3qv8hi): Ditto.
13481 (aarch64_rax1qv2di): Ditto.
13482 (aarch64_xarqv2di): Ditto.
13483 (aarch64_bcaxqv8hi): Ditto.
13484 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13485 (vsha512h2q_u64): Ditto.
13486 (vsha512su0q_u64): Ditto.
13487 (vsha512su1q_u64): Ditto.
13488 (veor3q_u16): Ditto.
13489 (vrax1q_u64): Ditto.
13490 (vxarq_u64): Ditto.
13491 (vbcaxq_u16): Ditto.
13492 * config/arm/types.md (crypto_sha512): New type attribute.
13493 (crypto_sha3): Ditto.
13494 (doc/invoke.texi): Document new sha3 option.
13496 2018-01-10 Michael Collison <michael.collison@arm.com>
13498 * config/aarch64/aarch64-builtins.c:
13499 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13500 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13501 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13502 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13503 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13504 (AARCH64_ISA_SM4): New ISA flag.
13505 (TARGET_SM4): New feature flag for sm4.
13506 * config/aarch64/aarch64-simd-builtins.def
13507 (aarch64_sm3ss1qv4si): Ditto.
13508 (aarch64_sm3tt1aq4si): Ditto.
13509 (aarch64_sm3tt1bq4si): Ditto.
13510 (aarch64_sm3tt2aq4si): Ditto.
13511 (aarch64_sm3tt2bq4si): Ditto.
13512 (aarch64_sm3partw1qv4si): Ditto.
13513 (aarch64_sm3partw2qv4si): Ditto.
13514 (aarch64_sm4eqv4si): Ditto.
13515 (aarch64_sm4ekeyqv4si): Ditto.
13516 * config/aarch64/aarch64-simd.md:
13517 (aarch64_sm3ss1qv4si): Ditto.
13518 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13519 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13520 (aarch64_sm4eqv4si): Ditto.
13521 (aarch64_sm4ekeyqv4si): Ditto.
13522 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13523 (sm3part_op): Ditto.
13524 (CRYPTO_SM3TT): Ditto.
13525 (CRYPTO_SM3PART): Ditto.
13526 (UNSPEC_SM3SS1): New unspec.
13527 (UNSPEC_SM3TT1A): Ditto.
13528 (UNSPEC_SM3TT1B): Ditto.
13529 (UNSPEC_SM3TT2A): Ditto.
13530 (UNSPEC_SM3TT2B): Ditto.
13531 (UNSPEC_SM3PARTW1): Ditto.
13532 (UNSPEC_SM3PARTW2): Ditto.
13533 (UNSPEC_SM4E): Ditto.
13534 (UNSPEC_SM4EKEY): Ditto.
13535 * config/aarch64/constraints.md (Ui2): New constraint.
13536 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13537 * config/arm/types.md (crypto_sm3): New type attribute.
13538 (crypto_sm4): Ditto.
13539 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13540 (vsm3tt1aq_u32): Ditto.
13541 (vsm3tt1bq_u32): Ditto.
13542 (vsm3tt2aq_u32): Ditto.
13543 (vsm3tt2bq_u32): Ditto.
13544 (vsm3partw1q_u32): Ditto.
13545 (vsm3partw2q_u32): Ditto.
13546 (vsm4eq_u32): Ditto.
13547 (vsm4ekeyq_u32): Ditto.
13548 (doc/invoke.texi): Document new sm4 option.
13550 2018-01-10 Michael Collison <michael.collison@arm.com>
13552 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13553 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13554 (AARCH64_FL_FOR_ARCH8_4): New.
13555 (AARCH64_FL_V8_4): New flag.
13556 (doc/invoke.texi): Document new armv8.4-a option.
13558 2018-01-10 Michael Collison <michael.collison@arm.com>
13560 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13561 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13562 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13563 * config/aarch64/aarch64-option-extension.def: Add
13564 AARCH64_OPT_EXTENSION of 'sha2'.
13565 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13566 (crypto): Disable sha2 and aes if crypto disabled.
13567 (crypto): Enable aes and sha2 if enabled.
13568 (simd): Disable sha2 and aes if simd disabled.
13569 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13571 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13572 (TARGET_SHA2): New feature flag for sha2.
13573 (TARGET_AES): New feature flag for aes.
13574 * config/aarch64/aarch64-simd.md:
13575 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13576 conditional on TARGET_AES.
13577 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13578 (aarch64_crypto_sha1hsi): Make pattern conditional
13580 (aarch64_crypto_sha1hv4si): Ditto.
13581 (aarch64_be_crypto_sha1hv4si): Ditto.
13582 (aarch64_crypto_sha1su1v4si): Ditto.
13583 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13584 (aarch64_crypto_sha1su0v4si): Ditto.
13585 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13586 (aarch64_crypto_sha256su0v4si): Ditto.
13587 (aarch64_crypto_sha256su1v4si): Ditto.
13588 (doc/invoke.texi): Document new aes and sha2 options.
13590 2018-01-10 Martin Sebor <msebor@redhat.com>
13592 PR tree-optimization/83781
13593 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13596 2018-01-11 Martin Sebor <msebor@gmail.com>
13597 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13599 PR tree-optimization/83501
13600 PR tree-optimization/81703
13602 * tree-ssa-strlen.c (get_string_cst): Rename...
13603 (get_string_len): ...to this. Handle global constants.
13604 (handle_char_store): Adjust.
13606 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13607 Jim Wilson <jimw@sifive.com>
13609 * config/riscv/riscv-protos.h (riscv_output_return): New.
13610 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13611 (riscv_attribute_table, riscv_output_return),
13612 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13613 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13614 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13616 (riscv_expand_prologue): Add early return for naked function.
13617 (riscv_expand_epilogue): Likewise.
13618 (riscv_function_ok_for_sibcall): Return false for naked function.
13619 (riscv_set_current_function): New.
13620 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13621 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13622 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13623 * doc/extend.texi (RISC-V Function Attributes): New.
13625 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13627 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13628 check for 128-bit long double before checking TCmode.
13629 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13630 128-bit long doubles before checking TFmode or TCmode.
13631 (FLOAT128_IBM_P): Likewise.
13633 2018-01-10 Martin Sebor <msebor@redhat.com>
13635 PR tree-optimization/83671
13636 * builtins.c (c_strlen): Unconditionally return zero for the empty
13638 Use -Warray-bounds for warnings.
13639 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13640 for non-constant array indices with COMPONENT_REF, arrays of
13641 arrays, and pointers to arrays.
13642 (gimple_fold_builtin_strlen): Determine and set length range for
13643 non-constant character arrays.
13645 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13647 PR middle-end/81897
13648 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13651 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13653 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13655 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13658 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13659 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13660 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13661 indexed_or_indirect_operand predicate.
13662 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13663 (*vsx_le_perm_load_v8hi): Likewise.
13664 (*vsx_le_perm_load_v16qi): Likewise.
13665 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13666 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13667 (*vsx_le_perm_store_v8hi): Likewise.
13668 (*vsx_le_perm_store_v16qi): Likewise.
13669 (eight unnamed splitters): Likewise.
13671 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13673 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13674 * config/rs6000/emmintrin.h: Likewise.
13675 * config/rs6000/mmintrin.h: Likewise.
13676 * config/rs6000/xmmintrin.h: Likewise.
13678 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13681 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13683 * tree.c (tree_nop_conversion): Return true for location wrapper
13685 (maybe_wrap_with_location): New function.
13686 (selftest::check_strip_nops): New function.
13687 (selftest::test_location_wrappers): New function.
13688 (selftest::tree_c_tests): Call it.
13689 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13690 (maybe_wrap_with_location): New decl.
13691 (EXPR_LOCATION_WRAPPER_P): New macro.
13692 (location_wrapper_p): New inline function.
13693 (tree_strip_any_location_wrapper): New inline function.
13695 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13698 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13699 stack_realign_offset for the largest alignment of stack slot
13701 (ix86_find_max_used_stack_alignment): New function.
13702 (ix86_finalize_stack_frame_flags): Use it. Set
13703 max_used_stack_alignment if we don't realign stack.
13704 * config/i386/i386.h (machine_function): Add
13705 max_used_stack_alignment.
13707 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13709 * config/arm/arm.opt (-mbranch-cost): New option.
13710 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13713 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13716 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13717 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13719 2018-01-10 Richard Biener <rguenther@suse.de>
13722 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13723 early out so it also covers the case where we have a non-NULL
13726 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13728 PR tree-optimization/83753
13729 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13730 for non-strided grouped accesses if the number of elements is 1.
13732 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13735 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13736 * i386.h (TARGET_USE_GATHER): Define.
13737 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13739 2018-01-10 Martin Liska <mliska@suse.cz>
13742 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13743 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13745 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13746 CLEANUP_NO_PARTITIONING is not set.
13748 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13750 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13751 for vectors, as a partial revert of r254296.
13752 * rtl.h (const_vec_p): Delete.
13753 (const_vec_duplicate_p): Don't test for vector CONSTs.
13754 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13755 * expmed.c (make_tree): Likewise.
13758 * common.md (E, F): Use CONSTANT_P instead of checking for
13760 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13761 checking for CONST_VECTOR.
13763 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13765 PR middle-end/83575
13766 * predict.c (force_edge_cold): Handle in more sane way edges
13767 with no prediction.
13769 2018-01-09 Carl Love <cel@us.ibm.com>
13771 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13773 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13774 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13775 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13776 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13777 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13778 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13779 * config/rs6000/rs6000-protos.h: Add extern defition for
13780 rs6000_generate_float2_double_code.
13781 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13783 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13784 (float2_v2df): Add define_expand.
13786 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13789 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13790 op_mode in the force_to_mode call.
13792 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13794 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13795 instead of checking each element individually.
13796 (aarch64_evpc_uzp): Likewise.
13797 (aarch64_evpc_zip): Likewise.
13798 (aarch64_evpc_ext): Likewise.
13799 (aarch64_evpc_rev): Likewise.
13800 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13801 instead of checking each element individually. Return true without
13803 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13804 whether all selected elements come from the same input, instead of
13805 checking each element individually. Remove calls to gen_rtx_REG,
13806 start_sequence and end_sequence and instead assert that no rtl is
13809 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13811 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13812 order of HIGH and CONST checks.
13814 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13816 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13817 if the destination isn't an SSA_NAME.
13819 2018-01-09 Richard Biener <rguenther@suse.de>
13821 PR tree-optimization/83668
13822 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13824 (canonicalize_loop_form): ... here, renamed from ...
13825 (canonicalize_loop_closed_ssa_form): ... this and amended to
13826 swap successor edges for loop exit blocks to make us use
13827 the RPO order we need for initial schedule generation.
13829 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13831 PR tree-optimization/64811
13832 * match.pd: When optimizing comparisons with Inf, avoid
13833 introducing or losing exceptions from comparisons with NaN.
13835 2018-01-09 Martin Liska <mliska@suse.cz>
13838 * asan.c (shadow_mem_size): Add gcc_assert.
13840 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13842 Don't save registers in main().
13845 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13846 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13847 * config/avr/avr.c (avr_set_current_function): Don't error if
13848 naked, OS_task or OS_main are specified at the same time.
13849 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13851 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13853 * common/config/avr/avr-common.c (avr_option_optimization_table):
13854 Switch on -mmain-is-OS_task for optimizing compilations.
13856 2018-01-09 Richard Biener <rguenther@suse.de>
13858 PR tree-optimization/83572
13859 * graphite.c: Include cfganal.h.
13860 (graphite_transform_loops): Connect infinite loops to exit
13861 and remove fake edges at the end.
13863 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13865 * ipa-inline.c (edge_badness): Revert accidental checkin.
13867 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13870 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13871 symbols; not inline clones.
13873 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13876 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13877 hard registers. Formatting fixes.
13879 PR preprocessor/83722
13880 * gcc.c (try_generate_repro): Pass
13881 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13882 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13885 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13886 Kito Cheng <kito.cheng@gmail.com>
13888 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13889 (riscv_leaf_function_p): Delete.
13890 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13892 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13894 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13896 (do_ifelse): New function.
13897 (do_isel): New function.
13898 (do_sub3): New function.
13899 (do_add3): New function.
13900 (do_load_mask_compare): New function.
13901 (do_overlap_load_compare): New function.
13902 (expand_compare_loop): New function.
13903 (expand_block_compare): Call expand_compare_loop() when appropriate.
13904 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13905 option description.
13906 (-mblock-compare-inline-loop-limit): New option.
13908 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13911 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13912 Reverse order of second and third operands in first alternative.
13913 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13914 of first and second elements in UNSPEC_VPERMR vector.
13915 (altivec_expand_vec_perm_le): Likewise.
13917 2018-01-08 Jeff Law <law@redhat.com>
13919 PR rtl-optimizatin/81308
13920 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13921 (process_switch): If group_case_labels makes a change, then set
13923 (pass_convert_switch::execute): If a switch is converted, then
13924 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13926 PR rtl-optimization/81308
13927 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13930 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13932 PR target/83663 - Revert r255946
13933 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13934 generation for cases where splatting a value is not useful.
13935 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13936 across a vec_duplicate and a paradoxical subreg forming a vector
13937 mode to a vec_concat.
13939 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13941 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13942 -march=armv8.3-a variants.
13943 * config/arm/t-multilib: Likewise.
13944 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13946 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13948 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13950 (cceq_ior_compare_complement): Give it a name so I can use it, and
13951 change boolean_or_operator predicate to boolean_operator so it can
13952 be used to generate a crand.
13953 (eqne): New code iterator.
13954 (bd/bd_neg): New code_attrs.
13955 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13956 a single define_insn.
13957 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13958 decrement (bdnzt/bdnzf/bdzt/bdzf).
13959 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13960 with the new names of the branch decrement patterns, and added the
13961 names of the branch decrement conditional patterns.
13963 2018-01-08 Richard Biener <rguenther@suse.de>
13965 PR tree-optimization/83563
13966 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13969 2018-01-08 Richard Biener <rguenther@suse.de>
13971 PR middle-end/83713
13972 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13974 2018-01-08 Richard Biener <rguenther@suse.de>
13976 PR tree-optimization/83685
13977 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13978 references to abnormals.
13980 2018-01-08 Richard Biener <rguenther@suse.de>
13983 * dwarf2out.c (output_indirect_strings): Handle empty
13984 skeleton_debug_str_hash.
13985 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13987 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13989 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13990 (emit_store_direct): Likewise.
13991 (arc_trampoline_adjust_address): Likewise.
13992 (arc_asm_trampoline_template): New function.
13993 (arc_initialize_trampoline): Use asm_trampoline_template.
13994 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13995 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13996 * config/arc/arc.md (flush_icache): Delete pattern.
13998 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14000 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14001 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14004 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14007 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14008 by not USED_FOR_TARGET.
14009 (make_pass_resolve_sw_modes): Likewise.
14011 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14013 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14016 2018-01-08 Richard Biener <rguenther@suse.de>
14018 PR middle-end/83580
14019 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14021 2018-01-08 Richard Biener <rguenther@suse.de>
14023 PR middle-end/83517
14024 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14026 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14028 PR middle-end/81897
14029 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14030 basic blocks with a small number of successors.
14031 (convert_control_dep_chain_into_preds): Improve handling of
14033 (dump_predicates): Split apart into...
14034 (dump_pred_chain): ...here...
14035 (dump_pred_info): ...and here.
14036 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14037 (can_chain_union_be_invalidated_p): Improve check for invalidation
14039 (uninit_uses_cannot_happen): Avoid unnecessary if
14040 convert_control_dep_chain_into_preds yielded nothing.
14042 2018-01-06 Martin Sebor <msebor@redhat.com>
14044 PR tree-optimization/83640
14045 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14046 subtracting negative offset from size.
14047 (builtin_access::overlap): Adjust offset bounds of the access to fall
14048 within the size of the object if possible.
14050 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14052 PR rtl-optimization/83699
14053 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14054 extract_bit_field_as_subreg to cases in which the extracted
14055 value is also a vector.
14057 * lra-constraints.c (process_alt_operands): Test for the equivalence
14058 substitutions when detecting a possible reload cycle.
14060 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14063 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14064 by default if flag_selective_schedling{,2}. Formatting fixes.
14066 PR rtl-optimization/83682
14067 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14068 if it has non-VECTOR_MODE element mode.
14069 (vec_duplicate_p): Likewise.
14071 PR middle-end/83694
14072 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14073 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14075 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14078 * config/i386/i386-builtin.def
14079 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14080 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14081 Require also OPTION_MASK_ISA_AVX512F in addition to
14082 OPTION_MASK_ISA_GFNI.
14083 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14084 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14085 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14086 to OPTION_MASK_ISA_GFNI.
14087 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14088 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14089 OPTION_MASK_ISA_AVX512BW.
14090 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14091 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14092 addition to OPTION_MASK_ISA_GFNI.
14093 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14094 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14095 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14096 to OPTION_MASK_ISA_GFNI.
14097 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14098 a requirement for all ISAs rather than any of them with a few
14100 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14102 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14103 bitmasks to be enabled with 3 exceptions, instead of requiring any
14104 enabled ISA with lots of exceptions.
14105 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14106 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14107 Change avx512bw in isa attribute to avx512f.
14108 * config/i386/sgxintrin.h: Add license boilerplate.
14109 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14110 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14111 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14112 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14114 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14115 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14116 temporarily sse2 rather than sse if not enabled already.
14119 * config/i386/sse.md (VI248_VLBW): Rename to ...
14120 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14121 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14122 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14123 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14124 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14125 mode iterator instead of VI248_VLBW.
14127 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14129 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14130 (record_modified): Skip clobbers; add debug output.
14131 (param_change_prob): Use sreal frequencies.
14133 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14135 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14136 punt for user-aligned variables.
14138 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14140 * tree-chrec.c (chrec_contains_symbols): Return true for
14143 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14146 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14147 of (x|y) == x for BICS pattern.
14149 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14151 PR tree-optimization/83605
14152 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14153 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14156 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14158 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14159 * config/epiphany/rtems.h: New file.
14161 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14162 Uros Bizjak <ubizjak@gmail.com>
14165 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14166 QIreg_operand instead of register_operand predicate.
14167 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14168 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14169 comments instead of -fmitigate[-_]rop.
14171 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14174 * cgraphunit.c (symbol_table::compile): Switch to text_section
14175 before calling assembly_start debug hook.
14176 * run-rtl-passes.c (run_rtl_passes): Likewise.
14179 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14181 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14182 range_int_cst_p rather than !symbolic_range_p before calling
14183 extract_range_from_multiplicative_op_1.
14185 2018-01-04 Jeff Law <law@redhat.com>
14187 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14188 redundant test in assertion.
14190 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14192 * doc/rtl.texi: Document machine_mode wrapper classes.
14194 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14196 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14197 using tree_to_uhwi.
14199 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14201 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14202 the VEC_PERM_EXPR fold to fail.
14204 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14207 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14208 to switched_sections.
14210 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14213 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14214 test for d.testing.
14216 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14219 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14220 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14222 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14225 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14226 is BLKmode and bitpos not zero or mode change is needed.
14228 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14231 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14234 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14237 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14238 instead of MULT rtx. Update all corresponding splitters.
14239 (*saddl_se): Ditto.
14240 (*ssub<modesuffix>): Ditto.
14241 (*ssubl_se): Ditto.
14242 (*cmp_sadd_di): Update split patterns.
14243 (*cmp_sadd_si): Ditto.
14244 (*cmp_sadd_sidi): Ditto.
14245 (*cmp_ssub_di): Ditto.
14246 (*cmp_ssub_si): Ditto.
14247 (*cmp_ssub_sidi): Ditto.
14248 * config/alpha/predicates.md (const23_operand): New predicate.
14249 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14250 Look for ASHIFT, not MULT inner operand.
14251 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14253 2018-01-04 Martin Liska <mliska@suse.cz>
14255 PR gcov-profile/83669
14256 * gcov.c (output_intermediate_file): Add version to intermediate
14258 * doc/gcov.texi: Document new field 'version' in intermediate
14259 file format. Fix location of '-k' option of gcov command.
14261 2018-01-04 Martin Liska <mliska@suse.cz>
14264 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14266 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14268 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14270 2018-01-03 Martin Sebor <msebor@redhat.com>
14272 PR tree-optimization/83655
14273 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14274 checking calls with invalid arguments.
14276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14278 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14279 (vectorizable_mask_load_store): Delete.
14280 (vectorizable_call): Return false for masked loads and stores.
14281 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14282 instead of gimple_assign_rhs1.
14283 (vectorizable_load): Handle IFN_MASK_LOAD.
14284 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14288 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14290 (vectorizable_mask_load_store): ...here.
14291 (vectorizable_load): ...and here.
14293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14295 * tree-vect-stmts.c (vect_build_all_ones_mask)
14296 (vect_build_zero_merge_argument): New functions, split out from...
14297 (vectorizable_load): ...here.
14299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14301 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14303 (vectorizable_mask_load_store): ...here.
14304 (vectorizable_store): ...and here.
14306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14308 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14310 (vectorizable_mask_load_store): ...here.
14312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14314 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14315 (vect_model_store_cost): Take a vec_load_store_type instead of a
14317 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14318 (vect_model_store_cost): Take a vec_load_store_type instead of a
14320 (vectorizable_mask_load_store): Update accordingly.
14321 (vectorizable_store): Likewise.
14322 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14326 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14327 IFN_MASK_LOAD calls here rather than...
14328 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14331 Alan Hayward <alan.hayward@arm.com>
14332 David Sherwood <david.sherwood@arm.com>
14334 * expmed.c (extract_bit_field_1): For vector extracts,
14335 fall back to extract_bit_field_as_subreg if vec_extract
14338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14339 Alan Hayward <alan.hayward@arm.com>
14340 David Sherwood <david.sherwood@arm.com>
14342 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14343 they are variable or constant sized.
14344 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14345 slots for constant-sized data.
14347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14348 Alan Hayward <alan.hayward@arm.com>
14349 David Sherwood <david.sherwood@arm.com>
14351 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14352 handling COND_EXPRs with boolean comparisons, try to find a better
14353 basis for the mask type than the boolean itself.
14355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14357 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14358 is calculated and how it can be overridden.
14359 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14360 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14362 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14366 Alan Hayward <alan.hayward@arm.com>
14367 David Sherwood <david.sherwood@arm.com>
14369 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14370 Remove the mode argument.
14371 (aarch64_simd_valid_immediate): Remove the mode and inverse
14373 * config/aarch64/iterators.md (bitsize): New iterator.
14374 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14375 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14376 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14377 aarch64_simd_valid_immediate.
14378 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14379 (aarch64_reg_or_bic_imm): Likewise.
14380 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14381 with an insn_type enum and msl with a modifier_type enum.
14382 Replace element_width with a scalar_mode. Change the shift
14383 to unsigned int. Add constructors for scalar_float_mode and
14384 scalar_int_mode elements.
14385 (aarch64_vect_float_const_representable_p): Delete.
14386 (aarch64_can_const_movi_rtx_p)
14387 (aarch64_simd_scalar_immediate_valid_for_move)
14388 (aarch64_simd_make_constant): Update call to
14389 aarch64_simd_valid_immediate.
14390 (aarch64_advsimd_valid_immediate_hs): New function.
14391 (aarch64_advsimd_valid_immediate): Likewise.
14392 (aarch64_simd_valid_immediate): Remove mode and inverse
14393 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14394 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14395 and aarch64_float_const_representable_p on the result.
14396 (aarch64_output_simd_mov_immediate): Remove mode argument.
14397 Update call to aarch64_simd_valid_immediate and use of
14398 simd_immediate_info.
14399 (aarch64_output_scalar_simd_mov_immediate): Update call
14402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14403 Alan Hayward <alan.hayward@arm.com>
14404 David Sherwood <david.sherwood@arm.com>
14406 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14407 (mode_nunits): Likewise CONST_MODE_NUNITS.
14408 * machmode.def (ADJUST_NUNITS): Document.
14409 * genmodes.c (mode_data::need_nunits_adj): New field.
14410 (blank_mode): Update accordingly.
14411 (adj_nunits): New variable.
14412 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14414 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14415 listed in adj_nunits.
14416 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14417 listed in adj_nunits. Don't emit case statements for such modes.
14418 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14419 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14420 nothing if adj_nunits is nonnull.
14421 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14422 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14423 (emit_mode_fbit): Update use of print_maybe_const_decl.
14424 (emit_move_size): Likewise. Treat the array as non-const
14426 (emit_mode_adjustments): Handle adj_nunits.
14428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14430 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14431 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14432 (VECTOR_MODES): Use it.
14433 (make_vector_modes): Take the prefix as an argument.
14435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14436 Alan Hayward <alan.hayward@arm.com>
14437 David Sherwood <david.sherwood@arm.com>
14439 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14440 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14441 for MODE_VECTOR_BOOL.
14442 * machmode.def (VECTOR_BOOL_MODE): Document.
14443 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14444 (make_vector_bool_mode): New function.
14445 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14447 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14448 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14450 * stor-layout.c (int_mode_for_mode): Likewise.
14451 * tree.c (build_vector_type_for_mode): Likewise.
14452 * varasm.c (output_constant_pool_2): Likewise.
14453 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14454 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14455 for MODE_VECTOR_BOOL.
14456 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14457 of mode class checks.
14458 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14459 instead of a list of mode class checks.
14460 (expand_vector_scalar_condition): Likewise.
14461 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14464 Alan Hayward <alan.hayward@arm.com>
14465 David Sherwood <david.sherwood@arm.com>
14467 * machmode.h (mode_size): Change from unsigned short to
14469 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14470 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14471 or if measurement_type is not polynomial.
14472 (fixed_size_mode::includes_p): Check for constant-sized modes.
14473 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14474 return a poly_uint16 rather than an unsigned short.
14475 (emit_mode_size): Change the type of mode_size from unsigned short
14476 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14477 (emit_mode_adjustments): Cope with polynomial vector sizes.
14478 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14480 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14482 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14483 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14484 * caller-save.c (setup_save_areas): Likewise.
14485 (replace_reg_with_saved_mem): Likewise.
14486 * calls.c (emit_library_call_value_1): Likewise.
14487 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14488 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14489 (gen_lowpart_for_combine): Likewise.
14490 * convert.c (convert_to_integer_1): Likewise.
14491 * cse.c (equiv_constant, cse_insn): Likewise.
14492 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14493 (cselib_subst_to_values): Likewise.
14494 * dce.c (word_dce_process_block): Likewise.
14495 * df-problems.c (df_word_lr_mark_ref): Likewise.
14496 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14497 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14498 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14499 (rtl_for_decl_location): Likewise.
14500 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14501 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14502 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14503 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14504 (expand_expr_real_1): Likewise.
14505 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14506 (pad_below): Likewise.
14507 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14508 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14509 * ira.c (get_subreg_tracking_sizes): Likewise.
14510 * ira-build.c (ira_create_allocno_objects): Likewise.
14511 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14512 (ira_sort_regnos_for_alter_reg): Likewise.
14513 * ira-costs.c (record_operand_costs): Likewise.
14514 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14515 (resolve_simple_move): Likewise.
14516 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14517 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14518 (lra_constraints): Likewise.
14519 (CONST_POOL_OK_P): Reject variable-sized modes.
14520 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14521 (add_pseudo_to_slot, lra_spill): Likewise.
14522 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14523 * optabs-query.c (get_best_extraction_insn): Likewise.
14524 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14525 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14526 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14527 * recog.c (offsettable_address_addr_space_p): Likewise.
14528 * regcprop.c (maybe_mode_change): Likewise.
14529 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14530 * regrename.c (build_def_use): Likewise.
14531 * regstat.c (dump_reg_info): Likewise.
14532 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14533 (find_reloads, find_reloads_subreg_address): Likewise.
14534 * reload1.c (eliminate_regs_1): Likewise.
14535 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14536 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14537 (simplify_binary_operation_1, simplify_subreg): Likewise.
14538 * targhooks.c (default_function_arg_padding): Likewise.
14539 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14540 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14541 (verify_gimple_assign_ternary): Likewise.
14542 * tree-inline.c (estimate_move_cost): Likewise.
14543 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14544 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14545 (get_address_cost_ainc): Likewise.
14546 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14547 (vect_supportable_dr_alignment): Likewise.
14548 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14549 (vectorizable_reduction): Likewise.
14550 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14551 (vectorizable_operation, vectorizable_load): Likewise.
14552 * tree.c (build_same_sized_truth_vector_type): Likewise.
14553 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14554 * var-tracking.c (emit_note_insn_var_location): Likewise.
14555 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14556 (ADDR_VEC_ALIGN): Likewise.
14558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14559 Alan Hayward <alan.hayward@arm.com>
14560 David Sherwood <david.sherwood@arm.com>
14562 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14564 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14565 or if measurement_type is polynomial.
14566 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14567 * combine.c (make_extraction): Likewise.
14568 * dse.c (find_shift_sequence): Likewise.
14569 * dwarf2out.c (mem_loc_descriptor): Likewise.
14570 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14571 (extract_bit_field, extract_low_bits): Likewise.
14572 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14573 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14574 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14575 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14576 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14577 * reload.c (find_reloads): Likewise.
14578 * reload1.c (alter_reg): Likewise.
14579 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14580 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14581 * tree-if-conv.c (predicate_mem_writes): Likewise.
14582 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14583 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14584 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14585 * valtrack.c (dead_debug_insert_temp): Likewise.
14586 * varasm.c (mergeable_constant_section): Likewise.
14587 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14590 Alan Hayward <alan.hayward@arm.com>
14591 David Sherwood <david.sherwood@arm.com>
14593 * expr.c (expand_assignment): Cope with polynomial mode sizes
14594 when assigning to a CONCAT.
14596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14597 Alan Hayward <alan.hayward@arm.com>
14598 David Sherwood <david.sherwood@arm.com>
14600 * machmode.h (mode_precision): Change from unsigned short to
14602 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14604 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14605 or if measurement_type is not polynomial.
14606 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14607 in which the mode is already known to be a scalar_int_mode.
14608 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14609 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14611 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14612 for GET_MODE_PRECISION.
14613 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14614 for GET_MODE_PRECISION.
14615 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14617 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14618 (expand_field_assignment, make_extraction): Likewise.
14619 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14620 (get_last_value): Likewise.
14621 * convert.c (convert_to_integer_1): Likewise.
14622 * cse.c (cse_insn): Likewise.
14623 * expr.c (expand_expr_real_1): Likewise.
14624 * lra-constraints.c (simplify_operand_subreg): Likewise.
14625 * optabs-query.c (can_atomic_load_p): Likewise.
14626 * optabs.c (expand_atomic_load): Likewise.
14627 (expand_atomic_store): Likewise.
14628 * ree.c (combine_reaching_defs): Likewise.
14629 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14630 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14631 * tree.h (type_has_mode_precision_p): Likewise.
14632 * ubsan.c (instrument_si_overflow): Likewise.
14634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14635 Alan Hayward <alan.hayward@arm.com>
14636 David Sherwood <david.sherwood@arm.com>
14638 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14639 polynomial numbers of units.
14640 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14641 (valid_vector_subparts_p): New function.
14642 (build_vector_type): Remove temporary shim and take the number
14643 of units as a poly_uint64 rather than an int.
14644 (build_opaque_vector_type): Take the number of units as a
14645 poly_uint64 rather than an int.
14646 * tree.c (build_vector_from_ctor): Handle polynomial
14647 TYPE_VECTOR_SUBPARTS.
14648 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14649 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14650 (build_vector_from_val): If the number of units is variable,
14651 use build_vec_duplicate_cst for constant operands and
14652 VEC_DUPLICATE_EXPR otherwise.
14653 (make_vector_type): Remove temporary is_constant ().
14654 (build_vector_type, build_opaque_vector_type): Take the number of
14655 units as a poly_uint64 rather than an int.
14656 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14658 * cfgexpand.c (expand_debug_expr): Likewise.
14659 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14660 (store_constructor, expand_expr_real_1): Likewise.
14661 (const_scalar_mask_from_tree): Likewise.
14662 * fold-const-call.c (fold_const_reduction): Likewise.
14663 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14664 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14665 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14666 (fold_relational_const): Likewise.
14667 (native_interpret_vector): Likewise. Change the size from an
14668 int to an unsigned int.
14669 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14670 TYPE_VECTOR_SUBPARTS.
14671 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14672 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14673 duplicating a non-constant operand into a variable-length vector.
14674 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14675 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14676 * ipa-icf.c (sem_variable::equals): Likewise.
14677 * match.pd: Likewise.
14678 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14679 * print-tree.c (print_node): Likewise.
14680 * stor-layout.c (layout_type): Likewise.
14681 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14682 * tree-cfg.c (verify_gimple_comparison): Likewise.
14683 (verify_gimple_assign_binary): Likewise.
14684 (verify_gimple_assign_ternary): Likewise.
14685 (verify_gimple_assign_single): Likewise.
14686 * tree-pretty-print.c (dump_generic_node): Likewise.
14687 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14688 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14689 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14690 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14691 (vect_shift_permute_load_chain): Likewise.
14692 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14693 (expand_vector_condition, optimize_vector_constructor): Likewise.
14694 (lower_vec_perm, get_compute_type): Likewise.
14695 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14696 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14697 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14698 (vect_recog_mask_conversion_pattern): Likewise.
14699 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14700 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14701 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14702 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14703 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14704 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14705 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14706 (supportable_widening_operation): Likewise.
14707 (supportable_narrowing_operation): Likewise.
14708 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14710 * varasm.c (output_constant): Likewise.
14712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14713 Alan Hayward <alan.hayward@arm.com>
14714 David Sherwood <david.sherwood@arm.com>
14716 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14717 so that both the length == 3 and length != 3 cases set up their
14718 own permute vectors. Add comments explaining why we know the
14719 number of elements is constant.
14720 (vect_permute_load_chain): Likewise.
14722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14723 Alan Hayward <alan.hayward@arm.com>
14724 David Sherwood <david.sherwood@arm.com>
14726 * machmode.h (mode_nunits): Change from unsigned char to
14728 (ONLY_FIXED_SIZE_MODES): New macro.
14729 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14730 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14731 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14733 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14734 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14735 or if measurement_type is not polynomial.
14736 * genmodes.c (ZERO_COEFFS): New macro.
14737 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14739 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14740 Use ZERO_COEFFS when emitting initializers.
14741 * data-streamer.h (bp_pack_poly_value): New function.
14742 (bp_unpack_poly_value): Likewise.
14743 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14744 for GET_MODE_NUNITS.
14745 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14746 for GET_MODE_NUNITS.
14747 * tree.c (make_vector_type): Remove temporary shim and make
14748 the real function take the number of units as a poly_uint64
14749 rather than an int.
14750 (build_vector_type_for_mode): Handle polynomial nunits.
14751 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14752 * emit-rtl.c (const_vec_series_p_1): Likewise.
14753 (gen_rtx_CONST_VECTOR): Likewise.
14754 * fold-const.c (test_vec_duplicate_folding): Likewise.
14755 * genrecog.c (validate_pattern): Likewise.
14756 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14757 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14758 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14759 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14760 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14761 * rtlanal.c (subreg_get_info): Likewise.
14762 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14763 (vect_grouped_load_supported): Likewise.
14764 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14765 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14766 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14767 (simplify_const_unary_operation, simplify_binary_operation_1)
14768 (simplify_const_binary_operation, simplify_ternary_operation)
14769 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14770 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14771 instead of CONST_VECTOR_NUNITS.
14772 * varasm.c (output_constant_pool_2): Likewise.
14773 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14774 explicit-encoded elements in the XVEC for variable-length vectors.
14776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14778 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14781 Alan Hayward <alan.hayward@arm.com>
14782 David Sherwood <david.sherwood@arm.com>
14784 * coretypes.h (fixed_size_mode): Declare.
14785 (fixed_size_mode_pod): New typedef.
14786 * builtins.h (target_builtins::x_apply_args_mode)
14787 (target_builtins::x_apply_result_mode): Change type to
14788 fixed_size_mode_pod.
14789 * builtins.c (apply_args_size, apply_result_size, result_vector)
14790 (expand_builtin_apply_args_1, expand_builtin_apply)
14791 (expand_builtin_return): Update accordingly.
14793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14795 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14796 * cselib.c (cselib_hash_rtx): Likewise.
14797 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14798 CONST_VECTOR encoding.
14800 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14801 Jeff Law <law@redhat.com>
14804 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14805 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14806 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14807 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14810 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14811 explicitly probe *sp in a noreturn function if there were any callee
14812 register saves or frame pointer is needed.
14814 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14817 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14818 BLKmode for ternary, binary or unary expressions.
14821 * var-tracking.c (delete_vta_debug_insn): New inline function.
14822 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14823 insns from get_insns () to NULL instead of each bb separately.
14824 Use delete_vta_debug_insn. No longer static.
14825 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14826 delete_vta_debug_insns callers.
14827 * rtl.h (delete_vta_debug_insns): Declare.
14828 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14829 instead of variable_tracking_main.
14831 2018-01-03 Martin Sebor <msebor@redhat.com>
14833 PR tree-optimization/83603
14834 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14835 arguments past the endof the argument list in functions declared
14836 without a prototype.
14837 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14838 Avoid checking when arguments are null.
14840 2018-01-03 Martin Sebor <msebor@redhat.com>
14843 * doc/extend.texi (attribute const): Fix a typo.
14844 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14845 issuing -Wsuggest-attribute for void functions.
14847 2018-01-03 Martin Sebor <msebor@redhat.com>
14849 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14850 offset_int::from instead of wide_int::to_shwi.
14851 (maybe_diag_overlap): Remove assertion.
14852 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14853 * gimple-ssa-sprintf.c (format_directive): Same.
14854 (parse_directive): Same.
14855 (sprintf_dom_walker::compute_format_length): Same.
14856 (try_substitute_return_value): Same.
14858 2018-01-03 Jeff Law <law@redhat.com>
14860 PR middle-end/83654
14861 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14862 non-constant residual for zero at runtime and avoid probing in
14863 that case. Reorganize code for trailing problem to mirror handling
14866 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14868 PR tree-optimization/83501
14869 * tree-ssa-strlen.c (get_string_cst): New.
14870 (handle_char_store): Call get_string_cst.
14872 2018-01-03 Martin Liska <mliska@suse.cz>
14874 PR tree-optimization/83593
14875 * tree-ssa-strlen.c: Include tree-cfg.h.
14876 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14877 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14878 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14880 (strlen_dom_walker::before_dom_children): Call
14881 gimple_purge_dead_eh_edges. Dump tranformation with details
14883 (strlen_dom_walker::before_dom_children): Update call by adding
14884 new argument cleanup_eh.
14885 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14887 2018-01-03 Martin Liska <mliska@suse.cz>
14890 * cif-code.def (VARIADIC_THUNK): New enum value.
14891 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14894 2018-01-03 Jan Beulich <jbeulich@suse.com>
14896 * sse.md (mov<mode>_internal): Tighten condition for when to use
14897 vmovdqu<ssescalarsize> for TI and OI modes.
14899 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14901 Update copyright years.
14903 2018-01-03 Martin Liska <mliska@suse.cz>
14906 * ipa-visibility.c (function_and_variable_visibility): Skip
14907 functions with noipa attribure.
14909 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14911 * gcc.c (process_command): Update copyright notice dates.
14912 * gcov-dump.c (print_version): Ditto.
14913 * gcov.c (print_version): Ditto.
14914 * gcov-tool.c (print_version): Ditto.
14915 * gengtype.c (create_file): Ditto.
14916 * doc/cpp.texi: Bump @copying's copyright year.
14917 * doc/cppinternals.texi: Ditto.
14918 * doc/gcc.texi: Ditto.
14919 * doc/gccint.texi: Ditto.
14920 * doc/gcov.texi: Ditto.
14921 * doc/install.texi: Ditto.
14922 * doc/invoke.texi: Ditto.
14924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14926 * vector-builder.h (vector_builder::m_full_nelts): Change from
14927 unsigned int to poly_uint64.
14928 (vector_builder::full_nelts): Update prototype accordingly.
14929 (vector_builder::new_vector): Likewise.
14930 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14931 (vector_builder::operator ==): Likewise.
14932 (vector_builder::finalize): Likewise.
14933 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14934 Take the number of elements as a poly_uint64 rather than an
14936 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14937 from unsigned int to poly_uint64.
14938 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14939 (vec_perm_indices::new_vector): Likewise.
14940 (vec_perm_indices::length): Likewise.
14941 (vec_perm_indices::nelts_per_input): Likewise.
14942 (vec_perm_indices::input_nelts): Likewise.
14943 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14944 number of elements per input as a poly_uint64 rather than an
14945 unsigned int. Use the original encoding for variable-length
14946 vectors, rather than clamping each individual element.
14947 For the second and subsequent elements in each pattern,
14948 clamp the step and base before clamping their sum.
14949 (vec_perm_indices::series_p): Handle polynomial element counts.
14950 (vec_perm_indices::all_in_range_p): Likewise.
14951 (vec_perm_indices_to_tree): Likewise.
14952 (vec_perm_indices_to_rtx): Likewise.
14953 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14954 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14955 (tree_vector_builder::new_binary_operation): Handle polynomial
14956 element counts. Return false if we need to know the number
14957 of elements at compile time.
14958 * fold-const.c (fold_vec_perm): Punt if the number of elements
14959 isn't known at compile time.
14961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14963 * vec-perm-indices.h (vec_perm_builder): Change element type
14964 from HOST_WIDE_INT to poly_int64.
14965 (vec_perm_indices::element_type): Update accordingly.
14966 (vec_perm_indices::clamp): Handle polynomial element_types.
14967 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14968 (vec_perm_indices::all_in_range_p): Likewise.
14969 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14971 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14972 polynomial vec_perm_indices element types.
14973 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14974 * fold-const.c (fold_vec_perm): Likewise.
14975 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14976 * tree-vect-generic.c (lower_vec_perm): Likewise.
14977 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14978 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14979 element type to HOST_WIDE_INT.
14981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14982 Alan Hayward <alan.hayward@arm.com>
14983 David Sherwood <david.sherwood@arm.com>
14985 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14986 rather than an int. Use plus_constant.
14987 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14988 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14991 Alan Hayward <alan.hayward@arm.com>
14992 David Sherwood <david.sherwood@arm.com>
14994 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14995 a HOST_WIDE_INT to a poly_int64.
14997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14998 Alan Hayward <alan.hayward@arm.com>
14999 David Sherwood <david.sherwood@arm.com>
15001 * calls.c (load_register_parameters): Cope with polynomial
15002 mode sizes. Require a constant size for BLKmode parameters
15003 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15004 forces a parameter to be padded at the lsb end in order to
15005 fill a complete number of words, require the parameter size
15006 to be ordered wrt UNITS_PER_WORD.
15008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15009 Alan Hayward <alan.hayward@arm.com>
15010 David Sherwood <david.sherwood@arm.com>
15012 * reload1.c (spill_stack_slot_width): Change element type
15013 from unsigned int to poly_uint64_pod.
15014 (alter_reg): Treat mode sizes as polynomial.
15016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15017 Alan Hayward <alan.hayward@arm.com>
15018 David Sherwood <david.sherwood@arm.com>
15020 * reload.c (complex_word_subreg_p): New function.
15021 (reload_inner_reg_of_subreg, push_reload): Use it.
15023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15024 Alan Hayward <alan.hayward@arm.com>
15025 David Sherwood <david.sherwood@arm.com>
15027 * lra-constraints.c (process_alt_operands): Reject matched
15028 operands whose sizes aren't ordered.
15029 (match_reload): Refer to this check here.
15031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15032 Alan Hayward <alan.hayward@arm.com>
15033 David Sherwood <david.sherwood@arm.com>
15035 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15036 that the mode size is in the set {1, 2, 4, 8, 16}.
15038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15039 Alan Hayward <alan.hayward@arm.com>
15040 David Sherwood <david.sherwood@arm.com>
15042 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15043 Use plus_constant instead of gen_rtx_PLUS.
15045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15046 Alan Hayward <alan.hayward@arm.com>
15047 David Sherwood <david.sherwood@arm.com>
15049 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15050 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15051 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15052 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15053 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15054 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15055 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15056 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15057 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15058 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15060 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15061 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15062 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15063 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15064 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15065 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15066 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15067 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15068 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15069 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15071 * expr.c (emit_move_resolve_push): Treat the input and result
15072 of PUSH_ROUNDING as a poly_int64.
15073 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15074 (emit_push_insn): Likewise.
15075 * lra-eliminations.c (mark_not_eliminable): Likewise.
15076 * recog.c (push_operand): Likewise.
15077 * reload1.c (elimination_effects): Likewise.
15078 * rtlanal.c (nonzero_bits1): Likewise.
15079 * calls.c (store_one_arg): Likewise. Require the padding to be
15080 known at compile time.
15082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15083 Alan Hayward <alan.hayward@arm.com>
15084 David Sherwood <david.sherwood@arm.com>
15086 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15087 Use plus_constant instead of gen_rtx_PLUS.
15089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15090 Alan Hayward <alan.hayward@arm.com>
15091 David Sherwood <david.sherwood@arm.com>
15093 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15094 rather than an int.
15096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15097 Alan Hayward <alan.hayward@arm.com>
15098 David Sherwood <david.sherwood@arm.com>
15100 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15101 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15102 via stack temporaries. Treat the mode size as polynomial too.
15104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15105 Alan Hayward <alan.hayward@arm.com>
15106 David Sherwood <david.sherwood@arm.com>
15108 * expr.c (expand_expr_real_2): When handling conversions involving
15109 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15110 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15111 as a poly_uint64 too.
15113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15114 Alan Hayward <alan.hayward@arm.com>
15115 David Sherwood <david.sherwood@arm.com>
15117 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15120 Alan Hayward <alan.hayward@arm.com>
15121 David Sherwood <david.sherwood@arm.com>
15123 * combine.c (can_change_dest_mode): Handle polynomial
15124 REGMODE_NATURAL_SIZE.
15125 * expmed.c (store_bit_field_1): Likewise.
15126 * expr.c (store_constructor): Likewise.
15127 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15128 and polynomial REGMODE_NATURAL_SIZE.
15129 (gen_lowpart_common): Likewise.
15130 * reginfo.c (record_subregs_of_mode): Likewise.
15131 * rtlanal.c (read_modify_subreg_p): Likewise.
15133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15134 Alan Hayward <alan.hayward@arm.com>
15135 David Sherwood <david.sherwood@arm.com>
15137 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15138 numbers of elements.
15140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15141 Alan Hayward <alan.hayward@arm.com>
15142 David Sherwood <david.sherwood@arm.com>
15144 * match.pd: Cope with polynomial numbers of vector elements.
15146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15147 Alan Hayward <alan.hayward@arm.com>
15148 David Sherwood <david.sherwood@arm.com>
15150 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15151 in a POINTER_PLUS_EXPR.
15153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15154 Alan Hayward <alan.hayward@arm.com>
15155 David Sherwood <david.sherwood@arm.com>
15157 * omp-simd-clone.c (simd_clone_subparts): New function.
15158 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15159 (ipa_simd_modify_function_body): Likewise.
15161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15162 Alan Hayward <alan.hayward@arm.com>
15163 David Sherwood <david.sherwood@arm.com>
15165 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15166 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15167 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15168 (expand_vector_condition, vector_element): Likewise.
15169 (subparts_gt): New function.
15170 (get_compute_type): Use subparts_gt.
15171 (count_type_subparts): Delete.
15172 (expand_vector_operations_1): Use subparts_gt instead of
15173 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15180 (vect_compile_time_alias): ...this new function. Do the calculation
15181 on poly_ints rather than trees.
15182 (vect_prune_runtime_alias_test_list): Update call accordingly.
15184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15185 Alan Hayward <alan.hayward@arm.com>
15186 David Sherwood <david.sherwood@arm.com>
15188 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15190 (vect_schedule_slp_instance): Likewise.
15192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15193 Alan Hayward <alan.hayward@arm.com>
15194 David Sherwood <david.sherwood@arm.com>
15196 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15197 constant and extern definitions for variable-length vectors.
15198 (vect_get_constant_vectors): Note that the number of units
15199 is known to be constant.
15201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15202 Alan Hayward <alan.hayward@arm.com>
15203 David Sherwood <david.sherwood@arm.com>
15205 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15206 of units as polynomial. Choose between WIDE and NARROW based
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 * tree-vect-stmts.c (simd_clone_subparts): New function.
15214 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15217 Alan Hayward <alan.hayward@arm.com>
15218 David Sherwood <david.sherwood@arm.com>
15220 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15221 vectors as polynomial. Use build_index_vector for
15222 IFN_GOMP_SIMD_LANE.
15224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15225 Alan Hayward <alan.hayward@arm.com>
15226 David Sherwood <david.sherwood@arm.com>
15228 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15229 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15230 for variable-length vectors.
15231 (vectorizable_mask_load_store): Treat the number of units as
15232 polynomial, asserting that it is constant if the condition has
15233 already been enforced.
15234 (vectorizable_store, vectorizable_load): Likewise.
15236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15237 Alan Hayward <alan.hayward@arm.com>
15238 David Sherwood <david.sherwood@arm.com>
15240 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15241 of units as polynomial. Punt if we can't tell at compile time
15242 which vector contains the final result.
15244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15245 Alan Hayward <alan.hayward@arm.com>
15246 David Sherwood <david.sherwood@arm.com>
15248 * tree-vect-loop.c (vectorizable_induction): Treat the number
15249 of units as polynomial. Punt on SLP inductions. Use an integer
15250 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15251 cast of such a series for variable-length floating-point
15254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15255 Alan Hayward <alan.hayward@arm.com>
15256 David Sherwood <david.sherwood@arm.com>
15258 * tree.h (build_index_vector): Declare.
15259 * tree.c (build_index_vector): New function.
15260 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15261 of units as polynomial, forcibly converting it to a constant if
15262 vectorizable_reduction has already enforced the condition.
15263 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15264 to create a {1,2,3,...} vector.
15265 (vectorizable_reduction): Treat the number of units as polynomial.
15266 Choose vectype_in based on the largest scalar element size rather
15267 than the smallest number of units. Enforce the restrictions
15270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15271 Alan Hayward <alan.hayward@arm.com>
15272 David Sherwood <david.sherwood@arm.com>
15274 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15275 number of units as polynomial.
15277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15278 Alan Hayward <alan.hayward@arm.com>
15279 David Sherwood <david.sherwood@arm.com>
15281 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15282 * target.def (autovectorize_vector_sizes): Return the vector sizes
15283 by pointer, using vector_sizes rather than a bitmask.
15284 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15285 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15286 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15288 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15289 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15290 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15291 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15292 * omp-general.c (omp_max_vf): Likewise.
15293 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15294 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15295 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15296 * tree-vect-slp.c (vect_slp_bb): Likewise.
15297 * doc/tm.texi: Regenerate.
15298 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15300 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15301 the vector size as a poly_uint64 rather than an unsigned int.
15302 (current_vector_size): Change from an unsigned int to a poly_uint64.
15303 (get_vectype_for_scalar_type): Update accordingly.
15304 * tree.h (build_truth_vector_type): Take the size and number of
15305 units as a poly_uint64 rather than an unsigned int.
15306 (build_vector_type): Add a temporary overload that takes
15307 the number of units as a poly_uint64 rather than an unsigned int.
15308 * tree.c (make_vector_type): Likewise.
15309 (build_truth_vector_type): Take the number of units as a poly_uint64
15310 rather than an unsigned int.
15312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15313 Alan Hayward <alan.hayward@arm.com>
15314 David Sherwood <david.sherwood@arm.com>
15316 * target.def (get_mask_mode): Take the number of units and length
15317 as poly_uint64s rather than unsigned ints.
15318 * targhooks.h (default_get_mask_mode): Update accordingly.
15319 * targhooks.c (default_get_mask_mode): Likewise.
15320 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15321 * doc/tm.texi: Regenerate.
15323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15324 Alan Hayward <alan.hayward@arm.com>
15325 David Sherwood <david.sherwood@arm.com>
15327 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15328 * omp-general.c (omp_max_vf): Likewise.
15329 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15330 (expand_omp_simd): Handle polynomial safelen.
15331 * omp-low.c (omplow_simd_context): Add a default constructor.
15332 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15333 (lower_rec_simd_input_clauses): Update accordingly.
15334 (lower_rec_input_clauses): Likewise.
15336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15337 Alan Hayward <alan.hayward@arm.com>
15338 David Sherwood <david.sherwood@arm.com>
15340 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15341 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15342 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15343 (vect_analyze_slp_cost): Likewise.
15344 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15345 (vect_model_load_cost): Likewise.
15347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15348 Alan Hayward <alan.hayward@arm.com>
15349 David Sherwood <david.sherwood@arm.com>
15351 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15352 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15353 from an unsigned int * to a poly_uint64_pod *.
15354 (calculate_unrolling_factor): New function.
15355 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15358 Alan Hayward <alan.hayward@arm.com>
15359 David Sherwood <david.sherwood@arm.com>
15361 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15362 from an unsigned int to a poly_uint64.
15363 (_loop_vec_info::slp_unrolling_factor): Likewise.
15364 (_loop_vec_info::vectorization_factor): Change from an int
15366 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15367 (vect_get_num_vectors): New function.
15368 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15369 (vect_get_num_copies): Use vect_get_num_vectors.
15370 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15371 to an unsigned int *.
15372 (vect_analyze_data_refs): Change min_vf from an int * to a
15374 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15375 than an unsigned HOST_WIDE_INT.
15376 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15377 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15378 to an unsigned int *.
15379 (vect_analyze_data_ref_dependences): Likewise.
15380 (vect_compute_data_ref_alignment): Handle polynomial vf.
15381 (vect_enhance_data_refs_alignment): Likewise.
15382 (vect_prune_runtime_alias_test_list): Likewise.
15383 (vect_shift_permute_load_chain): Likewise.
15384 (vect_supportable_dr_alignment): Likewise.
15385 (dependence_distance_ge_vf): Take the vectorization factor as a
15386 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15387 (vect_analyze_data_refs): Change min_vf from an int * to a
15389 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15390 vfm1 as a poly_uint64 rather than an int. Make the same change
15391 for the returned bound_scalar.
15392 (vect_gen_vector_loop_niters): Handle polynomial vf.
15393 (vect_do_peeling): Likewise. Update call to
15394 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15395 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15397 * tree-vect-loop.c (vect_determine_vectorization_factor)
15398 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15399 (vect_get_known_peeling_cost): Likewise.
15400 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15401 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15402 (vect_transform_loop): Likewise. Use the lowest possible VF when
15403 updating the upper bounds of the loop.
15404 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15405 rather than an int.
15406 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15407 polynomial unroll factors.
15408 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15409 (vect_make_slp_decision): Likewise.
15410 (vect_supported_load_permutation_p): Likewise, and polynomial
15412 (vect_analyze_slp_cost): Handle polynomial vf.
15413 (vect_slp_analyze_node_operations): Likewise.
15414 (vect_slp_analyze_bb_1): Likewise.
15415 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15416 than an unsigned HOST_WIDE_INT.
15417 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15418 (vectorizable_load): Handle polynomial vf.
15419 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15421 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15424 Alan Hayward <alan.hayward@arm.com>
15425 David Sherwood <david.sherwood@arm.com>
15427 * match.pd: Handle bit operations involving three constants
15428 and try to fold one pair.
15430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15432 * tree-vect-loop-manip.c: Include gimple-fold.h.
15433 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15434 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15435 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15436 Add a path that uses a step of VF instead of 1, but disable it
15438 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15439 and niters_no_overflow parameters. Update calls to
15440 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15441 Create a new SSA name if the latter choses to use a ste other
15442 than zero, and return it via niters_vector_mult_vf_var.
15443 * tree-vect-loop.c (vect_transform_loop): Update calls to
15444 vect_do_peeling, vect_gen_vector_loop_niters and
15445 slpeel_make_loop_iterate_ntimes.
15446 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15447 (vect_gen_vector_loop_niters): Update declarations after above changes.
15449 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15451 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15452 128-bit round to integer instructions.
15453 (ceil<mode>2): Likewise.
15454 (btrunc<mode>2): Likewise.
15455 (round<mode>2): Likewise.
15457 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15459 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15460 unaligned VSX load/store on P8/P9.
15461 (expand_block_clear): Allow the use of unaligned VSX
15462 load/store on P8/P9.
15464 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15466 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15468 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15469 swap associated with both a load and a store.
15471 2018-01-02 Andrew Waterman <andrew@sifive.com>
15473 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15474 * config/riscv/riscv.md (clear_cache): Use it.
15476 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15478 * web.c: Remove out-of-date comment.
15480 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15482 * expr.c (fixup_args_size_notes): Check that any existing
15483 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15484 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15485 (emit_single_push_insn): ...here.
15487 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15489 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15490 (const_vector_encoded_nelts): New function.
15491 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15492 (const_vector_int_elt, const_vector_elt): Declare.
15493 * emit-rtl.c (const_vector_int_elt_1): New function.
15494 (const_vector_elt): Likewise.
15495 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15496 of CONST_VECTOR_ELT.
15498 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15500 * expr.c: Include rtx-vector-builder.h.
15501 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15502 directly on the tree encoding.
15503 (const_vector_from_tree): Likewise.
15504 * optabs.c: Include rtx-vector-builder.h.
15505 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15506 sequence of "u" values.
15507 * vec-perm-indices.c: Include rtx-vector-builder.h.
15508 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15509 directly on the vec_perm_indices encoding.
15511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15513 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15514 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15515 * rtx-vector-builder.h: New file.
15516 * rtx-vector-builder.c: Likewise.
15517 * rtl.h (rtx_def::u2): Add a const_vector field.
15518 (CONST_VECTOR_NPATTERNS): New macro.
15519 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15520 (CONST_VECTOR_DUPLICATE_P): Likewise.
15521 (CONST_VECTOR_STEPPED_P): Likewise.
15522 (CONST_VECTOR_ENCODED_ELT): Likewise.
15523 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15524 (unwrap_const_vec_duplicate): Likewise.
15525 (const_vec_series_p): Check for a non-duplicated vector encoding.
15526 Say that the function only returns true for integer vectors.
15527 * emit-rtl.c: Include rtx-vector-builder.h.
15528 (gen_const_vec_duplicate_1): Delete.
15529 (gen_const_vector): Call gen_const_vec_duplicate instead of
15530 gen_const_vec_duplicate_1.
15531 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15532 (gen_const_vec_duplicate): Use rtx_vector_builder.
15533 (gen_const_vec_series): Likewise.
15534 (gen_rtx_CONST_VECTOR): Likewise.
15535 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15536 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15537 Build a new vector rather than modifying a CONST_VECTOR in-place.
15538 (handle_special_swappables): Update call accordingly.
15539 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15540 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15541 Build a new vector rather than modifying a CONST_VECTOR in-place.
15542 (handle_special_swappables): Update call accordingly.
15544 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15546 * simplify-rtx.c (simplify_const_binary_operation): Use
15547 CONST_VECTOR_ELT instead of XVECEXP.
15549 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15551 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15552 the selector elements to be different from the data elements
15553 if the selector is a VECTOR_CST.
15554 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15555 ssizetype for the selector.
15557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15559 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15560 before testing each element individually.
15561 * tree-vect-generic.c (lower_vec_perm): Likewise.
15563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15565 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15566 * selftest-run-tests.c (selftest::run_tests): Call it.
15567 * vector-builder.h (vector_builder::operator ==): New function.
15568 (vector_builder::operator !=): Likewise.
15569 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15570 (vec_perm_indices::all_from_input_p): New function.
15571 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15572 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15573 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15574 instead of reading the VECTOR_CST directly. Detect whether both
15575 vector inputs are the same before constructing the vec_perm_indices,
15576 and update the number of inputs argument accordingly. Use the
15577 utility functions added above. Only construct sel2 if we need to.
15579 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15581 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15582 the broadcast of the low byte.
15583 (expand_mult_highpart): Use an explicit encoding for the permutes.
15584 * optabs-query.c (can_mult_highpart_p): Likewise.
15585 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15586 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15587 (vectorizable_bswap): Likewise.
15588 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15589 explicit encoding for the power-of-2 permutes.
15590 (vect_permute_store_chain): Likewise.
15591 (vect_grouped_load_supported): Likewise.
15592 (vect_permute_load_chain): Likewise.
15594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15596 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15597 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15598 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15599 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15600 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15601 (vect_gen_perm_mask_any): Likewise.
15603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15605 * int-vector-builder.h: New file.
15606 * vec-perm-indices.h: Include int-vector-builder.h.
15607 (vec_perm_indices): Redefine as an int_vector_builder.
15608 (auto_vec_perm_indices): Delete.
15609 (vec_perm_builder): Redefine as a stand-alone class.
15610 (vec_perm_indices::vec_perm_indices): New function.
15611 (vec_perm_indices::clamp): Likewise.
15612 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15613 (vec_perm_indices::new_vector): New function.
15614 (vec_perm_indices::new_expanded_vector): Update for new
15615 vec_perm_indices class.
15616 (vec_perm_indices::rotate_inputs): New function.
15617 (vec_perm_indices::all_in_range_p): Operate directly on the
15618 encoded form, without computing elided elements.
15619 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15620 encoding. Update for new vec_perm_indices class.
15621 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15622 the given vec_perm_builder.
15623 (expand_vec_perm_var): Update vec_perm_builder constructor.
15624 (expand_mult_highpart): Use vec_perm_builder instead of
15625 auto_vec_perm_indices.
15626 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15627 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15628 or double series encoding as appropriate.
15629 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15630 vec_perm_indices instead of auto_vec_perm_indices.
15631 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15632 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15633 (vect_permute_store_chain): Likewise.
15634 (vect_grouped_load_supported): Likewise.
15635 (vect_permute_load_chain): Likewise.
15636 (vect_shift_permute_load_chain): Likewise.
15637 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15638 (vect_transform_slp_perm_load): Likewise.
15639 (vect_schedule_slp_instance): Likewise.
15640 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15641 (vectorizable_mask_load_store): Likewise.
15642 (vectorizable_bswap): Likewise.
15643 (vectorizable_store): Likewise.
15644 (vectorizable_load): Likewise.
15645 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15646 vec_perm_indices instead of auto_vec_perm_indices. Use
15647 tree_to_vec_perm_builder to read the vector from a tree.
15648 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15649 vec_perm_builder instead of a vec_perm_indices.
15650 (have_whole_vector_shift): Use vec_perm_builder and
15651 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15652 truncation to calc_vec_perm_mask_for_shift.
15653 (vect_create_epilog_for_reduction): Likewise.
15654 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15655 from auto_vec_perm_indices to vec_perm_indices.
15656 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15657 instead of changing individual elements.
15658 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15659 the vector in d.perm.
15660 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15661 from auto_vec_perm_indices to vec_perm_indices.
15662 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15663 instead of changing individual elements.
15664 (arm_vectorize_vec_perm_const): Use new_vector to install
15665 the vector in d.perm.
15666 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15667 Update vec_perm_builder constructor.
15668 (rs6000_expand_interleave): Likewise.
15669 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15670 (rs6000_expand_interleave): Likewise.
15672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15674 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15675 to qimode could truncate the indices.
15676 * optabs.c (expand_vec_perm_var): Likewise.
15678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15680 * Makefile.in (OBJS): Add vec-perm-indices.o.
15681 * vec-perm-indices.h: New file.
15682 * vec-perm-indices.c: Likewise.
15683 * target.h (vec_perm_indices): Replace with a forward class
15685 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15686 * optabs.h: Include vec-perm-indices.h.
15687 (expand_vec_perm): Delete.
15688 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15689 (expand_vec_perm_const): Declare.
15690 * target.def (vec_perm_const_ok): Replace with...
15691 (vec_perm_const): ...this new hook.
15692 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15693 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15694 * doc/tm.texi: Regenerate.
15695 * optabs.def (vec_perm_const): Delete.
15696 * doc/md.texi (vec_perm_const): Likewise.
15697 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15698 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15699 expand_vec_perm for constant permutation vectors. Assert that
15700 the mode of variable permutation vectors is the integer equivalent
15701 of the mode that is being permuted.
15702 * optabs-query.h (selector_fits_mode_p): Declare.
15703 * optabs-query.c: Include vec-perm-indices.h.
15704 (selector_fits_mode_p): New function.
15705 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15706 is defined, instead of checking whether the vec_perm_const_optab
15707 exists. Use targetm.vectorize.vec_perm_const instead of
15708 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15709 fit in the vector mode before using a variable permute.
15710 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15711 vec_perm_indices instead of an rtx.
15712 (expand_vec_perm): Replace with...
15713 (expand_vec_perm_const): ...this new function. Take the selector
15714 as a vec_perm_indices rather than an rtx. Also take the mode of
15715 the selector. Update call to shift_amt_for_vec_perm_mask.
15716 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15717 Use vec_perm_indices::new_expanded_vector to expand the original
15718 selector into bytes. Check whether the indices fit in the vector
15719 mode before using a variable permute.
15720 (expand_vec_perm_var): Make global.
15721 (expand_mult_highpart): Use expand_vec_perm_const.
15722 * fold-const.c: Includes vec-perm-indices.h.
15723 * tree-ssa-forwprop.c: Likewise.
15724 * tree-vect-data-refs.c: Likewise.
15725 * tree-vect-generic.c: Likewise.
15726 * tree-vect-loop.c: Likewise.
15727 * tree-vect-slp.c: Likewise.
15728 * tree-vect-stmts.c: Likewise.
15729 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15731 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15732 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15733 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15734 (aarch64_vectorize_vec_perm_const): ...this new function.
15735 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15736 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15737 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15738 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15739 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15740 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15741 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15743 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15744 check for NEON modes.
15745 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15746 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15747 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15748 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15750 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15751 the old VEC_PERM_CONST conditions.
15752 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15753 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15754 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15755 (ia64_vectorize_vec_perm_const_ok): Merge into...
15756 (ia64_vectorize_vec_perm_const): ...this new function.
15757 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15758 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15759 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15760 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15761 * config/mips/mips.c (mips_expand_vec_perm_const)
15762 (mips_vectorize_vec_perm_const_ok): Merge into...
15763 (mips_vectorize_vec_perm_const): ...this new function.
15764 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15765 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15766 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15767 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15768 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15769 (rs6000_expand_vec_perm_const): Delete.
15770 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15772 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15773 (altivec_expand_vec_perm_const_le): Take each operand individually.
15774 Operate on constant selectors rather than rtxes.
15775 (altivec_expand_vec_perm_const): Likewise. Update call to
15776 altivec_expand_vec_perm_const_le.
15777 (rs6000_expand_vec_perm_const): Delete.
15778 (rs6000_vectorize_vec_perm_const_ok): Delete.
15779 (rs6000_vectorize_vec_perm_const): New function.
15780 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15781 an element count and rtx array.
15782 (rs6000_expand_extract_even): Update call accordingly.
15783 (rs6000_expand_interleave): Likewise.
15784 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15785 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15786 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15787 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15788 (rs6000_expand_vec_perm_const): Delete.
15789 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15790 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15791 (altivec_expand_vec_perm_const_le): Take each operand individually.
15792 Operate on constant selectors rather than rtxes.
15793 (altivec_expand_vec_perm_const): Likewise. Update call to
15794 altivec_expand_vec_perm_const_le.
15795 (rs6000_expand_vec_perm_const): Delete.
15796 (rs6000_vectorize_vec_perm_const_ok): Delete.
15797 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15798 reference to the SPE evmerge intructions.
15799 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15800 an element count and rtx array.
15801 (rs6000_expand_extract_even): Update call accordingly.
15802 (rs6000_expand_interleave): Likewise.
15803 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15804 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15806 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15810 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15811 vector mode and that that mode matches the mode of the data
15813 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15814 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15815 directly using expand_vec_perm_1 when forcing selectors into
15817 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15819 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15821 * optabs-query.h (can_vec_perm_p): Delete.
15822 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15823 * optabs-query.c (can_vec_perm_p): Split into...
15824 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15825 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15826 particular selector is valid.
15827 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15828 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15829 (vect_grouped_load_supported): Likewise.
15830 (vect_shift_permute_load_chain): Likewise.
15831 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15832 (vect_transform_slp_perm_load): Likewise.
15833 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15834 (vectorizable_bswap): Likewise.
15835 (vect_gen_perm_mask_checked): Likewise.
15836 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15837 implementations of variable permutation vectors into account
15838 when deciding which selector to use.
15839 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15840 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15841 with a false third argument.
15842 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15843 to test whether the constant selector is valid and can_vec_perm_var_p
15844 to test whether a variable selector is valid.
15846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15848 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15849 * optabs-query.c (can_vec_perm_p): Likewise.
15850 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15851 instead of vec_perm_indices.
15852 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15853 (vect_gen_perm_mask_checked): Likewise,
15854 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15855 (vect_gen_perm_mask_checked): Likewise,
15857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15859 * optabs-query.h (qimode_for_vec_perm): Declare.
15860 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15861 (qimode_for_vec_perm): ...this new function.
15862 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15864 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15866 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15867 does not have a conditional at the top.
15869 2018-01-02 Richard Biener <rguenther@suse.de>
15871 * ipa-inline.c (big_speedup_p): Fix expression.
15873 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15876 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15879 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15883 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15884 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15885 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15886 cond_taken_branch_cost 3->4.
15888 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15890 PR tree-optimization/83581
15891 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15892 TODO_cleanup_cfg if any changes have been made.
15894 PR middle-end/83608
15895 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15896 convert_modes if target mode has the right side, but different mode
15899 PR middle-end/83609
15900 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15901 last argument when extracting from CONCAT. If either from_real or
15902 from_imag is NULL, use expansion through memory. If result is not
15903 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15904 the parts directly to inner mode, if even that fails, use expansion
15907 PR middle-end/83623
15908 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15909 check for bswap in mode rather than HImode and use that in expand_unop
15912 Copyright (C) 2018 Free Software Foundation, Inc.
15914 Copying and distribution of this file, with or without modification,
15915 are permitted in any medium without royalty provided the copyright
15916 notice and this notice are preserved.