1 2018-06-20 Maya Rashish <coypu@sdf.org>
3 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
5 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8 * config/aarch64/aarch64.c (xgene1_tunings): Add
9 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10 (aarch64_mode_valid_for_sched_fusion_p):
12 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
13 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
17 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
18 * config/aarch64/iterators.md (VQ2): New mode iterator.
20 2018-06-20 Martin Liska <mliska@suse.cz>
22 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
23 Change default ratio from 10 to 8.
25 2018-06-20 Martin Liska <mliska@suse.cz>
27 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
29 (bit_test_cluster::find_bit_tests): Likewise.
30 (switch_decision_tree::analyze_switch_statement): Find clusters.
31 * tree-switch-conversion.h (struct jump_table_cluster): Document
34 2018-06-20 Martin Liska <mliska@suse.cz>
36 * tree-switch-conversion.c (switch_conversion::collect):
37 Record m_uniq property.
38 (switch_conversion::expand): Bail out for special conditions.
39 (group_cluster::~group_cluster): New.
40 (group_cluster::group_cluster): Likewise.
41 (group_cluster::dump): Likewise.
42 (jump_table_cluster::emit): New.
43 (switch_decision_tree::fix_phi_operands_for_edges): New.
44 (struct case_node): Remove struct.
45 (jump_table_cluster::can_be_handled): New.
46 (case_values_threshold): Moved to header.
47 (reset_out_edges_aux): Likewise.
48 (jump_table_cluster::is_beneficial): New.
49 (bit_test_cluster::can_be_handled): Likewise.
50 (add_case_node): Remove.
51 (bit_test_cluster::is_beneficial): New.
52 (case_bit_test::cmp): New.
53 (bit_test_cluster::emit): New.
54 (expand_switch_as_decision_tree_p): Remove.
55 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
56 (fix_phi_operands_for_edge): Likewise.
57 (switch_decision_tree::analyze_switch_statement): New.
58 (compute_cases_per_edge): Move ...
59 (switch_decision_tree::compute_cases_per_edge): ... here.
60 (try_switch_expansion): Likewise.
61 (switch_decision_tree::try_switch_expansion): Likewise.
62 (record_phi_operand_mapping): Likewise.
63 (switch_decision_tree::record_phi_operand_mapping): Likewise.
64 (emit_case_decision_tree): Likewise.
65 (switch_decision_tree::emit): Likewise.
66 (balance_case_nodes): Likewise.
67 (switch_decision_tree::balance_case_nodes): Likewise.
68 (dump_case_nodes): Likewise.
69 (switch_decision_tree::dump_case_nodes): Likewise.
70 (emit_jump): Likewise.
71 (switch_decision_tree::emit_jump): Likewise.
72 (emit_cmp_and_jump_insns): Likewise.
73 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
74 (emit_case_nodes): Likewise.
75 (switch_decision_tree::emit_case_nodes): Likewise.
76 (conditional_probability): Remove.
77 * tree-switch-conversion.h (enum cluster_type): New.
79 (struct cluster): Likewise.
80 (cluster::cluster): Likewise.
81 (struct simple_cluster): Likewise.
82 (simple_cluster::simple_cluster): Likewise.
83 (struct group_cluster): Likewise.
84 (struct jump_table_cluster): Likewise.
85 (struct bit_test_cluster): Likewise.
86 (struct min_cluster_item): Likewise.
87 (struct case_tree_node): Likewise.
88 (case_tree_node::case_tree_node): Likewise.
89 (jump_table_cluster::case_values_threshold): Likewise.
90 (struct case_bit_test): Likewise.
91 (struct switch_decision_tree): Likewise.
92 (struct switch_conversion): Likewise.
93 (switch_decision_tree::reset_out_edges_aux): Likewise.
95 2018-06-20 Martin Liska <mliska@suse.cz>
97 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
98 (hoist_edge_and_branch_if_true): Likewise.
99 (expand_switch_using_bit_tests_p): Likewise.
100 (struct case_bit_test): Likewise.
101 (case_bit_test_cmp): Likewise.
102 (emit_case_bit_tests): Likewise.
103 (switch_conversion::switch_conversion): New class.
104 (struct switch_conv_info): Remove old struct.
105 (collect_switch_conv_info): More to ...
106 (switch_conversion::collect): ... this.
107 (check_range): Likewise.
108 (switch_conversion::check_range): Likewise.
109 (check_all_empty_except_final): Likewise.
110 (switch_conversion::check_all_empty_except_final): Likewise.
111 (check_final_bb): Likewise.
112 (switch_conversion::check_final_bb): Likewise.
113 (create_temp_arrays): Likewise.
114 (switch_conversion::create_temp_arrays): Likewise.
115 (free_temp_arrays): Likewise.
116 (gather_default_values): Likewise.
117 (switch_conversion::gather_default_values): Likewise.
118 (build_constructors): Likewise.
119 (switch_conversion::build_constructors): Likewise.
120 (constructor_contains_same_values_p): Likewise.
121 (switch_conversion::contains_same_values_p): Likewise.
122 (array_value_type): Likewise.
123 (switch_conversion::array_value_type): Likewise.
124 (build_one_array): Likewise.
125 (switch_conversion::build_one_array): Likewise.
126 (build_arrays): Likewise.
127 (switch_conversion::build_arrays): Likewise.
128 (gen_def_assigns): Likewise.
129 (switch_conversion::gen_def_assigns): Likewise.
130 (prune_bbs): Likewise.
131 (switch_conversion::prune_bbs): Likewise.
132 (fix_phi_nodes): Likewise.
133 (switch_conversion::fix_phi_nodes): Likewise.
134 (gen_inbound_check): Likewise.
135 (switch_conversion::gen_inbound_check): Likewise.
136 (process_switch): Use the newly created class.
137 (switch_conversion::expand): New.
138 (switch_conversion::~switch_conversion): New.
139 * tree-switch-conversion.h: New file.
141 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
143 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
144 tree-vect-patterns.c.
145 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
146 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
147 (vect_recog_sad_pattern): Likewise.
148 (vect_recog_widen_sum_pattern): Likewise.
149 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
150 (vect_recog_widen_shift_pattern): Remove the type_in argument.
151 (vect_recog_rotate_pattern): Likewise.
152 (vect_recog_mult_pattern): Likewise.
153 (vect_recog_vector_vector_shift_pattern): Likewise.
154 (vect_recog_divmod_pattern): Likewise.
155 (vect_recog_mixed_size_cond_pattern): Likewise.
156 (vect_recog_bool_pattern): Likewise.
157 (vect_recog_mask_conversion_pattern): Likewise.
158 (vect_try_gather_scatter_pattern): Likewise.
159 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
160 (vect_recog_over_widening_pattern): Likewise.
161 (vect_recog_gather_scatter_pattern): Likewise.
162 (vect_recog_func_ptr): Move from tree-vectorizer.h
163 (vect_vect_recog_func_ptrs): Move further down the file.
164 (vect_recog_func): Likewise. Remove the third argument.
165 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
166 (vect_pattern_recog_1): Expect the pattern function to do any
167 necessary target tests. Also expect it to provide a vector type.
168 Remove the type_in handling.
170 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
172 * tree-vect-patterns.c (vect_pattern_detected): New function.
173 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
174 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
175 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
176 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
177 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
178 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
179 (vect_recog_mask_conversion_pattern)
180 (vect_try_gather_scatter_pattern): Likewise.
182 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
184 * tree-vect-patterns.c (vect_get_internal_def): New function.
185 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
186 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
187 (search_type_for_mask_1): Use it.
189 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
191 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
192 redundant WIDEN_SUM_EXPR handling.
193 (vect_recog_sad_pattern): Likewise.
195 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
197 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
198 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
199 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
200 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
201 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
203 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
205 * tree-vect-stmts.c (vectorizable_call): Make sure that we
206 use the stmt_vec_info of the original bb statement for the
207 new zero assignment, even if the call is part of a pattern.
209 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
211 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
212 that the sequence is attached to the original statement rather
213 than the pattern statement.
214 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
215 PATTERN_DEF_SEQ from the original statement rather than
216 the main pattern statement.
217 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
218 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
219 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
221 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
223 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
224 definition statements before the early exit for statements that aren't
226 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
228 (vect_transform_loop): ...here. Process pattern definition
229 statements without first checking whether the main pattern
230 statement is live or relevant.
232 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
234 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
235 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
237 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
239 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
240 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
241 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
242 (expand_block_compare): Change select_block_compare_mode call.
243 (expand_strncmp_align_check): Use new functions, fix comment.
244 (emit_final_str_compare_gpr): New function.
245 (expand_strn_compare): Refactor and clean up code.
246 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
248 2018-06-19 Tony Reix <tony.reix@atos.com>
249 Damien Bergamini <damien.bergamini@atos.com>
250 David Edelsohn <dje.gcc@gmail.com>
252 * collect2.c (static_obj): New variable.
253 (static_libs): New variable.
254 (is_in_list): Uncomment declaration.
255 (main): Track AIX libraries linked statically.
256 (is_in_list): Uncomment definition.
257 (scan_prog_file): Don't add AIX shared libraries initializer
258 to constructor list if linking statically.
260 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
262 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
264 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
266 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
268 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
271 2018-06-19 Martin Liska <mliska@suse.cz>
273 * config/i386/i386.c (ix86_can_inline_p): Do not use
274 ipa_fn_summaries::get_create.
275 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
277 (devirtualization_time_bonus): Likewise.
278 (ipcp_propagate_stage): Likewise.
279 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
280 (edge_set_predicate): Likewise.
281 (evaluate_conditions_for_known_args): Likewise.
282 (evaluate_properties_for_edge): Likewise.
283 (ipa_call_summary::reset): Tranform to ...
284 (ipa_call_summary::~ipa_call_summary): ... this.
285 (ipa_fn_summary::reset): Transform to ...
286 (ipa_fn_summary::~ipa_fn_summary): ... this.
287 (ipa_fn_summary_t::remove): Rename to ...
288 (ipa_fn_summary_t::remove_callees): ... this.
289 (ipa_fn_summary_t::duplicate): Use placement new
290 instead of memory copy.
291 (ipa_call_summary_t::duplicate): Likewise.
292 (ipa_call_summary_t::remove): Remove.
293 (dump_ipa_call_summary): Change get_create to get.
294 (ipa_dump_fn_summary): Dump only when summary exists.
295 (analyze_function_body): Use symbol_summary::get instead
297 (compute_fn_summary): Likewise.
298 (estimate_edge_devirt_benefit): Likewise.
299 (estimate_edge_size_and_time): Likewise.
300 (inline_update_callee_summaries): Likewise.
301 (remap_edge_change_prob): Likewise.
302 (remap_edge_summaries): Likewise.
303 (ipa_merge_fn_summary_after_inlining): Likewise.
304 (write_ipa_call_summary): Likewise.
305 (ipa_fn_summary_write): Likewise.
306 (ipa_free_fn_summary): Likewise.
307 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
308 (struct ipa_call_summary): Likewise.
309 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
311 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
312 (estimate_size_after_inlining): Likewise.
313 (estimate_growth): Likewise.
314 (growth_likely_positive): Likewise.
315 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
316 (inline_call): Likewise.
317 * ipa-inline.c (caller_growth_limits): Likewise.
318 (can_inline_edge_p): Likewise.
319 (can_inline_edge_by_limits_p): Likewise.
320 (compute_uninlined_call_time): Likewise.
321 (compute_inlined_call_time): Likewise.
322 (want_inline_small_function_p): Likewise.
323 (edge_badness): Likewise.
324 (update_caller_keys): Likewise.
325 (update_callee_keys): Likewise.
326 (inline_small_functions): Likewise.
327 (inline_to_all_callers_1): Likewise.
328 (dump_overall_stats): Likewise.
329 (early_inline_small_functions): Likewise.
330 (early_inliner): Likewise.
331 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
332 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
333 * ipa-pure-const.c (malloc_candidate_p): Likewise.
334 * ipa-split.c (execute_split_functions): Likewise.
335 * symbol-summary.h: Likewise.
336 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
338 2018-06-19 Richard Biener <rguenther@suse.de>
340 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
341 (vectorize_loops): ... here. Fix dbgcnt handling.
342 (try_vectorize_loop): Wrap try_vectorize_loop_1.
344 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
347 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
348 ieee128 argument takes up only one (vector) register, not two (floating
351 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
353 * gimplify.c (gimplify_init_constructor): Really never clear for an
354 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
356 2018-06-19 Richard Biener <rguenther@suse.de>
358 PR tree-optimization/86179
359 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
360 after failed recognition.
362 2018-06-18 Martin Sebor <msebor@redhat.com>
365 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
366 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
367 Handle integer subtraction.
368 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
369 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
371 2018-06-18 David Malcolm <dmalcolm@redhat.com>
373 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
374 param from rtx to rtx_insn *.
375 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
377 (frv_ifcvt_modify_insn): Likwise.
378 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
379 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
380 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
381 as_a <rtx_insn *> cast to local "unprotected_region" once
382 it's been established that it's not NULL or pc_rtx.
383 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
384 param "sethi" from rtx to rtx_insn *.
385 (nds32_group_float_insns): Likewise for param "insn".
386 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
388 (vax_output_int_subtract): Likewise.
389 * config/vax/vax.c (vax_output_int_add): Likewise for param
391 (vax_output_int_subtract): Likewise.
392 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
393 (emit_pattern_after): Likewise for param "after".
394 (emit_insn_after): Likewise.
395 (emit_jump_insn_after): Likewise.
396 (emit_call_insn_after): Likewise.
397 (emit_debug_insn_after): Likewise.
398 (emit_pattern_before): Likewise for param "before".
399 (emit_insn_before): Likewise.
400 (emit_jump_insn_before): Likewise.
401 * final.c (get_insn_template): Likewise for param "insn", removing
403 * output.h (get_insn_template): Likewise for 2nd param.
404 * rtl.h (emit_insn_before): Likewise.
405 (emit_jump_insn_before): Likewise.
406 (emit_debug_insn_before_noloc): Likewise.
407 (emit_insn_after): Likewise.
408 (emit_jump_insn_after): Likewise.
409 (emit_call_insn_after): Likewise.
410 (emit_debug_insn_after): Likewise.
411 (set_insn_deleted): Likewise for param.
413 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
416 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
417 floating point modes, so that IFmode is numerically greater than
418 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
419 to declare the ordering. This prevents IFmode from being
420 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
421 machine. Include rs6000-modes.h to share the fractional values
422 between genmodes* and the rest of the compiler.
426 * config/rs6000/rs6000-modes.h: New file.
427 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
428 meaning of rs6000_long_double_size so that 126..128 selects an
429 appropriate 128-bit floating point type.
430 (rs6000_option_override_internal): Likewise.
431 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
432 (TARGET_LONG_DOUBLE_128): Change the meaning of
433 rs6000_long_double_size so that 126..128 selects an appropriate
434 128-bit floating point type.
435 (LONG_DOUBLE_TYPE_SIZE): Update comment.
436 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
437 source and destination to match the standard usage.
438 (truncifkf2): Likewise.
439 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
440 ISA 2.07 to use an explicit clobber, instead of passing in a
442 (copysign<mode>3_soft): Likewise.
444 2018-06-18 David Malcolm <dmalcolm@redhat.com>
446 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
447 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
448 (vect_slp_analyze_instance_dependence): Likewise.
449 (vect_enhance_data_refs_alignment): Likewise.
450 (vect_analyze_data_refs_alignment): Likewise.
451 (vect_slp_analyze_and_verify_instance_alignment
452 (vect_analyze_data_ref_accesses): Likewise.
453 (vect_prune_runtime_alias_test_list): Likewise.
454 (vect_analyze_data_refs): Likewise.
455 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
456 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
457 (vect_analyze_scalar_cycles_1): Likewise.
458 (vect_get_loop_niters): Likewise.
459 (vect_analyze_loop_form_1): Likewise.
460 (vect_update_vf_for_slp): Likewise.
461 (vect_analyze_loop_operations): Likewise.
462 (vect_analyze_loop): Likewise.
463 (vectorizable_induction): Likewise.
464 (vect_transform_loop): Likewise.
465 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
466 * tree-vect-slp.c (vect_analyze_slp): Likewise.
467 (vect_make_slp_decision): Likewise.
468 (vect_detect_hybrid_slp): Likewise.
469 (vect_slp_analyze_operations): Likewise.
470 (vect_slp_bb): Likewise.
471 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
472 (vectorizable_bswap): Likewise.
473 (vectorizable_call): Likewise.
474 (vectorizable_simd_clone_call): Likewise.
475 (vectorizable_conversion): Likewise.
476 (vectorizable_assignment): Likewise.
477 (vectorizable_shift): Likewise.
478 (vectorizable_operation): Likewise.
479 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
481 2018-06-18 Martin Sebor <msebor@redhat.com>
483 PR tree-optimization/81384
484 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
485 * builtins.c (expand_builtin_strnlen): New function.
486 (expand_builtin): Call it.
487 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
488 * builtins.def (BUILT_IN_STRNLEN): New.
489 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
490 Warn for bounds in excess of maximum object size.
491 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
492 single-value ranges. Handle strnlen.
493 (handle_builtin_strlen): Handle strnlen.
494 (strlen_check_and_optimize_stmt): Same.
495 * doc/extend.texi (Other Builtins): Document strnlen.
497 2018-06-18 Maya Rashish <coypu@sdf.org>
499 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
500 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
501 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
503 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
505 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
507 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
509 * tree.c (escaped_string::escape): Replace cast to char * by
510 const_cast<char *> (unescaped).
512 2018-06-18 Nick Clifton <nickc@redhat.com>
515 * tree.c (escaped_string): New class. Converts an unescaped
516 string into its escaped equivalent.
517 (warn_deprecated_use): Use the new class to convert the
518 deprecation message, if present.
519 (test_escaped_strings): New self test.
520 (test_c_tests): Add test_escaped_strings.
521 * doc/extend.texi (deprecated): Add a note that the
522 deprecation message is affected by the -fmessage-length
523 option, and that control characters will be escaped.
524 (#pragma GCC error): Document this pragma.
525 (#pragma GCC warning): Likewise.
526 * doc/invoke.texi (-fmessage-length): Document this option's
527 effect on the #warning and #error preprocessor directives and
528 the deprecated attribute.
530 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
532 * tree.c (decl_value_expr_lookup): Revert latest change.
533 (decl_value_expr_insert): Likewise.
535 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
537 * gimplify.c (nonlocal_vlas): Delete.
538 (nonlocal_vla_vars): Likewise.
539 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
541 (gimplify_body): Do not create and destroy nonlocal_vlas.
542 * tree-nested.c: Include diagnostic.h.
543 (use_pointer_in_frame): Tweak.
544 (lookup_field_for_decl): Add assertion and declare the transformation.
545 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
546 internal error when the reference is in a wrong context. Do not
547 create a debug decl by default.
548 (note_nonlocal_block_vlas): Delete.
549 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
550 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
551 create a debug decl by default.
552 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
553 call is in a wrong context.
554 (fixup_vla_decls): New function.
555 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
556 debug variables were created.
557 * tree.c (decl_value_expr_lookup): Add checking assertion.
558 (decl_value_expr_insert): Likewise.
560 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
563 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
564 * tree-scalar-evolution.c (interpret_expr): Likewise.
565 (expression_expensive_p): Likewise.
566 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
567 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
568 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
569 (ssa_defined_by_minus_one_stmt_p): New.
571 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
574 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
575 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
576 * dojump.c (do_jump): Likewise.
577 * expr.c (expand_expr_real_2): Check operand type's sign.
578 * fold-const.c (const_unop): Handle ABSU_EXPR.
579 (fold_abs_const): Likewise.
580 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
581 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
582 (strip_sign_op_1): Likesise.
583 * match.pd: Add new pattern to generate ABSU_EXPR.
584 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
585 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
586 * tree-eh.c (operation_could_trap_helper_p): Likewise.
587 * tree-inline.c (estimate_operator_cost): Likewise.
588 * tree-pretty-print.c (dump_generic_node): Likewise.
589 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
590 * tree.def (ABSU_EXPR): New.
592 2018-06-16 Jakub Jelinek <jakub@redhat.com>
595 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
596 documented as preserved for backward compatibility only.
597 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
599 PR rtl-optimization/86108
600 * bb-reorder.c (create_forwarder_block): Renamed to ...
601 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
602 jump from new landing pad to the second part.
603 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
606 2018-06-15 Jakub Jelinek <jakub@redhat.com>
609 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
610 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
611 Only call store_expr for halves if the mode is the same.
614 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
615 Fix up comment formatting.
617 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
619 * typed-splay-tree.h (typed_splay_tree::remove): New function.
620 (typed_splay_tree::closure,
621 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
622 (typed_splay_tree::typed_splay_tree,
623 typed_splay_tree::operator =): Declared private.
624 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
625 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
626 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
627 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
628 typed_splay_tree::splay_tree_splay,
629 typed_splay_tree::splay_tree_foreach_helper,
630 typed_splay_tree::splay_tree_insert,
631 typed_splay_tree::splay_tree_remove,
632 typed_splay_tree::splay_tree_lookup,
633 typed_splay_tree::splay_tree_predecessor,
634 typed_splay_tree::splay_tree_successor,
635 typed_splay_tree::splay_tree_min,
636 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
637 (typed_splay_tree::root, typed_splay_tree::comp,
638 typed_splay_tree::delete_key,
639 typed_splay_tree::delete_value): New data members.
640 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
641 typed_splay_tree::remove.
643 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
645 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
646 -mginv and -mno-ginv to the assembler.
647 * config/mips/mips.opt (-mcrc): New option.
649 * doc/invoke.text (-mcrc): Document.
652 2018-06-15 Nick Clifton <nickc@redhat.com>
655 * tree.c (escaped_string): New class. Converts an unescaped
656 string into its escaped equivalent.
657 (warn_deprecated_use): Use the new class to convert the
658 deprecation message, if present.
659 (test_escaped_strings): New self test.
660 (test_c_tests): Add test_escaped_strings.
661 * doc/extend.texi (deprecated): Add a note that the
662 deprecation message is affected by the -fmessage-length
663 option, and that control characters will be escaped.
664 (#pragma GCC error): Document this pragma.
665 (#pragma GCC warning): Likewise.
666 * doc/invoke.texi (-fmessage-length): Document this option's
667 effect on the #warning and #error preprocessor directives and
668 the deprecated attribute.
670 2018-06-15 Richard Biener <rguenther@suse.de>
672 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
673 here, also noting vector size used.
674 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
675 size used in MSG_OPTIMIZED_LOCATIONS dump.
676 (pass_slp_vectorize::execute): Adjust.
678 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
681 * config/arc/arc.c (arc_return_address_register): Fix
684 2018-06-15 Richard Biener <rguenther@suse.de>
687 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
688 leave useless conversion stripping to force_gimple_operand_gsi.
689 (gimplify_build2): Likewise.
690 (gimplify_build1): Likewise.
692 2018-06-15 Richard Biener <rguenther@suse.de>
695 * tree-cfg.c (move_stmt_op): unshare invariant addresses
696 before adjusting their block.
698 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
700 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
701 multilibs for *-*-rtems*.
702 * config/riscv/t-rtems: New file.
704 2018-06-14 Jakub Jelinek <jakub@redhat.com>
707 * match.pd ((A +- CST1) +- CST2): Punt if last resort
708 unsigned_type_for returns NULL.
711 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
712 subregs of multi-word pseudos unless the float mode has word size.
714 2018-06-14 Richard Biener <rguenther@suse.de>
717 * tree-vect-generic.c (build_word_mode_vector_type): Remove
718 duplicate and harmful type_hash_canon.
719 * tree.c (type_hash_canon): Assert we didn't find ourselves.
721 2018-06-14 Richard Biener <rguenther@suse.de>
724 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
727 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
729 * config/rtems.h (STDINT_LONG32): Define.
731 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
732 Prachi Godbole <prachi.godbole@imgtec.com>
734 * config/mips/mips-cpus.def: Define P6600.
735 * config/mips/mips-tables.opt: Regenerate.
736 * config/mips/mips.c (mips_ucbranch_type): New enum.
737 (mips_rtx_cost_data): Add support for P6600.
738 (mips_issue_rate): Likewise.
739 (mips_multipass_dfa_lookahead): Likewise.
740 (mips_avoid_hazard): Likewise.
741 (mips_reorg_process_insns): Likewise.
742 (mips_classify_branch_p6600): New function.
743 * config/mips/mips.h (TUNE_P6600): New define.
744 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
745 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
746 * config/mips/mips.md: Include p6600.md.
747 (processor): Add p6600.
748 * config/mips/p6600.md: New file.
749 * doc/invoke.texi: Add p6600 to supported architectures.
751 2018-06-13 Martin Sebor <msebor@redhat.com>
753 PR tree-optimization/86114
754 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
756 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
758 2018-06-13 Richard Biener <rguenther@suse.de>
760 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
761 Properly set vector type of the intermediate stmt.
762 * tree-vect-stmts.c (vectorizable_operation): The destination
763 var always has vectype_out type.
765 2018-06-13 Jeff Law <law@redhat.com>
767 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
768 integer 0 for argument to print_rtl_with_bb.
769 (rl78_reorg): Likewise.
771 2018-06-13 David Malcolm <dmalcolm@redhat.com>
773 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
774 from rtx to rtx_insn *.
775 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
777 (add_sched_insns_for_speculation): Likewise for local "target",
778 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
779 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
780 from rtx_insn ** to rtx_code_label **.
781 (reorg_emit_nops): Likewise.
782 (c6x_reorg): Likewise for local "call_labels".
783 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
785 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
786 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
787 the loops over LABEL_REFS.
788 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
790 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
791 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
792 (split_branches): Strengthen local "olabel" from rtx to
793 rtx_insn *, adding a safe_as_a cast.
794 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
796 (add_insn_after): Likewise for first two params.
797 (add_insn_before): Likewise.
798 (remove_insn): Likewise for param.
799 (emit_pattern_before_noloc): Likewise for second and third params.
800 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
801 (emit_call_insn_before_noloc): Likewise.
802 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
804 (emit_barrier_before): Likewise.
805 (emit_label_before): Strengthen "label" param from "rtx" to
806 "rtx_code_label *". Strengthen "before" param from "rtx" to
808 (emit_insn_after_1): Strengthen "after" param from "rtx" to
810 (emit_pattern_after_noloc): Likewise.
811 (emit_insn_after_noloc): Likewise.
812 (emit_jump_insn_after_noloc): Likewise.
813 (emit_call_insn_after_noloc): Likewise.
814 (emit_debug_insn_after_noloc): Likewise.
815 (emit_barrier_after): Likewise.
816 (emit_label_after): Likewise for both params.
817 (emit_pattern_after_setloc): Likewise for "after" param. Convert
818 "loc" param from "int" to "location_t".
819 (emit_insn_after_setloc): Likewise.
820 (emit_jump_insn_after_setloc): Likewise.
821 (emit_call_insn_after_setloc): Likewise.
822 (emit_debug_insn_after_setloc): Likewise.
823 (emit_pattern_before_setloc): Likewise for "before" param. Convert
824 "loc" param from "int" to "location_t".
825 (emit_pattern_before): Convert NULL_RTX to NULL.
826 (emit_insn_before_setloc): Convert "loc" param from "int" to
828 (emit_jump_insn_before_setloc): Likewise.
829 (emit_call_insn_before_setloc): Likewise.
830 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
831 rtx_insn *. Convert "loc" param from "int" to "location_t".
832 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
833 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
834 Convert 3rd param from "int" to "location_t".
835 (emit_barrier_before, emit_barrier_after, next_real_insn):
836 Strengthen param from rtx to rtx_insn *.
837 (emit_label_before): Strengthen 1st param from "rtx" to
838 "rtx_code_label *". Strengthen 2nd param from "rtx" to
840 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
841 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
842 Strengthen 2nd param from "rtx" to "rtx_insn *".
843 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
844 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
845 Likewise. Convert 3rd param from "int" to "location_t".
846 (emit_label_after): Strengthen 1st param from "rtx" to
848 (next_real_insn, remove_insn): Strengthen param from "rtx" to
850 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
851 from "rtx" to "rtx_insn *".
853 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
855 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
856 bodies streamed in with -Q.
857 * dumpfile.c (dump_files): Add lto-stream-out dump file.
858 * dumpfile.h (tree_dump_index): Add lto_stream_out.
859 * gimple-streamer-out.c: Include gimple-pretty-print.h
860 (output_bb): Dump stmts streamed.
861 * lto-section-out.c: Include print-tree.h
862 (lto_begin_section): Dump sections created.
863 (lto_output_decl_index): Dump decl encoded.
864 * lto-streamer-out.c: Include print-tree.h
865 (create_output_block): Dump output block created.
866 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
867 (output_function): Dump function output.
868 (output_constructor): Dump constructor streamed.
869 (write_global_stream): Output indexes encoded.
870 (produce_asm_for_decls): Dump streams encoded.
871 * lto-streamer.c (streamer_dump_file): New global var.
872 * lto-streamer.h (streamer_dump_file): Declare.
873 * passes.c (ipa_write_summaries): Initialize streamer dump.
874 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
877 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
880 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
881 offsets for register save directives. Emit a second batch of save
882 directives, if need be, when the function accesses prior frames.
884 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
886 * config/arc/fpu.md (fmasf4): Force operand to register.
889 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
891 * config/arc/arc-protos.h (arc_pad_return): Remove.
892 * config/arc/arc.c (machine_function): Remove force_short_suffix
894 (arc_print_operand): Adjust printing of '&'.
895 (arc_verify_short): Remove conditional printing of short suffix.
896 (arc_final_prescan_insn): Remove reference to size_reason.
897 (pad_return): New function.
898 (arc_reorg): Call pad_return.
899 (arc_pad_return): Remove.
900 (arc_init_machine_status): Remove reference to force_short_suffix.
901 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
902 (attr length): When attribute iscompact is true force to 2
903 regardless; in the case of maybe check if we want to force the
904 instruction to have 4 bytes length.
905 (nopv): Change it to generate 4 byte long nop as well.
906 (blockage): New pattern.
907 (simple_return): Remove call to arc_pad_return.
908 (p_return_i): Likewise.
910 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
912 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
914 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
916 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
919 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
921 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
922 for ARC700 and ARCv2.
924 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
927 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
928 operands[2] instead of operands[1].
931 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
933 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
934 case, check whether the outer register overlaps an unallocatable
935 register, not just whether it fits the required class.
937 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
939 * poly-int.h (can_div_trunc_p): Add new overload in which all values
941 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
942 (memrefs_conflict_p): Likewise.
943 (init_alias_analysis): Likewise.
944 * cfgexpand.c (expand_debug_expr): Likewise.
945 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
946 * cse.c (fold_rtx): Likewise.
947 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
948 * expr.c (emit_block_move_hints): Likewise.
949 (clear_storage_hints, push_block, emit_push_insn): Likewise.
950 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
951 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
952 (emit_group_store): Likewise.
953 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
954 to read the PRE/POST_MODIFY increment.
955 * calls.c (store_one_arg): Use strip_offset.
956 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
958 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
960 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
961 (simplify_binary_operation_1): Extend CONST_INT handling to
963 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
964 than a HOST_WIDE_INT.
965 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
967 (adjust_mems, add_stores): Update accodingly.
968 (vt_canonicalize_addr): Track polynomial offsets.
969 (emit_note_insn_var_location): Likewise.
970 (vt_add_function_parameter): Likewise.
971 (vt_initialize): Likewise.
973 2018-06-12 Jeff Law <law@redhat.com>
975 * config.gcc (alpha*-*-freebsd*): Remove.
976 * config/alpha/freebsd.h: Remove.
978 2018-06-12 David Malcolm <dmalcolm@redhat.com>
981 * spellcheck-tree.c (levenshtein_distance): Rename to...
982 (get_edit_distance): ...this, and update for underlying renaming.
983 * spellcheck-tree.h (levenshtein_distance): Rename to...
984 (get_edit_distance): ...this.
985 * spellcheck.c (levenshtein_distance): Rename to...
986 (get_edit_distance): ...this. Convert from Levenshtein distance
987 to Damerau-Levenshtein distance by supporting transpositions of
988 adjacent characters. Rename "v1" to "v_next" and "v0" to
990 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
991 (selftest::test_edit_distance_unit_test_oneway): ...this, and
992 update for underlying renaming.
993 (selftest::levenshtein_distance_unit_test): Rename to...
994 (selftest::test_get_edit_distance_unit): ...this, and update for
996 (selftest::test_find_closest_string): Add example from PR 69968
997 where transposition helps
998 (selftest::test_metric_conditions): Update for renaming.
999 (selftest::test_metric_conditions): Likewise.
1000 (selftest::spellcheck_c_tests): Likewise.
1001 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1002 (get_edit_distance): ...this.
1003 (best_match::consider): Update for renaming.
1005 2018-06-12 Martin Sebor <msebor@redhat.com>
1007 PR tree-optimization/85259
1008 * builtins.c (compute_objsize): Handle constant offsets.
1009 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1010 true iff a warning has been issued.
1011 * gimple.h (gimple_nonartificial_location): New function.
1012 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1013 gimple_nonartificial_location and handle -Wno-system-headers.
1014 (handle_builtin_stxncpy): Same.
1016 2018-06-12 Martin Sebor <msebor@redhat.com>
1019 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1021 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1023 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1024 BUILTIN_VEC_XST entries for pointer to double and long long.
1026 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1029 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1031 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1034 2018-06-12 Martin Liska <mliska@suse.cz>
1036 * doc/options.texi: Document IntegerRange.
1038 2018-06-12 Martin Liska <mliska@suse.cz>
1040 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1041 * opt-functions.awk: Handle Deprecated flag.
1042 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1044 (read_cmdline_option): Report warning for a deprecated option.
1045 * opts.h (struct cl_option): Add new field cl_deprecated.
1046 (CL_ERR_DEPRECATED): New.
1048 2018-06-12 Martin Liska <mliska@suse.cz>
1050 * doc/options.texi: Document Deprecated option flag.
1052 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1054 * config/arc/arc-arch.h (arc_extras): New enum.
1055 (arc_cpu_t):Add field extra.
1056 (arc_cpu_types): Consider the extras.
1057 * config/arc/arc-cpus.def: Add extras info.
1058 * config/arc/arc-opts.h (processor_type): Consider extra field.
1059 * config/arc/arc.c (arc_override_options): Handle extra field.
1061 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1063 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1064 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1066 * config/arc/arc.c (arc_init): Update pic warning.
1067 (irq_range): Update irq range parsing warnings.
1068 (arc_override_options): Update various warning messages.
1069 (arc_handle_aux_attribute): Likewise.
1071 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1073 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1075 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1077 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1078 compiler messages directives.
1080 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1082 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1083 * config/mips/mips-tables.opt: Regenerate.
1084 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1086 * doc/invoke.texi: Document -march=i6500.
1088 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1090 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1091 (i6400_gpmul): Add cpu_unit.
1092 (i6400_gpdiv): Likewise.
1093 (i6400_msa_add_d): Update reservations.
1094 (i6400_msa_int_add) Likewise.
1095 (i6400_msa_short_logic3) Likewise.
1096 (i6400_msa_short_logic2) Likewise.
1097 (i6400_msa_short_logic) Likewise.
1098 (i6400_msa_move) Likewise.
1099 (i6400_msa_cmp) Likewise.
1100 (i6400_msa_short_float2) Likewise.
1101 (i6400_msa_div_d) Likewise.
1102 (i6400_msa_long_logic1) Likewise.
1103 (i6400_msa_long_logic2) Likewise.
1104 (i6400_msa_mult) Likewise.
1105 (i6400_msa_long_float2) Likewise.
1106 (i6400_msa_long_float4) Likewise.
1107 (i6400_msa_long_float5) Likewise.
1108 (i6400_msa_long_float8) Likewise.
1109 (i6400_fpu_fadd): Include frint type.
1110 (i6400_fpu_store): New define_insn_reservation.
1111 (i6400_fpu_load): Likewise.
1112 (i6400_fpu_move): Likewise.
1113 (i6400_fpu_fcmp): Likewise.
1114 (i6400_fpu_fmadd): Likewise.
1115 (i6400_int_mult): Include imul3nc type and update reservation.
1116 (i6400_int_div): Include idiv3 type and update reservation.
1117 (i6400_int_load): Update to check type not move_type.
1118 (i6400_int_store): Likewise.
1119 (i6400_int_prefetch): Set zero latency.
1121 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1123 * gcc.c: Document new %@{...} sequence.
1124 (LINK_COMMAND_SPEC): Use it for the -L switches.
1125 (cpp_unique_options): Use it for the -I switches.
1126 (at_file_argbuf): New global variable.
1127 (in_at_file): Likewise.
1128 (alloc_args): Create at_file_argbuf.
1129 (clear_args): Truncate at_file_argbuf.
1130 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1131 (open_at_file): New function.
1132 (close_at_file): Likewise.
1133 (create_at_file): Delete.
1134 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1137 (validate_switches_from_spec): Deal with %@{...} sequence.
1138 (validate_switches): Likewise.
1139 (driver::finalize): Call clear_args.
1141 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1143 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1145 2018-06-11 Martin Sebor <msebor@redhat.com>
1147 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1148 (Wno-class-memaccess): Add @opindex.
1149 (Wno-templates, Wno-multiple-inheritance): Same.
1150 (Wno-virtual-inheritance, Wno-namespaces): Same.
1151 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1152 (Wno-format-overflow, Wno-format-truncation): Same.
1153 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1154 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1155 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1156 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1157 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1158 (Wno-misspelled-isr): Same.
1160 2018-06-11 Martin Sebor <msebor@redhat.com>
1162 * PR tree-optimization/86083
1163 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1165 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1167 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1169 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1172 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1173 on the correct operand.
1174 (*movdi_internal64): Ditto.
1176 2018-06-11 Martin Liska <mliska@suse.cz>
1178 PR tree-optimization/86089
1179 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1181 2018-06-11 Julia Koval <julia.koval@intel.com>
1183 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1184 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1185 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1187 2018-06-11 Olivier Hainque <hainque@adacore.com>
1189 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1190 for Ada with strict dwarf2.
1192 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1195 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1198 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1200 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1202 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1204 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1207 2018-06-08 Martin Liska <mliska@suse.cz>
1209 * tree-cfg.h (debug_function): Fix argument type to match
1212 2018-06-08 Martin Liska <mliska@suse.cz>
1214 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1215 Remove usage of MPX-related (and removed) fields.
1216 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1218 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1220 * cfg.c (debug): Use TDF_NONE rather than 0.
1221 * cfghooks.c (debug): Likewise.
1222 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1223 (struct dump_option_value_info): Convert to...
1224 (struct kv_pair): ...this template type.
1225 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1227 (optinfo_verbosity_options): Likewise.
1228 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1230 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1231 than int for "optgroup_flags" param.
1232 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1234 (dump_dec): Likewise.
1235 (dump_finish): Use TDF_NONE rather than 0.
1236 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1237 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1238 than 0. Update for change to option_ptr.
1239 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1240 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1241 0. Update for changes to optinfo_verbosity_options and
1243 (opt_info_switch_p): Convert optgroup_flags from int to
1245 (dump_basic_block): Use dump_flags_t rather than int
1246 for "dump_kind" param.
1247 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1248 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1249 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1250 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1251 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1252 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1253 TDF_NONE): Convert from macros to...
1254 (enum dump_flag): ...this new enum.
1255 (dump_flags_t): Update to use enum.
1256 (operator|, operator&, operator~, operator|=, operator&=):
1257 Implement for dump_flags_t.
1258 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1259 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1260 Convert from macros to...
1261 (enum optgroup_flag): ...this new enum.
1262 (optgroup_flags_t): New typedef.
1263 (operator|, operator|=): Implement for optgroup_flags_t.
1264 (struct dump_file_info): Convert field "alt_flags" to
1265 dump_flags_t. Convert field "optgroup_flags" to
1267 (dump_basic_block): Use dump_flags_t rather than int for param.
1268 (dump_generic_expr_loc): Likewise.
1269 (dump_dec): Likewise.
1270 (dump_register): Convert param "optgroup_flags" to
1272 (opt_info_enable_passes): Likewise.
1273 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1275 * gimple-pretty-print.c (debug): Likewise.
1276 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1277 (merged_store_group::apply_stores): Likewise.
1278 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1279 * gimple.c (verify_gimple_pp): Likewise.
1280 * graphite-poly.c (print_pbb_body): Likewise.
1281 * passes.c (pass_manager::register_one_dump_file): Convert
1282 local "optgroup_flags" to optgroup_flags_t.
1283 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1285 (debug_body): Likewise.
1286 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1287 to optgroup_flags_t.
1288 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1290 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1291 (convert_mult_to_fma): Likewise.
1292 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1293 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1294 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1295 "dump_kind" to dump_flags_t.
1297 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1299 * config/rs6000/rs6000.c (min, max): Delete.
1301 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1303 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1304 -mabi=spe and -mabi=no-spe.
1306 2018-06-08 Martin Liska <mliska@suse.cz>
1308 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1309 where we expect an existing summary.
1311 2018-06-08 Martin Liska <mliska@suse.cz>
1313 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1314 * ipa-inline.h (estimate_edge_growth): Likewise.
1316 2018-06-08 Martin Liska <mliska@suse.cz>
1318 * cgraph.c (function_version_hasher::hash): Use
1319 cgraph_node::get_uid ().
1320 (function_version_hasher::equal):
1321 * cgraph.h (cgraph_node::get_uid): New method.
1322 * ipa-inline.c (update_caller_keys): Use
1323 cgraph_node::get_uid ().
1324 (update_callee_keys): Likewise.
1325 * ipa-utils.c (searchc): Likewise.
1326 (ipa_reduced_postorder): Likewise.
1327 * lto-cgraph.c (input_node): Likewise.
1328 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1329 * symbol-summary.h (symtab_insertion): Likewise.
1330 (symtab_removal): Likewise.
1331 (symtab_duplication): Likewise.
1332 * tree-pretty-print.c (dump_function_header): Likewise.
1333 * tree-sra.c (convert_callers_for_node): Likewise.
1335 2018-06-08 Martin Liska <mliska@suse.cz>
1337 * cgraph.c (symbol_table::create_edge): Always assign a new
1339 (symbol_table::free_edge): Do not recycle numbers.
1340 * cgraph.h (cgraph_edge::get): New method.
1341 * symbol-summary.h (symtab_removal): Use it.
1342 (symtab_duplication): Likewise.
1343 (call_summary::hashable_uid): Remove.
1345 2018-06-08 Martin Liska <mliska@suse.cz>
1347 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1348 (initialize_growth_caches): Remove.
1349 (free_growth_caches): Likewise.
1350 (do_estimate_edge_time): Use edge_growth_cache.
1351 (do_estimate_edge_size): Likewise.
1352 (do_estimate_edge_hints): Likewise.
1353 * ipa-inline.c (reset_edge_caches): Likewise.
1354 (recursive_inlining): Likewise.
1355 (inline_small_functions): Likewise.
1356 * ipa-inline.h (initialize_growth_caches): Remove.
1357 (estimate_edge_size): Likewise.
1358 (estimate_edge_time): Likewise.
1359 (estimate_edge_hints): Likewise.
1360 (reset_edge_growth_cache): Likewise.
1361 * symbol-summary.h (call_summary::remove): New method.
1363 2018-06-08 Martin Liska <mliska@suse.cz>
1365 * ipa-cp.c (class edge_clone_summary): New summary.
1366 (grow_edge_clone_vectors): Remove.
1367 (ipcp_edge_duplication_hook): Remove.
1368 (class edge_clone_summary_t): New call_summary class.
1369 (ipcp_edge_removal_hook): Remove.
1370 (edge_clone_summary_t::duplicate): New function.
1371 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1372 (create_specialized_node): Likewise.
1373 (ipcp_driver): Initialize edge_clone_summaries and do not
1376 2018-06-08 Martin Liska <mliska@suse.cz>
1378 * symbol-summary.h (get): New function.
1379 (call_summary::m_initialize_when_cloning): New class member.
1381 2018-06-08 Martin Liska <mliska@suse.cz>
1383 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1384 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1385 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1386 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1387 (remove_cgraph_node_from_order): Use the removed_nodes set.
1388 (do_per_function_toporder): Likwise.
1389 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1390 instead of summary_uid.
1391 (symtab_removal): Likewise.
1392 (symtab_duplication): Likewise.
1394 2018-06-08 Martin Liska <mliska@suse.cz>
1396 * ipa-cp.c (ipcp_store_bits_results): Use
1397 ipcp_transformation_sum.
1398 (ipcp_store_vr_results): Likewise.
1399 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1401 (ipcp_transformation_initialize): ... this.
1402 (ipa_set_node_agg_value_chain):
1403 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1404 (write_ipcp_transformation_info): Likewise.
1405 (read_ipcp_transformation_info): Likewise.
1406 (ipcp_update_bits): Likewise.
1407 (ipcp_update_vr): Likewise.
1408 (ipcp_transform_function): Likewise.
1409 * ipa-prop.h: Rename ipcp_transformation_summary to
1410 ipcp_transformation.
1411 (class ipcp_transformation_t): New function summary.
1412 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1413 (ipa_get_agg_replacements_for_node): Likewise.
1415 2018-06-08 Martin Liska <mliska@suse.cz>
1417 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1419 (class funct_state_summary_t): New function_summary class.
1420 (has_function_state): Remove.
1421 (get_function_state): Likewise.
1422 (set_function_state): Likewise.
1423 (add_new_function): Likewise.
1424 (funct_state_summary_t::insert): New function.
1425 (duplicate_node_data): Remove.
1426 (remove_node_data): Remove.
1427 (funct_state_summary_t::duplicate): New function.
1428 (register_hooks): Create new funct_state_summaries.
1429 (pure_const_generate_summary): Use it.
1430 (pure_const_write_summary): Likewise.
1431 (pure_const_read_summary): Likewise.
1432 (propagate_pure_const): Likewise.
1433 (propagate_nothrow): Likewise.
1434 (dump_malloc_lattice): Likewise.
1435 (propagate_malloc): Likewise.
1436 (execute): Do not register hooks, just remove summary
1438 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1441 2018-06-08 Martin Liska <mliska@suse.cz>
1443 * ipa-reference.c (remove_node_data): Remove.
1444 (duplicate_node_data): Likewise.
1445 (class ipa_ref_var_info_summary_t): New class.
1446 (class ipa_ref_opt_summary_t): Likewise.
1447 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1448 (get_reference_optimization_summary): Use
1449 ipa_ref_opt_sum_summaries.
1450 (set_reference_vars_info): Remove.
1451 (set_reference_optimization_summary): Likewise.
1452 (ipa_init): Create summaries.
1453 (init_function_info): Use function summary.
1454 (ipa_ref_opt_summary_t::duplicate): New function.
1455 (ipa_ref_opt_summary_t::remove): New function.
1456 (get_read_write_all_from_node): Fix GNU coding style.
1457 (propagate): Use function summary.
1458 (write_node_summary_p): Fix GNU coding style.
1459 (stream_out_bitmap): Likewise.
1460 (ipa_reference_read_optimization_summary): Use function summary.
1461 (ipa_reference_c_finalize): Do not release hooks.
1463 2018-06-08 Martin Liska <mliska@suse.cz>
1465 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1466 (analyze_function_body): Extract multiple calls of get_create.
1467 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1468 * ipa-inline.c (recursive_inlining): Use ::get method.
1469 * ipa-inline.h (estimate_edge_growth): Likewise.
1471 2018-06-08 Martin Liska <mliska@suse.cz>
1473 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1475 (hsa_function_summary::hsa_function_summary): Use the new enum
1477 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1478 * hsa-gen.c (hsa_get_host_function): Likewise.
1479 (get_brig_function_name): Likewise.
1480 * ipa-hsa.c (process_hsa_functions): Likewise.
1481 (ipa_hsa_write_summary): Likewise.
1482 * symbol-summary.h (symtab_duplication): Use ::get function/
1483 (get): New function.
1485 2018-06-08 Martin Liska <mliska@suse.cz>
1487 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1489 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1490 (hsa_register_kernel): Likewise.
1491 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1492 * hsa-gen.c (hsa_get_host_function): Likewise.
1493 (get_brig_function_name): Likewise.
1494 (generate_hsa): Likewise.
1495 (pass_gen_hsail::execute): Likewise.
1496 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1497 (devirtualization_time_bonus): Likewise.
1498 (ipcp_propagate_stage): Likewise.
1499 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1500 (edge_set_predicate): Likewise.
1501 (evaluate_conditions_for_known_args): Likewise.
1502 (evaluate_properties_for_edge): Likewise.
1503 (ipa_fn_summary::reset): Likewise.
1504 (ipa_fn_summary_t::duplicate): Likewise.
1505 (dump_ipa_call_summary): Likewise.
1506 (ipa_dump_fn_summary): Likewise.
1507 (analyze_function_body): Likewise.
1508 (compute_fn_summary): Likewise.
1509 (estimate_edge_devirt_benefit): Likewise.
1510 (estimate_edge_size_and_time): Likewise.
1511 (estimate_calls_size_and_time): Likewise.
1512 (estimate_node_size_and_time): Likewise.
1513 (inline_update_callee_summaries): Likewise.
1514 (remap_edge_change_prob): Likewise.
1515 (remap_edge_summaries): Likewise.
1516 (ipa_merge_fn_summary_after_inlining): Likewise.
1517 (ipa_update_overall_fn_summary): Likewise.
1518 (read_ipa_call_summary): Likewise.
1519 (inline_read_section): Likewise.
1520 (write_ipa_call_summary): Likewise.
1521 (ipa_fn_summary_write): Likewise.
1522 (ipa_free_fn_summary): Likewise.
1523 * ipa-hsa.c (process_hsa_functions): Likewise.
1524 (ipa_hsa_write_summary): Likewise.
1525 (ipa_hsa_read_section): Likewise.
1526 * ipa-icf.c (sem_function::merge): Likewise.
1527 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1528 (do_estimate_edge_time): Likewise.
1529 (estimate_size_after_inlining): Likewise.
1530 (estimate_growth): Likewise.
1531 (growth_likely_positive): Likewise.
1532 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1533 (inline_call): Likewise.
1534 * ipa-inline.c (caller_growth_limits): Likewise.
1535 (can_inline_edge_p): Likewise.
1536 (can_inline_edge_by_limits_p): Likewise.
1537 (compute_uninlined_call_time): Likewise.
1538 (compute_inlined_call_time): Likewise.
1539 (want_inline_small_function_p): Likewise.
1540 (edge_badness): Likewise.
1541 (update_caller_keys): Likewise.
1542 (update_callee_keys): Likewise.
1543 (recursive_inlining): Likewise.
1544 (inline_small_functions): Likewise.
1545 (inline_to_all_callers_1): Likewise.
1546 (dump_overall_stats): Likewise.
1547 (early_inline_small_functions): Likewise.
1548 (early_inliner): Likewise.
1549 * ipa-inline.h (estimate_edge_growth): Likewise.
1550 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1551 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1552 * ipa-prop.h (IPA_NODE_REF): Likewise.
1553 (IPA_EDGE_REF): Likewise.
1554 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1555 (propagate_malloc): Likewise.
1556 * ipa-split.c (execute_split_functions): Likewise.
1557 * symbol-summary.h: Rename get to get_create.
1559 (get_create): Likewise.
1560 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1562 2018-06-08 Martin Liska <mliska@suse.cz>
1564 * symbol-summary.h (release): Move definition out of class
1566 (symtab_removal): Likewise.
1567 (symtab_duplication): Likewise.
1569 2018-06-08 Martin Liska <mliska@suse.cz>
1571 * symbol-summary.h (function_summary): Move constructor
1572 implementation out of class declaration.
1573 (release): Likewise.
1574 (symtab_insertion): Likewise.
1575 (symtab_removal): Likewise.
1576 (symtab_duplication): Likewise.
1579 2018-06-08 Martin Liska <mliska@suse.cz>
1581 * Makefile.in: Remove support for MPX (macros, related functions,
1582 fields in cgraph_node, ...).
1583 * builtin-types.def (BT_BND): Likewise.
1584 (BT_FN_BND_CONST_PTR): Likewise.
1585 (BT_FN_CONST_PTR_BND): Likewise.
1586 (BT_FN_VOID_PTR_BND): Likewise.
1587 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1588 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1589 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1590 (expand_builtin_mempcpy_with_bounds): Likewise.
1591 (expand_builtin_memset_with_bounds): Likewise.
1592 (expand_builtin_memset_args): Likewise.
1593 (std_expand_builtin_va_start): Likewise.
1594 (expand_builtin): Likewise.
1595 (expand_builtin_with_bounds): Likewise.
1596 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1597 (DEF_LIB_BUILTIN_CHKP): Likewise.
1598 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1599 (DEF_CHKP_BUILTIN): Likewise.
1600 (BUILT_IN_MEMCPY): Likewise.
1601 (BUILT_IN_MEMMOVE): Likewise.
1602 (BUILT_IN_MEMPCPY): Likewise.
1603 (BUILT_IN_MEMSET): Likewise.
1604 (BUILT_IN_STPCPY): Likewise.
1605 (BUILT_IN_STRCAT): Likewise.
1606 (BUILT_IN_STRCHR): Likewise.
1607 (BUILT_IN_STRCPY): Likewise.
1608 (BUILT_IN_STRLEN): Likewise.
1609 (BUILT_IN_MEMCPY_CHK): Likewise.
1610 (BUILT_IN_MEMMOVE_CHK): Likewise.
1611 (BUILT_IN_MEMPCPY_CHK): Likewise.
1612 (BUILT_IN_MEMSET_CHK): Likewise.
1613 (BUILT_IN_STPCPY_CHK): Likewise.
1614 (BUILT_IN_STRCAT_CHK): Likewise.
1615 (BUILT_IN_STRCPY_CHK): Likewise.
1616 * calls.c (store_bounds): Likewise.
1617 (emit_call_1): Likewise.
1618 (special_function_p): Likewise.
1619 (maybe_warn_nonstring_arg): Likewise.
1620 (initialize_argument_information): Likewise.
1621 (finalize_must_preallocate): Likewise.
1622 (compute_argument_addresses): Likewise.
1623 (expand_call): Likewise.
1624 * cfgexpand.c (expand_call_stmt): Likewise.
1625 (expand_return): Likewise.
1626 (expand_gimple_stmt_1): Likewise.
1627 (pass_expand::execute): Likewise.
1628 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1629 (cgraph_node::remove): Likewise.
1630 (cgraph_node::dump): Likewise.
1631 (cgraph_node::verify_node): Likewise.
1632 * cgraph.h (chkp_function_instrumented_p): Likewise.
1633 (symtab_node::get_alias_target): Likewise.
1634 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1635 (cgraph_local_p): Likewise.
1636 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1637 (cgraph_edge::rebuild_references): Likewise.
1638 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1639 (walk_polymorphic_call_targets): Likewise.
1640 (cgraph_node::expand_thunk): Likewise.
1641 (symbol_table::output_weakrefs): Likewise.
1642 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1643 (ix86_handle_option): Likewise.
1644 * config/i386/constraints.md: Likewise.
1645 * config/i386/i386-builtin-types.def (BND): Likewise.
1649 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1650 (BDESC_FIRST): Likewise.
1652 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1653 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1654 * config/i386/i386.c (enum reg_class): Likewise.
1655 (ix86_target_string): Likewise.
1656 (ix86_option_override_internal): Likewise.
1657 (ix86_conditional_register_usage): Likewise.
1658 (ix86_valid_target_attribute_inner_p): Likewise.
1659 (ix86_set_indirect_branch_type): Likewise.
1660 (ix86_set_current_function): Likewise.
1661 (ix86_function_arg_regno_p): Likewise.
1662 (init_cumulative_args): Likewise.
1663 (ix86_function_arg_advance): Likewise.
1664 (ix86_function_arg): Likewise.
1665 (ix86_pass_by_reference): Likewise.
1666 (ix86_function_value_regno_p): Likewise.
1667 (ix86_function_value_1): Likewise.
1668 (ix86_function_value_bounds): Likewise.
1669 (ix86_return_in_memory): Likewise.
1670 (ix86_setup_incoming_vararg_bounds): Likewise.
1671 (ix86_va_start): Likewise.
1672 (indirect_thunk_need_prefix): Likewise.
1673 (print_reg): Likewise.
1674 (ix86_print_operand): Likewise.
1675 (ix86_expand_call): Likewise.
1676 (ix86_output_function_return): Likewise.
1677 (reg_encoded_number): Likewise.
1678 (BDESC_VERIFYS): Likewise.
1679 (ix86_init_mpx_builtins): Likewise.
1680 (ix86_init_builtins): Likewise.
1681 (ix86_emit_cmove): Likewise.
1682 (ix86_emit_move_max): Likewise.
1683 (ix86_expand_builtin): Likewise.
1684 (ix86_builtin_mpx_function): Likewise.
1685 (ix86_get_arg_address_for_bt): Likewise.
1686 (ix86_load_bounds): Likewise.
1687 (ix86_store_bounds): Likewise.
1688 (ix86_load_returned_bounds): Likewise.
1689 (ix86_store_returned_bounds): Likewise.
1690 (ix86_class_likely_spilled_p): Likewise.
1691 (ix86_hard_regno_mode_ok): Likewise.
1692 (x86_order_regs_for_local_alloc): Likewise.
1693 (ix86_mitigate_rop): Likewise.
1694 (ix86_bnd_prefixed_insn_p): Likewise.
1695 (ix86_mpx_bound_mode): Likewise.
1696 (ix86_make_bounds_constant): Likewise.
1697 (ix86_initialize_bounds): Likewise.
1698 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1699 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1700 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1701 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1702 (TARGET_CHKP_BOUND_MODE): Likewise.
1703 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1704 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1705 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1706 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1707 * config/i386/i386.h (TARGET_MPX): Likewise.
1708 (TARGET_MPX_P): Likewise.
1709 (VALID_BND_REG_MODE): Likewise.
1710 (FIRST_BND_REG): Likewise.
1711 (LAST_BND_REG): Likewise.
1712 (enum reg_class): Likewise.
1713 (BND_REG_P): Likewise.
1714 (BND_REGNO_P): Likewise.
1715 (BNDmode): Likewise.
1716 (ADJUST_INSN_LENGTH): Likewise.
1717 * config/i386/i386.md: Likewise.
1718 * config/i386/i386.opt: Likewise.
1719 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1720 (defined): Likewise.
1721 (LINK_MPX): Likewise.
1722 (MPX_SPEC): Likewise.
1723 (LIBMPX_SPEC): Likewise.
1724 (LIBMPXWRAPPERS_SPEC): Likewise.
1725 (CHKP_SPEC): Likewise.
1726 * config/i386/predicates.md: Likewise.
1727 * dbxout.c (dbxout_type): Likewise.
1728 * doc/extend.texi: Likewise.
1729 * doc/invoke.texi: Likewise.
1730 * doc/md.texi: Likewise.
1731 * doc/tm.texi: Likewise.
1732 * doc/tm.texi.in: Likewise.
1733 * dwarf2out.c (is_base_type): Likewise.
1734 (gen_formal_types_die): Likewise.
1735 (gen_subprogram_die): Likewise.
1736 (gen_type_die_with_usage): Likewise.
1737 (gen_decl_die): Likewise.
1738 (dwarf2out_late_global_decl): Likewise.
1739 * expr.c (expand_assignment): Likewise.
1740 (emit_storent_insn): Likewise.
1741 (store_expr_with_bounds): Likewise.
1742 (store_expr): Likewise.
1743 (expand_expr_real_1): Likewise.
1744 * expr.h (store_expr_with_bounds): Likewise.
1745 * function.c (use_register_for_decl): Likewise.
1746 (struct bounds_parm_data): Likewise.
1747 (assign_parms_augmented_arg_list): Likewise.
1748 (assign_parm_find_entry_rtl): Likewise.
1749 (assign_parm_is_stack_parm): Likewise.
1750 (assign_parm_load_bounds): Likewise.
1751 (assign_bounds): Likewise.
1752 (assign_parms): Likewise.
1753 (expand_function_start): Likewise.
1754 * gcc.c (CHKP_SPEC): Likewise.
1755 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1756 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1757 (wrestrict_dom_walker::check_call): Likewise.
1758 * gimple.c (gimple_build_call_from_tree): Likewise.
1759 * gimple.h (enum gf_mask): Likewise.
1760 (gimple_call_with_bounds_p): Likewise.
1761 (gimple_call_set_with_bounds): Likewise.
1762 * gimplify.c (gimplify_init_constructor): Likewise.
1763 * ipa-cp.c (initialize_node_lattices): Likewise.
1764 (propagate_constants_across_call): Likewise.
1765 (find_more_scalar_values_for_callers_subset): Likewise.
1766 * ipa-hsa.c (process_hsa_functions): Likewise.
1767 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1768 * ipa-icf.c (sem_function::merge): Likewise.
1769 * ipa-inline.c (early_inliner): Likewise.
1770 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1771 (warn_function_cold): Likewise.
1772 (propagate_pure_const): Likewise.
1773 * ipa-ref.h (enum GTY): Likewise.
1774 * ipa-split.c (find_retbnd): Likewise.
1775 (consider_split): Likewise.
1776 (split_function): Likewise.
1777 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1778 * ipa.c (walk_polymorphic_call_targets): Likewise.
1779 (symbol_table::remove_unreachable_nodes): Likewise.
1780 (process_references): Likewise.
1781 (cgraph_build_static_cdtor_1): Likewise.
1782 * lto-cgraph.c (lto_output_node): Likewise.
1783 (output_refs): Likewise.
1784 (compute_ltrans_boundary): Likewise.
1785 (input_overwrite_node): Likewise.
1786 (input_node): Likewise.
1787 (input_cgraph_1): Likewise.
1788 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1789 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1790 (class pass_chkp_instrumentation_passes): Likewise.
1791 (make_pass_chkp_instrumentation_passes): Likewise.
1792 * passes.def: Likewise.
1793 * rtl.h (struct GTY): Likewise.
1794 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1795 * stor-layout.c (layout_type): Likewise.
1796 * symtab.c: Likewise.
1797 * target.def: Likewise.
1798 * targhooks.c (default_chkp_bound_type): Likewise.
1799 (default_chkp_bound_mode): Likewise.
1800 (default_builtin_chkp_function): Likewise.
1801 (default_chkp_function_value_bounds): Likewise.
1802 (default_chkp_make_bounds_constant): Likewise.
1803 (default_chkp_initialize_bounds): Likewise.
1804 * targhooks.h (default_chkp_bound_type): Likewise.
1805 (default_chkp_bound_mode): Likewise.
1806 (default_builtin_chkp_function): Likewise.
1807 (default_chkp_function_value_bounds): Likewise.
1808 (default_chkp_make_bounds_constant): Likewise.
1809 (default_chkp_initialize_bounds): Likewise.
1810 * toplev.c (compile_file): Likewise.
1811 (process_options): Likewise.
1812 * tree-core.h (DEF_BUILTIN): Likewise.
1813 (DEF_BUILTIN_CHKP): Likewise.
1814 * tree-inline.c (declare_return_variable): Likewise.
1815 (remap_gimple_stmt): Likewise.
1816 (copy_bb): Likewise.
1817 (initialize_inlined_parameters): Likewise.
1818 (expand_call_inline): Likewise.
1819 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1820 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1821 (make_pass_ipa_chkp_produce_thunks): Likewise.
1822 (make_pass_chkp): Likewise.
1823 (make_pass_chkp_opt): Likewise.
1824 (make_pass_chkp_instrumentation_passes): Likewise.
1825 * tree-pretty-print.c (dump_generic_node): Likewise.
1826 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1827 * tree-ssa-dce.c (propagate_necessity): Likewise.
1828 (eliminate_unnecessary_stmts): Likewise.
1829 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1830 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1831 * tree-ssa-sccvn.h: Likewise.
1832 * tree-ssa-strlen.c (get_string_length): Likewise.
1833 (valid_builtin_call): Likewise.
1834 (adjust_last_stmt): Likewise.
1835 (handle_builtin_strchr): Likewise.
1836 (handle_builtin_strcpy): Likewise.
1837 (handle_builtin_stxncpy): Likewise.
1838 (handle_builtin_memcpy): Likewise.
1839 (handle_builtin_strcat): Likewise.
1840 (strlen_check_and_optimize_stmt): Likewise.
1841 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1842 * tree-streamer-in.c: Likewise.
1843 * tree-streamer.c (record_common_node): Likewise.
1844 * tree.c (tree_code_size): Likewise.
1845 (wide_int_to_tree_1): Likewise.
1846 (type_contains_placeholder_1): Likewise.
1847 (build_common_tree_nodes): Likewise.
1848 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1849 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1850 (POINTER_BOUNDS_P): Likewise.
1851 (BOUNDED_TYPE_P): Likewise.
1852 (BOUNDED_P): Likewise.
1853 (CALL_WITH_BOUNDS_P): Likewise.
1854 (pointer_bounds_type_node): Likewise.
1855 * value-prof.c (gimple_ic): Likewise.
1856 * var-tracking.c (vt_add_function_parameters): Likewise.
1857 * varasm.c (make_decl_rtl): Likewise.
1858 (assemble_start_function): Likewise.
1859 (output_constant): Likewise.
1860 (maybe_assemble_visibility): Likewise.
1861 * varpool.c (ctor_for_folding): Likewise.
1862 * chkp-builtins.def: Remove.
1863 * ipa-chkp.c: Remove.
1864 * ipa-chkp.h: Remove.
1865 * rtl-chkp.c: Remove.
1866 * rtl-chkp.h: Remove.
1867 * tree-chkp-opt.c: Remove.
1868 * tree-chkp.c: Remove.
1869 * tree-chkp.h: Remove.
1871 2018-06-07 Carl Love <cel@us.ibm.com>
1873 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1874 vextract_fp_from_shortl): Add BE support.
1876 2018-06-07 Paul Koning <ni1d@arrl.net>
1878 * compare-elim.c (try_merge_compare): Don't merge compare if
1879 address contains a side effect.
1880 (try_eliminate_compare): Likewise.
1882 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1884 * config.gcc: Support "tremont".
1885 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1886 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1888 * config/i386/i386.c (m_TREMONT): Define.
1889 (processor_target_table): Add "tremont".
1890 (PTA_TREMONT): Define.
1891 (ix86_lea_outperforms): Add TARGET_TREMONT.
1892 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1893 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1894 and M_INTEL_GOLDMONT_PLUS.
1895 (fold_builtin_cpu): Add "tremont".
1896 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1897 (ix86_option_override_internal): Add "tremont".
1898 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1899 (processor_type): Add PROCESSOR_TREMONT.
1900 * config/i386/x86-tune.def: Add m_TREMONT.
1901 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1903 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1905 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1906 symbol defined for msp430i* devices to be lower case.
1908 gcc/testsuite/gcc.target/msp430/
1909 * msp430i-device-symbol.c: New test.
1910 * msp430f-device-symbol.c: New test.
1911 * msp430.h: New test header file.
1913 2018-06-07 Richard Biener <rguenther@suse.de>
1915 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1916 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1917 Properly wrap signed arithmetic if overflow wraps.
1919 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1921 PR tree-optimization/69615
1922 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1923 of a cast from a same precision integral SSA_NAME in a bb dominated
1924 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1925 cast to utype if rhs2 has already a compatible type.
1927 2018-06-07 Richard Biener <rguenther@suse.de>
1929 PR tree-optimization/85935
1930 * graphite-scop-detection.c (find_params_in_bb): Analyze
1931 condition operands with respect to the correct loop. Assert
1932 the analysis doesn't fail.
1934 2018-06-04 Carl Love <cel@us.ibm.com>
1936 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1937 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1938 as it is slightly cheaper.
1939 (first_match_or_eos_index_<mode>):
1940 Calculate index using natural element order.
1941 (first_match_index_<mode>):
1942 Calculate index using natural element order.
1943 (first_match_or_eos_index_<mode>):
1944 Calculate index using natural order.
1945 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1946 for BE and LE modes.
1947 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1948 P9V_BUILTIN_VCLZLSBB_V16QI.
1949 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1952 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1954 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1955 indentation and line wrap for many prototypes. Add missing
1956 @smallexample directives around block of prototypes for vec_xl and
1959 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1961 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1962 track if we pass or return IEEE 128-bit floating point.
1963 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1964 C++ mangling that is compatible with GCC 8.1.
1965 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1966 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1967 floating point types.
1968 (rs6000_function_arg_advance_1): Likewise.
1969 (rs6000_mangle_type): Optionally generate mangled names that match
1970 what GCC 8.1 generated for IEEE 128-bit floating point types.
1971 (rs6000_globalize_decl_name): If we have an external function that
1972 passes or returns IEEE 128-bit types, generate a weak reference
1973 from the mangled name used in GCC 8.1 to the current mangled
1975 (rs6000_init_builtins): Make __ibm128 use the long double type if
1976 long double is IBM extended double. Make __float128 use the long
1977 double type if long double is IEEE 128-bit.
1980 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1981 macro for __ibm128 built-in functions.
1982 (PACK_IF): Add __ibm128 pack/unpack functions.
1983 (UNPACK_IF): Likewise.
1984 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1985 enable long double built-in functions if long double is IEEE
1986 128-bit floating point.
1987 (rs6000_invalid_builtin): Update long double built-in function
1989 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1990 functions, adjust the built-in function to use the long double
1991 built-in function if __ibm128 and long double are the same type.
1992 * doc/extend.texi (PowerPC builtins): Update documention for
1993 __builtin_{,un}pack_longdouble. Add documentation for
1994 __builtin_{,un}pack_ibm128.
1996 2018-06-06 Jim Wilson <jimw@sifive.com>
1998 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1999 (struct machine_function): New field interrupt_mode.
2000 (riscv_handle_type_attribute): New function. Add forward declaration.
2001 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2002 (riscv_expand_epilogue): Check interrupt_mode field.
2003 (riscv_set_current_function): Check interrupt attribute args and
2004 set interrupt_mode field.
2005 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2006 (riscv_sret, riscv_uret): New.
2007 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2008 new arguments to interrupt attribute.
2010 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2013 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2014 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2016 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2018 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2019 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2020 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2021 several redundant entries.
2023 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2025 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2026 type from "rtx" to "rtx_insn *".
2027 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2028 for local "call_insn", removing cast.
2029 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2031 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2033 PR tree-optimization/86066
2034 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2035 for BIT_INSERT_EXPR stores.
2037 2018-06-06 Richard Biener <rguenther@suse.de>
2039 PR tree-optimization/86062
2040 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2041 component refs ontop
2042 of to be offsetted base.
2044 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2046 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2047 to be static and remove check on interrupt attribute name.
2049 gcc/testsuite/gcc.target/msp430/
2050 * function-attributes-4.c: New test.
2051 * static-interrupts.c: New test.
2053 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2055 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2056 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2058 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2061 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2063 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2064 Remove second argument, change how error is called.
2065 (aarch64_layout_arg): Remove second argument from
2066 aarch64_err_no_fpadvsimd call.
2067 (aarch64_init_cumulative_args): Ditto.
2068 (aarch64_gimplify_va_arg_expr): Ditto.
2069 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2071 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2073 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2074 (*simple_return_indirect_internal<mode>): Rename from
2075 simple_return_indirect_internal. Use W mode iterator.
2076 (rstorssp): New expander.
2077 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2078 (clrssbsy): New expander.
2079 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2081 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2083 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2085 (cmse_check_pointed_object): Likewise.
2087 2018-06-05 Martin Liska <mliska@suse.cz>
2089 PR gcov-profile/47618
2090 * doc/invoke.texi: Document how -fprofile-dir format
2093 2018-06-05 Richard Biener <rguenther@suse.de>
2095 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2096 removal pretend DOM info isn't available so we do not update
2097 it and only remove edges, not dominated blocks. Actually free
2098 DOM info in case we removed something. Remove unreachable blocks.
2099 (mfb_keep_latches): Work with either DOM info or marked backedges.
2100 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2101 first. Mark backedges if DOM info isn't available.
2102 (Re-)compute DOM info after cleanup_control_flow_pre.
2104 2018-06-05 Richard Biener <rguenther@suse.de>
2106 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2107 (locus_discrim_hasher::hash): Adjust.
2108 (locus_discrim_hasher::equal): Likewise.
2109 (next_discriminator_for_locus): Work on line directly.
2110 (same_line_p): Pass in expanded locus1 as well.
2111 (assign_discriminators): Avoid redundant location expansions.
2113 2018-06-05 Richard Biener <rguenther@suse.de>
2115 PR tree-optimization/86046
2116 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2117 if required after clearing TREE_ADDRESSABLE.
2119 2018-06-05 Richard Biener <rguenther@suse.de>
2121 PR tree-optimization/86047
2122 * tree-ssa-loop.c (for_each_index): Glob handling of all
2123 decls and constants and really handle all of them.
2125 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2128 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2129 qualifier_void_pointer and qualifier_const_void_pointer.
2130 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2131 (arm_init_builtins): Handle the above.
2132 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2133 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2134 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2137 2018-06-05 Martin Liska <mliska@suse.cz>
2139 * auto-profile.c (read_autofdo_file): Do not use
2140 gcov_ctr_summary struct.
2141 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2142 * coverage.c (struct counts_entry): Likewise.
2143 (read_counts_file): Read just single summary entry.
2144 (get_coverage_counts): Use gcov_summary struct.
2145 * coverage.h (get_coverage_counts): Likewise.
2146 * gcov-dump.c (dump_working_sets): Likewise.
2147 (tag_summary): Dump just single summary.
2148 * gcov-io.c (gcov_write_summary): Write just histogram
2150 (gcov_read_summary): Read just single summary.
2151 (compute_working_sets): Use gcov_summary struct.
2152 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2153 of GCOV_COUNTERS_SUMMABLE.
2154 (GCOV_COUNTERS_SUMMABLE): Remove.
2155 (GCOV_FIRST_VALUE_COUNTER): Replace with
2156 GCOV_COUNTER_V_INTERVAL.
2157 (struct gcov_ctr_summary): Remove.
2158 (struct gcov_summary): Directly use fields of former
2160 (compute_working_sets): Use gcov_summary struct.
2161 * gcov.c (read_count_file): Do not use ctrs fields.
2162 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2164 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2166 * profile.c: Likewise.
2167 * profile.h: Likewise.
2169 2018-06-05 Martin Liska <mliska@suse.cz>
2171 PR gcov-profile/84846
2172 * gcov.c (output_lines): Print working directory only
2173 in intermediate format.
2175 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2177 * config/s390/s390-builtin-types.def: Add void function type.
2178 * config/s390/s390-builtins.def: Use the function type for the
2181 2018-06-04 Jim Wilson <jimw@sifive.com>
2183 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2185 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2186 and maybe_eh_return. Change regno to unsigned int. Use new args to
2187 handle EH_RETURN_DATA_REGNO registers properly.
2188 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2189 (riscv_expand_epilogue): Update comment. Change argument name and
2190 type. Update code to use new name and type. Pass new args to
2191 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2193 * config/riscv/riscv.md (NORMAL_RETURN): New.
2194 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2195 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2196 (eh_return): Call gen_eh_return_internal and emit barrier.
2197 (eh_return_internal): Call riscv_expand_epilogue.
2199 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2201 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2202 bit_insertion field and declare can_be_merged_into method.
2203 (merged_store_group::can_be_merged_into): New method.
2204 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2205 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2206 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2208 2018-06-04 Richard Biener <rguenther@suse.de>
2210 PR tree-optimization/85955
2211 * builtins.c (fold_builtin_sincos): Convert pointers to
2212 destination to appropriate type before dereferencing.
2214 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2216 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2218 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2220 * expr.c (expand_expr_real_1): Force the operand into memory if
2221 its TYPE_MODE is BLKmode and if there is no integer mode for
2222 the number of bits being extracted.
2224 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2228 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2229 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2231 2018-06-04 Richard Biener <rguenther@suse.de>
2233 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2234 (cleanup_tree_cfg_noloop): ... single caller. Do
2235 start_recording_case_labels later.
2237 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2239 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2240 to _IMMINTRIN_H_INCLUDED.
2241 * config/i386/pconfigintrin.h: Ditto.
2242 * config/i386/waitpkgintrin.h: Ditto.
2243 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2244 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2245 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2246 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2247 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2248 waitpkgintrin.h and cldemoteintrin.h.
2250 2018-06-04 Richard Biener <rguenther@suse.de>
2252 PR tree-optimization/86038
2253 * tracer.c (find_best_successor): Check probability for
2254 being initialized, bail out if not.
2256 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2259 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2260 of bits to ignore when comparing architectures.
2262 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2264 PR tree-optimization/69615
2265 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2266 maximum or minimum of the type, try to merge it also as if
2267 range1 is + [-, x - 1] or + [x + 1, -].
2270 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2272 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2274 PR tree-optimization/86034
2275 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2276 the unsigned bitfield type in a bit insertion sequence if it does not
2277 have a larger precision than the bitfield size.
2278 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2280 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2282 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2284 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2286 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2287 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2288 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2289 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2291 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2293 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2294 Disable -fdelete-null-pointer-checks for ELF toolchain.
2296 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2297 Kito Cheng <kito.cheng@gmail.com>
2299 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2300 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2301 (nds32*-*-*): Add glibc and uclibc conditions.
2302 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2303 (TARGET_EXCEPT_UNWIND_INFO): Define.
2304 * config/nds32/elf.h: New file.
2305 * config/nds32/linux.h: New file.
2306 * config/nds32/nds32-elf.opt: New file.
2307 * config/nds32/nds32-linux.opt: New file.
2308 * config/nds32/nds32-fp-as-gp.c
2309 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2310 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2312 (nds32_asm_file_end): Ditto.
2313 (nds32_print_operand): Ditto.
2314 (nds32_insert_attributes): Ditto.
2315 (nds32_init_libfuncs): New function.
2316 (TARGET_HAVE_TLS): Define.
2317 (TARGET_INIT_LIBFUNCS): Define.
2318 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2320 (TARGET_ELF): Apply different mcmodel setting.
2321 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2322 been migrated into elf.h and linux.h files.
2323 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2324 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2325 (mcmodel): The content has been migrated into nds32-elf.opt and
2326 nds32-linux.opt files.
2327 * config/nds32/t-elf: New file.
2328 * config/nds32/t-linux: New file.
2330 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2331 Shiva Chen <shiva0217@gmail.com>
2333 * config/nds32/constants.md (unspec_volatile_element): Add
2334 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2335 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2337 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2338 (make_pass_nds32_fp_as_gp): Declare.
2339 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2341 (nds32_asm_function_end_prologue): Remove unused asm output.
2342 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2343 (nds32_asm_file_start): Output necessary fp_as_gp information.
2344 (nds32_option_override): Adjust register usage.
2345 (nds32_expand_prologue): Consider fp_as_gp situation.
2346 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2347 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2350 (simple_return): Ditto.
2351 (omit_fp_begin): Output special directive for fp_as_gp.
2352 (omit_fp_end): Output special directive for fp_as_gp.
2353 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2354 mforbid-fp-as-gp): New options.
2356 2018-06-01 Mark Wielaard <mark@klomp.org>
2358 * dwarf2out.c (dwarf2out_finish): Remove generation of
2359 DW_AT_loclists_base.
2361 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2363 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2364 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2365 (struct merged_store_group): Add bit_insertion field.
2366 (dump_char_array): Use standard hexadecimal format.
2367 (merged_store_group::merged_store_group): Set bit_insertion to false.
2368 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2369 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2370 also print the mask in the dump file.
2371 (pass_store_merging::gate): Minor tweak.
2372 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2373 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2374 stores with INTEGER_CST stores.
2375 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2376 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2377 and use it throughout. Generate bit insertion sequences if need be.
2378 (pass_store_merging::process_store): Remove redundant condition.
2379 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2381 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2383 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2384 the 128-bit floating point types. Fix function comment.
2386 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2388 * config/aarch64/aarch64-simd.md
2389 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2391 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2394 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2396 PR tree-optimization/85989
2397 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2399 (backprop::intersect_uses): Check it when deciding whether this
2400 is a backedge reference.
2401 (backprop::process_block): Add each phi to m_visited_phis
2402 after visiting it, then clear it at the end.
2404 2018-06-01 Richard Biener <rguenther@suse.de>
2406 * tree-vectorizer.h (vect_dr_stmt): New function.
2407 (vect_get_load_cost): Adjust.
2408 (vect_get_store_cost): Likewise.
2409 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2410 Use vect_dr_stmt instead of DR_SMTT.
2411 (vect_record_base_alignments): Likewise.
2412 (vect_calculate_target_alignment): Likewise.
2413 (vect_compute_data_ref_alignment): Likewise and make static.
2414 (vect_update_misalignment_for_peel): Likewise.
2415 (vect_verify_datarefs_alignment): Likewise.
2416 (vector_alignment_reachable_p): Likewise.
2417 (vect_get_data_access_cost): Likewise. Pass down
2418 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2419 (vect_get_peeling_costs_all_drs): Likewise.
2420 (vect_peeling_hash_get_lowest_cost): Likewise.
2421 (vect_enhance_data_refs_alignment): Likewise.
2422 (vect_find_same_alignment_drs): Likewise.
2423 (vect_analyze_data_refs_alignment): Likewise.
2424 (vect_analyze_group_access_1): Likewise.
2425 (vect_analyze_group_access): Likewise.
2426 (vect_analyze_data_ref_access): Likewise.
2427 (vect_analyze_data_ref_accesses): Likewise.
2428 (vect_vfa_segment_size): Likewise.
2429 (vect_small_gap_p): Likewise.
2430 (vectorizable_with_step_bound_p): Likewise.
2431 (vect_prune_runtime_alias_test_list): Likewise.
2432 (vect_analyze_data_refs): Likewise.
2433 (vect_supportable_dr_alignment): Likewise.
2434 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2435 (vect_gen_prolog_loop_niters): Likewise.
2436 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2437 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2439 (vect_recog_mask_conversion_pattern): Likewise.
2440 (vect_try_gather_scatter_pattern): Likewise.
2441 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2442 to vect_get_store_cost.
2443 (vect_get_store_cost): Get stmt_info instead of DR.
2444 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2445 (vect_get_load_cost): Get stmt_info instead of DR.
2447 2018-06-01 Richard Biener <rguenther@suse.de>
2450 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2451 vars wrapped in MEM_REFs.
2453 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2455 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2456 Fix subreg tests so that we only return a choice between
2457 GENERAL_REGS and FP_REGS if the original classes included both.
2459 2018-06-01 Richard Biener <rguenther@suse.de>
2462 * tree-ssa-structalias.c (get_function_part_constraint):
2463 Handle NULL fi->decl.
2464 (find_func_aliases_for_call): Properly handle indirect
2465 fi from direct call.
2466 (find_func_clobbers): Likewise.
2467 (ipa_pta_execute): Likewise.
2468 (create_variable_info_for): For functions that are ifunc_resolver
2469 resolve to a varinfo that contains the result of the resolver call.
2470 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2473 2018-05-31 Michael Collison <michael.collison@arm.com>
2475 * config/aarch64/aarch64.md:
2476 (*fix_to_zero_extenddfdi2): New pattern.
2477 * gcc.target/aarch64/fix_extend1.c: New testcase.
2479 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2483 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2484 and BUILT_IN_STRNCMP_EQ.
2485 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2486 BUILT_IN_STRNCMP_EQ.
2487 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2488 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2489 (gimple_fold_builtin): Likewise.
2490 * tree-ssa-strlen.c (compute_string_length): New function.
2491 (determine_min_obsize): New function.
2492 (handle_builtin_string_cmp): New function to handle calls to
2493 string compare functions.
2494 (strlen_optimize_stmt): Add handling to builtin string compare
2496 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2497 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2498 * tree.c (build_common_builtin_nodes): Add new defines of
2499 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2501 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2504 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2505 functions with naked attribute.
2507 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2509 * config/i386/sse.md (avx_vec_concat<mode>):
2510 Substitute concat_tg_mode mode attribute with xtg_mode.
2511 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2512 (concat_tg_mode): Remove mode attribute.
2514 2018-05-31 Martin Sebor <msebor@redhat.com>
2517 * calls.c (alloc_max_size): Correct a logic error/typo.
2518 Treat excessive arguments as infinite. Warn for invalid arguments.
2519 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2521 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2524 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2525 and movx for Haswell.
2527 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2528 Cesar Philippidis <cesar@codesourcery.com>
2531 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2532 when emitting error on private/firstprivate reductions.
2533 * omp-low.c (lower_omp_target): Avoid reference-type processing
2534 on pointers for firstprivate clause.
2536 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2538 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2541 * config/aarch64/aarch64-simd.md
2542 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2543 (aarch64_ld1_x3_<mode>): Likewise
2544 (aarch64_st1x2<VALLDIF:mode>): Likewise
2545 (aarch64_st1_x2_<mode>): Likewise
2546 (aarch64_st1x3<VALLDIF:mode>): Likewise
2547 (aarch64_st1_x3_<mode>): Likewise
2548 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2549 (vld1_s8_x3): Likewise.
2550 (vld1_u16_x3): Likewise.
2551 (vld1_s16_x3): Likewise.
2552 (vld1_u32_x3): Likewise.
2553 (vld1_s32_x3): Likewise.
2554 (vld1_u64_x3): Likewise.
2555 (vld1_s64_x3): Likewise.
2556 (vld1_f16_x3): Likewise.
2557 (vld1_f32_x3): Likewise.
2558 (vld1_f64_x3): Likewise.
2559 (vld1_p8_x3): Likewise.
2560 (vld1_p16_x3): Likewise.
2561 (vld1_p64_x3): Likewise.
2562 (vld1q_u8_x3): Likewise.
2563 (vld1q_s8_x3): Likewise.
2564 (vld1q_u16_x3): Likewise.
2565 (vld1q_s16_x3): Likewise.
2566 (vld1q_u32_x3): Likewise.
2567 (vld1q_s32_x3): Likewise.
2568 (vld1q_u64_x3): Likewise.
2569 (vld1q_s64_x3): Likewise.
2570 (vld1q_f16_x3): Likewise.
2571 (vld1q_f32_x3): Likewise.
2572 (vld1q_f64_x3): Likewise.
2573 (vld1q_p8_x3): Likewise.
2574 (vld1q_p16_x3): Likewise.
2575 (vld1q_p64_x3): Likewise.
2576 (vst1_s64_x2): Likewise.
2577 (vst1_u64_x2): Likewise.
2578 (vst1_f64_x2): Likewise.
2579 (vst1_s8_x2): Likewise.
2580 (vst1_p8_x2): Likewise.
2581 (vst1_s16_x2): Likewise.
2582 (vst1_p16_x2): Likewise.
2583 (vst1_s32_x2): Likewise.
2584 (vst1_u8_x2): Likewise.
2585 (vst1_u16_x2): Likewise.
2586 (vst1_u32_x2): Likewise.
2587 (vst1_f16_x2): Likewise.
2588 (vst1_f32_x2): Likewise.
2589 (vst1_p64_x2): Likewise.
2590 (vst1q_s8_x2): Likewise.
2591 (vst1q_p8_x2): Likewise.
2592 (vst1q_s16_x2): Likewise.
2593 (vst1q_p16_x2): Likewise.
2594 (vst1q_s32_x2): Likewise.
2595 (vst1q_s64_x2): Likewise.
2596 (vst1q_u8_x2): Likewise.
2597 (vst1q_u16_x2): Likewise.
2598 (vst1q_u32_x2): Likewise.
2599 (vst1q_u64_x2): Likewise.
2600 (vst1q_f16_x2): Likewise.
2601 (vst1q_f32_x2): Likewise.
2602 (vst1q_f64_x2): Likewise.
2603 (vst1q_p64_x2): Likewise.
2604 (vst1_s64_x3): Likewise.
2605 (vst1_u64_x3): Likewise.
2606 (vst1_f64_x3): Likewise.
2607 (vst1_s8_x3): Likewise.
2608 (vst1_p8_x3): Likewise.
2609 (vst1_s16_x3): Likewise.
2610 (vst1_p16_x3): Likewise.
2611 (vst1_s32_x3): Likewise.
2612 (vst1_u8_x3): Likewise.
2613 (vst1_u16_x3): Likewise.
2614 (vst1_u32_x3): Likewise.
2615 (vst1_f16_x3): Likewise.
2616 (vst1_f32_x3): Likewise.
2617 (vst1_p64_x3): Likewise.
2618 (vst1q_s8_x3): Likewise.
2619 (vst1q_p8_x3): Likewise.
2620 (vst1q_s16_x3): Likewise.
2621 (vst1q_p16_x3): Likewise.
2622 (vst1q_s32_x3): Likewise.
2623 (vst1q_s64_x3): Likewise.
2624 (vst1q_u8_x3): Likewise.
2625 (vst1q_u16_x3): Likewise.
2626 (vst1q_u32_x3): Likewise.
2627 (vst1q_u64_x3): Likewise.
2628 (vst1q_f16_x3): Likewise.
2629 (vst1q_f32_x3): Likewise.
2630 (vst1q_f64_x3): Likewise.
2631 (vst1q_p64_x3): Likewise.
2633 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2635 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2636 user_label_prefix to name.
2638 * tree-core.h: Update comment about the format of NAME string
2639 passed to handler in attribute_spec.
2641 * config/msp430/msp430.md: Remove erroneous subreg expression from
2642 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2643 zero_extend{q,h}isi2.
2645 2018-05-30 Borislav Petkov <bp@suse.de>
2647 * doc/extend.texi: Document some architecture specific
2648 constraints and sort entries.
2650 2018-05-30 Martin Sebor <msebor@redhat.com>
2653 * builtins.c (expand_builtin_stpcpy_1): New function.
2654 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2655 only if the former succeeds.
2657 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2659 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2662 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2664 * doc/invoke.texi (-flinker-output): Document
2666 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2668 * passes.c (ipa_write_summaries): Only modify statements if body
2670 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2671 incrementally linking.
2672 (ipa_passes): Likewise.
2673 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2674 pass down resolution info.
2675 * common.opt (flag_incremental_link): Update info.
2676 * gcc.c (plugin specs): Turn flinker-output=* to
2677 -plugin-opt=-linker-output-known
2678 * toplev.c (compile_file): Also cut compilation when doing incremental
2680 * flag-types. (enum lto_partition_model): Add
2681 LTO_LINKER_OUTPUT_NOLTOREL.
2682 (invoke.texi): Add -flinker-output docs.
2683 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2684 link same way as WPA; do not stream in dead initializers.
2686 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2687 darf2out_register_external_decl): Support incremental link.
2689 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2691 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2693 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2695 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2696 it down to simple_object_copy_lto_debug_sections.
2697 (run_gcc): Determine incremental LTO link time and configure
2698 lto1 into non-wpa mode, disable renaming of debug sections.
2700 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2702 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2703 descriptions of various incorrectly documented functions.
2705 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2708 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2709 address check not strict.
2711 2018-05-30 Richard Biener <rguenther@suse.de>
2713 PR tree-optimization/85964
2714 * tracer.c (better_p): Drop initialized count check, we only
2715 call the function with initialized counts now.
2716 (find_best_successor): Do find a best edge if one
2717 has uninitialized count.
2718 (find_best_predecessor): Likewise. Do BB frequency check only
2719 if count is initialized.
2721 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2723 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2724 (aarch64_ldrstr_offset_compare): New.
2725 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2726 load/store orderings.
2727 (aarch64_gen_adjusted_ldpstp): Likewise.
2729 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2731 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2732 Check for subset of GENERAL_REGS and FP_REGS.
2733 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2736 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2738 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2739 and wi::to_poly_offset. Add the current offset and then check
2740 whether the sum fits, rather than using an unchecked addition of
2741 a checked term. Check for a shwi rather than a uhwi.
2742 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2743 (store_constructor): Use poly_int_tree_p.
2744 (expand_expr_real_1): Likewise.
2745 * function.c (assign_temp): Likewise.
2746 * fold-const.c (const_binop): Use poly_int_tree_p and
2748 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2750 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2751 to_poly_offset for MEM offsets.
2752 * ipa-icf.c (sem_variable::equals): Likewise.
2753 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2754 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2755 wi::to_poly_offset for BIT_FIELD_REF offsets.
2756 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2758 * var-tracking.c (emit_note_insn_var_location): Use
2759 tree_to_poly_uint64.
2761 2018-05-29 Jim Wilson <jimw@sifive.com>
2763 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2765 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2768 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2769 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2771 (sse4_1_round<mode>2): Use nonimmediate_operand
2772 for operand 1 predicate.
2774 2018-05-29 Martin Sebor <msebor@redhat.com>
2775 Richard Biener <rguenther@suse.de>
2778 * calls.c (get_size_range): Call determine_value_range instead
2779 of get_value_range..
2780 * tree-vrp.h (determine_value_range): Declared new function.
2781 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2783 2018-05-29 Richard Biener <rguenther@suse.de>
2785 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2786 sure to use non-pattern stmts for get_earlier_stmt arguments.
2787 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2788 called on pattern stmts.
2789 (get_later_stmt): Likewise.
2791 2018-05-29 Martin Liska <mliska@suse.cz>
2793 PR gcov-profile/85759
2794 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2797 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2799 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2801 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2805 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2806 VEC_PACK_FLOAT_EXPR): New tree codes.
2807 * tree-pretty-print.c (op_code_prio): Handle
2808 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2809 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2810 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2811 * tree-inline.c (estimate_operator_cost): Likewise.
2812 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2813 * fold-const.c (const_binop): Likewise.
2814 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2815 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2816 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2817 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2818 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2819 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2820 * expr.c (expand_expr_real_2): Likewise.
2821 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2822 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2823 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2825 * optabs.c (expand_widen_pattern_expr): For
2826 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2827 sign from result type rather than operand's type.
2828 (expand_binop_directly): For vec_packu_float_optab and
2829 vec_packs_float_optab allow result type to be different from operand's
2831 * optabs-tree.c (optab_for_tree_code): Handle
2832 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2833 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2834 * tree-vect-generic.c (expand_vector_operations_1): Handle
2835 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2836 VEC_PACK_FLOAT_EXPR.
2837 * tree-vect-stmts.c (supportable_widening_operation): Handle
2839 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2840 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2841 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2842 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2843 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2845 (vec_pack<floatprefix>_float_<mode>): New expander.
2846 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2848 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2849 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2850 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2851 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2852 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2854 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2855 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2856 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2857 VEC_PACK_FLOAT_EXPR): Document.
2859 2018-05-29 Richard Biener <rguenther@suse.de>
2861 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2863 (stmt_vec_info_vec): Make pointer.
2864 (init_stmt_vec_info_vec): Remove.
2865 (free_stmt_vec_info_vec): Likewise.
2866 (set_stmt_vec_info_vec): New function.
2867 (free_stmt_vec_infos): Likewise.
2868 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2869 (set_vinfo_for_stmt): Likewise.
2870 (get_earlier_stmt): Likewise.
2871 (get_later_stmt): Likewise.
2872 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2873 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2874 (vec_info::~vec_info): Free stmt_vec_infos.
2875 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2876 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2877 (pass_slp_vectorize::execute): Likewise.
2878 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2879 (free_stmt_vec_info_vec): Likewise.
2880 (set_stmt_vec_info_vec): New function.
2881 (free_stmt_vec_infos): Likewise.
2882 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2883 the global stmt_vec_info_vec.
2884 * tree-parloops.c (gather_scalar_reductions): Use
2885 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2888 2018-05-29 Richard Biener <rguenther@suse.de>
2890 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2892 2018-05-29 Martin Liska <mliska@suse.cz>
2893 David Malcolm <dmalcolm@redhat.com>
2895 * vec.c (test_reverse): New.
2896 (vec_c_tests): Add new test.
2897 * vec.h (vl_ptr>::reverse): New function.
2899 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2901 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2903 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2906 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2908 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2910 2018-05-28 Richard Biener <rguenther@suse.de>
2912 PR tree-optimization/85933
2913 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2914 look at stmts marked as vectorizable.
2916 2018-05-28 Richard Biener <rguenther@suse.de>
2918 PR tree-optimization/85934
2919 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2920 vector boolean check before scalar optimization.
2922 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2924 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2927 2018-05-28 Mark Wielaard <mark@klomp.org>
2929 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2930 if it is an expression containing a minus sign.
2932 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2934 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2936 2018-05-27 Paul Koning <ni1d@arrl.net>
2938 * config/pdp11/pdp11.md (truncsihi2): Remove.
2940 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2941 Chung-Ju Wu <jasonwucj@gmail.com>
2943 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2945 (unaligned_store_dw): Ditto.
2946 * config/nds32/nds32-memory-manipulation.c
2947 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2948 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2949 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2950 (emit_setmem_word_loop): Rename to ...
2951 (emit_setmem_doubleword_loop): ... this.
2952 (nds32_gen_dup_4_byte_to_word_value): New function.
2953 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2954 (nds32_expand_setmem_loop): Refine implementation.
2955 (nds32_expand_setmem_loop_v3m): Ditto.
2956 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2959 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2961 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2963 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2966 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2967 machine->attr_no_prologue_p.
2968 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2969 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2970 (nds32_expand_epilogue): Consider attr_naked_p.
2971 (nds32_expand_epilogue_v3pop): Likewise.
2972 (nds32_can_use_return_insn): Likewise.
2973 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2974 attr_no_prologue_p fields.
2975 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2977 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2980 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2982 * config/i386/sse.md
2983 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2985 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2987 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2989 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2991 (*<floatsuffix>floatv2div2sf2): Rename to ...
2992 (*float<floatunssuffix>v2div2sf2): ... this.
2993 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2994 (float<floatunssuffix>v2div2sf2_mask): ... this.
2995 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2996 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2997 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2999 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3001 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3003 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3005 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3007 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3009 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3010 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3011 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3012 gen_ufix_truncv8dfv8si2.
3013 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3014 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3015 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3016 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3017 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3018 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3019 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3020 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3022 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3026 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3028 2018-05-25 Jim Wilson <jimw@sifive.com>
3030 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3031 * config/riscv/riscv.c (struct machine_function): Add
3032 interrupt_handler_p and attribute_checked_p fields.
3033 (riscv_attribute_table): Add interrupt.
3034 (riscv_interrupt_type_p): New.
3035 (riscv_save_reg_p): Save extra regs for interrupt handler.
3036 (riscv_use_save_libcall): Return false for interrupt handler.
3037 (riscv_first_stack_step): Add forward declaration.
3038 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3039 for interrupt handler with large frame. Use it for saved reg list.
3040 (riscv_expand_prologue): Move flag_stack_usage_info support to
3041 eliminate duplication.
3042 (riscv_expand_epilogue): Generate mret for interrupt handler.
3043 (riscv_epilogue_uses): New.
3044 (riscv_can_use_return_insn): Return false for interrupt handler.
3045 (riscv_function_ok_for_sibcall): Likewise.
3046 (riscv_set_current_function): Add interrupt handler support.
3047 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3048 * config/riscv/riscv.md (UNSPECV_MRET): New.
3050 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3052 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3054 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3056 PR tree-optimization/85712
3057 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3058 this candidate has already been replaced in-situ by a copy.
3060 2018-05-25 Jason Merrill <jason@redhat.com>
3062 PR c++/80485 - inline function non-zero address.
3063 * symtab.c (nonzero_address): Check DECL_COMDAT.
3065 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3068 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3069 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3070 extension of SImode operation. Use const123_operand predicate.
3071 (*saddsi_1): Remove.
3072 (*saddl_se_1): Ditto.
3074 (*ssubl_se_1): Ditto.
3075 * config/alpha/predicates.md (const123_operand): New predicate.
3076 * config/alpha/constraints.md (P): Use IN_RANGE.
3078 2018-05-25 Richard Biener <rguenther@suse.de>
3080 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3082 (ref_maybe_used_by_stmt_p): Likewise.
3083 (stmt_may_clobber_ref_p): Likewise.
3084 (stmt_may_clobber_ref_p_1): Likewise.
3085 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3087 (ref_maybe_used_by_stmt_p): Likewise.
3088 (stmt_may_clobber_ref_p): Likewise.
3089 (stmt_may_clobber_ref_p_1): Likewise.
3090 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3091 the alias oracle to disambiguate DRs with stmts DR analysis
3093 (vect_analyze_data_refs): Do not give up on not analyzable
3094 DRs for BB vectorization. Remove code truncating the dataref
3097 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3100 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3101 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3102 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3104 2018-05-25 Richard Biener <rguenther@suse.de>
3106 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3107 function, combining stmt data ref gathering and fatal analysis
3109 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3110 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3111 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3112 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3113 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3115 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3117 PR tree-optimization/85720
3118 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3119 SCC if all partitions are builtins.
3120 (version_loop_by_alias_check): New parameter. Generate cancelable
3121 runtime alias check if all partitions are builtins.
3122 (distribute_loop): Update call to above function.
3124 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3126 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3127 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3128 (parm_default_def_partition_arg): Ditto.
3129 (set_parm_default_def_partition): Ditto.
3130 (get_parm_default_def_partitions): Ditto and make it static.
3131 (get_undefined_value_partitions): Ditto and make it static.
3132 (remove_ssa_form): Refactor call to init_var_map here.
3133 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3134 computation for loop region.
3135 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3136 (register_default_def): Delete.
3137 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3138 (parm_default_def_partition_arg): Ditto.
3139 (set_parm_default_def_partition): Ditto.
3140 (get_parm_default_def_partitions): Ditto and make it static.
3141 (get_undefined_value_partitions): Ditto and make it static.
3142 (coalesce_with_default, coalesce_with_default): Update comment.
3143 (create_coalesce_list_for_region): New func factored out from
3144 create_outofssa_var_map.
3145 (populate_coalesce_list_for_outofssa): New func factored out from
3146 create_outofssa_var_map and coalesce_ssa_name.
3147 (create_outofssa_var_map): Delete.
3148 (coalesce_ssa_name): Refactor to support live range computation.
3149 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3150 (get_parm_default_def_partitions): Delete.
3151 (get_undefined_value_partitions): Ditto.
3152 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3153 computation for loop region.
3154 (new_tree_live_info, loe_visit_block): Ditto.
3155 (live_worklist, set_var_live_on_entry): Ditto.
3156 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3157 * tree-ssa-live.h (struct _var_map): New fields.
3158 (init_var_map): Change decl.
3159 (region_contains_p): New.
3161 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3163 * tree-ssa-live.h (live_merge_and_clear): Delete.
3165 2018-05-25 Richard Biener <rguenther@suse.de>
3168 * tree-dump.c (dequeue_and_dump): Remove access to removed
3169 operand 2 of a SWITCH_EXPR.
3171 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3173 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3174 multiplication and division.
3175 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3176 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3177 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3178 (cond_udiv_optab, cond_umod_optab): New optabs.
3179 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3180 (IFN_COND_RDIV): New internal functions.
3181 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3182 TRUNC_MOD_EXPR and RDIV_EXPR.
3183 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3184 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3186 (SVE_INT_BINARY): Include mult.
3187 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3188 (optab, sve_int_op): Handle mult.
3189 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3191 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3192 for SVE_INT_BINARY_SD.
3194 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3196 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3197 (optab, sve_int_op): Handle div and udiv.
3198 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3199 for SVE_INT_BINARY_SD.
3200 (*<optab><mode>3): New insn for the same.
3202 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3204 * tree-vect-patterns.c: Include predict.h.
3205 (vect_recog_divmod_pattern): Restrict check for division support
3206 to when optimizing for size.
3208 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3210 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3211 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3212 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3213 (gimple_match_op::set_op): Likewise.
3214 (gimple_resimplify4): Declare.
3215 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3216 (expr::gen_transform): Likewise.
3217 (decision_tree::gen): Generate a simplification routine for 4 operands.
3218 * gimple-match-head.c (gimple_simplify): Add an overload for
3219 4 operands. In the top-level function, handle up to 4 call
3220 arguments and call gimple_resimplify4.
3221 (gimple_resimplify4): New function.
3222 (build_call_internal): Pass a fourth operand.
3223 (maybe_push_to_seq): Likewise.
3224 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3225 Fold VEC_COND_EXPRs of an operation and a default value into
3226 an IFN_COND_* function if possible.
3227 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3229 (SVE_COND_FP_BINARY): Include them.
3230 (optab, sve_fp_op): Handle them.
3231 (SVE_INT_BINARY_REV): New code iterator.
3232 (SVE_COND_FP_BINARY_REV): New int iterator.
3233 (commutative): New int attribute.
3234 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3236 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3238 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3239 (*cond_<optab><mode>): New patterns for reversed operands.
3241 2018-05-25 Richard Biener <rguenther@suse.de>
3243 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3244 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3245 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3246 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3247 * tree-vect-data-refs.c (everywhere): Adjust users.
3248 * tree-vect-loop.c (everywhere): Likewise.
3249 * tree-vect-slp.c (everywhere): Likewise.
3250 * tree-vect-stmts.c (everywhere): Likewise.
3251 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3253 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3255 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3257 (gcc_cv_as_section_exclude): ... this.
3258 Try Solaris as #exclude syntax.
3259 * configure: Regenerate.
3260 * config.in: Regenerate.
3261 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3263 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3264 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3266 * varasm.c (default_elf_asm_named_section): Don't check if
3267 HAVE_GAS_SECTION_EXCLUDE is defined.
3269 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3271 * doc/md.texi: Update the documentation of the cond_* optabs
3272 to mention the new final operand. Fix GET_MODE_NUNITS call.
3273 Describe the scalar case too.
3274 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3275 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3277 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3278 (get_conditional_internal_fn): Update comment.
3279 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3280 accumulator value as a final argument to conditional functions.
3281 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3282 a define_expand and add an "else" operand. Assert for now that
3283 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3284 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3285 (*cond_<optab><mode>): New patterns.
3286 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3287 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3288 (UNSPEC_COND_EOR): Delete.
3289 (optab): Remove associated mappings.
3290 (SVE_INT_BINARY): New code iterator.
3291 (sve_int_op): Remove int attribute and add "minus" to the code
3293 (SVE_COND_INT_OP): Delete.
3294 (SVE_COND_FP_OP): Rename to...
3295 (SVE_COND_FP_BINARY): ...this.
3297 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3299 * optabs.c (can_reuse_operands_p): New function.
3300 (maybe_legitimize_operands): Try to reuse the results for
3303 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3305 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3306 Add {q} suffix to insn mnemonic.
3308 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3310 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3311 (msp430_warn_func_return): New.
3313 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3315 * fold-const.c (tree_nonzero_bits): New function.
3316 * fold-const.h (tree_nonzero_bits): Likewise.
3317 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3318 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3320 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3324 * varasm.c (assemble_alias): Check ifunc_resolver only on
3327 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3330 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3331 when memory input operand is handled.
3333 2018-05-24 Luis Machado <luis.machado@linaro.org>
3335 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3337 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3339 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3341 * match.pd: Delay FMA folds until after vectorization.
3343 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3346 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3347 address check not strict.
3349 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3351 * gimple-match.h (gimple_match_op): New class.
3352 (mprts_hook): Replace parameters with a gimple_match_op *.
3353 (maybe_build_generic_op): Likewise.
3354 (gimple_simplified_result_is_gimple_val): Replace parameters with
3355 a const gimple_match_op *.
3356 (gimple_simplify): Replace code_helper * and tree * parameters with
3357 a gimple_match_op * parameter.
3358 (gimple_resimplify1): Replace code_helper *, tree and tree *
3359 parameters with a gimple_match_op * parameter.
3360 (gimple_resimplify2): Likewise.
3361 (gimple_resimplify3): Likewise.
3362 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3363 parameters with a gimple_match_op * parameter.
3364 * gimple-match-head.c (gimple_simplify): Change prototypes of
3365 auto-generated functions to take a gimple_match_op * instead of
3366 separate code_helper * and tree * parameters. Make the same
3367 change in the top-level overload and update calls to the
3368 gimple_resimplify routines. Update calls to the auto-generated
3369 functions and to maybe_push_res_to_seq in the publicly-facing
3370 operation-specific gimple_simplify overloads.
3371 (gimple_match_op::MAX_NUM_OPS): Define.
3372 (gimple_resimplify1): Replace rcode and ops with a single res_op
3373 parameter. Update call to gimple_simplify.
3374 (gimple_resimplify2): Likewise.
3375 (gimple_resimplify3): Likewise.
3376 (mprts_hook): Replace parameters with a gimple_match_op *.
3377 (maybe_build_generic_op): Likewise.
3378 (build_call_internal): Replace type, nargs and ops with
3379 a gimple_match_op *.
3380 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3381 with a single gimple_match_op *. Update calls to mprts_hook,
3382 build_call_internal and gimple_simplified_result_is_gimple_val.
3383 Factor out code that is common to the tree_code and combined_fn cases.
3384 * genmatch.c (expr::gen_transform): Replace tem_code and
3385 tem_ops with a gimple_match_op called tem_op. Update calls
3386 to the gimple_resimplify functions and maybe_push_res_to_seq.
3387 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3388 res_ops. Update call to the gimple_resimplify functions.
3389 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3390 (decision_tree::gen): Make the functions take a gimple_match_op *
3391 called res_op instead of separate res_code and res_ops parameters.
3392 Update call accordingly.
3393 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3394 and ops with a single res_op parameter. Update calls to
3395 maybe_build_generic_op and maybe_push_res_to_seq.
3396 (fold_stmt_1): Update calls to gimple_simplify and
3397 replace_stmt_with_simplification.
3398 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3399 and gimple_simplified_result_is_gimple_val.
3400 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3402 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3403 with a gimple_match_op *.
3404 (vn_nary_build_or_lookup): Likewise. Update call to
3405 vn_nary_build_or_lookup_1.
3406 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3407 gimple_match_op *. Update calls to the gimple_resimplify routines
3408 and to gimple_simplified_result_is_gimple_val.
3409 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3410 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3411 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3412 (visit_nary_op): Likewise.
3413 (visit_reference_op_load): Likewise.
3415 2018-05-23 Luis Machado <luis.machado@linaro.org>
3417 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3418 modifier for printing the step amount.
3420 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3423 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3426 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3428 * doc/sourcebuild.texi (Endianness): New subsubsection.
3430 2018-05-23 Luis Machado <luis.machado@linaro.org>
3432 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3433 <prefetch_dynamic_strides>: New const bool field.
3434 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3435 prefetch_dynamic_strides.
3436 (exynosm1_prefetch_tune): Likewise.
3437 (thunderxt88_prefetch_tune): Likewise.
3438 (thunderx_prefetch_tune): Likewise.
3439 (thunderx2t99_prefetch_tune): Likewise.
3440 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3442 (aarch64_override_options_internal): Update to set
3443 PARAM_PREFETCH_DYNAMIC_STRIDES.
3444 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3445 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3446 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3447 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3448 prefetch-dynamic-strides setting.
3450 2018-05-23 Luis Machado <luis.machado@linaro.org>
3452 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3453 <minimum_stride>: New const int field.
3454 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3455 minimum_stride field defaulting to -1.
3456 (exynosm1_prefetch_tune): Likewise.
3457 (thunderxt88_prefetch_tune): Likewise.
3458 (thunderx_prefetch_tune): Likewise.
3459 (thunderx2t99_prefetch_tune): Likewise.
3460 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3461 <default_opt_level>: Set to 3.
3462 (aarch64_override_options_internal): Update to set
3463 PARAM_PREFETCH_MINIMUM_STRIDE.
3464 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3465 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3466 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3467 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3468 stride is constant and is below the minimum stride threshold.
3470 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3472 * config/arm/arm-cpus.in (mode26): Delete.
3473 (armv4): Delete mode26 reference.
3474 * config/arm/arm.c (arm_configure_build_target): Delete use of
3477 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3479 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3481 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3482 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3484 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3485 (floatunsdidf2): Ditto.
3487 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3489 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3490 (fixuns_trunc<mode>si2_avx512f): Ditto.
3491 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3492 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3493 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3495 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3497 PR rtl-optimization/79985
3498 * df-scan.c (df_insn_refs_collect): Remove special case for
3499 global registers and asm statements.
3501 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3503 * extend.texi (Global Register Variables): Rewrite the bullet list.
3504 Note that the register is available for allocation. Note that access
3505 via inline asm must use constraints. Add note about async-signal
3506 handlers. Remove paragraph about automagic register selection.
3508 2018-05-23 Richard Biener <rguenther@suse.de>
3510 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3511 of fixed offset from memset VN.
3513 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3515 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3517 (alloc_cand_and_find_basis): Initialize first_interp field.
3518 (slsr_process_mul): Modify first_interp field.
3519 (slsr_process_add): Likewise.
3520 (slsr_process_cast): Modify first_interp field for each new
3522 (slsr_process_copy): Likewise.
3523 (dump_candidate): Dump first_interp field.
3524 (replace_mult_candidate): Process all interpretations, not just
3526 (replace_rhs_if_not_dup): Likewise.
3527 (replace_one_candidate): Likewise.
3529 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3531 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3533 (aarch64_needs_frame_chain): New function.
3534 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3536 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3539 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3540 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3541 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3542 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3543 as true for strict-align.
3544 (aarch64_can_inline_p): Perform checks even when callee has no
3545 attributes to check for strict alignment.
3546 * doc/extend.texi (AArch64 Function Attributes): Document
3548 * doc/invoke.texi: (AArch64 Options): Likewise.
3550 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3552 PR tree-optimization/85853
3553 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3554 the handling of the root of the node to...
3555 (vect_slp_analyze_node_operations_1): ...this new function,
3556 and run the whole thing with the child nodes' def types
3557 set according to their SLP node's def type.
3559 2018-05-23 Richard Biener <rguenther@suse.de>
3562 * tree-data-ref.c (create_runtime_alias_checks): Defer
3563 and ignore overflow warnings.
3565 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3567 PR tree-optimization/85822
3568 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3571 2018-05-23 Richard Biener <rguenther@suse.de>
3573 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3574 memset constants via native_interpret_expr.
3576 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3579 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3581 (cgraph_node::create_alias): Likewise.
3582 (cgraph_node::get_availability): Check ifunc_resolver instead
3583 of looking up ifunc attribute.
3584 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3585 * varasm.c (do_assemble_alias): Likewise.
3586 (assemble_alias): Likewise.
3587 (default_binds_local_p_3): Likewise.
3588 * cgraph.h (cgraph_node): Add ifunc_resolver.
3589 (cgraph_node::only_called_directly_or_aliased_p): Return false
3591 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3593 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3594 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3595 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3596 instead of looking up ifunc attribute.
3598 2018-05-22 Luis Machado <luis.machado@linaro.org>
3600 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3602 2018-05-22 Martin Sebor <msebor@redhat.com>
3605 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3606 only when expasion succeeds.
3607 (expand_builtin_strcmp): Same.
3608 (expand_builtin_strncmp): Same.
3610 2018-05-22 Martin Sebor <msebor@redhat.com>
3612 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3614 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3615 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3617 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3618 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3620 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3621 Add check for legitimate_address.
3622 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3623 (aarch64_swap_ldrstr_operands): New.
3624 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3627 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3628 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3630 * config/aarch64/aarch64.md: New patterns to generate stp
3632 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3633 single words and double words.
3634 (load_pair_sw, load_pair_dw): Likewise.
3635 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3637 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3639 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3640 for different mode ldpstp and add peephole for merged zero stores.
3642 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3644 (aarch64_gen_store_pair): Rename calls to match new patterns.
3645 (aarch64_gen_load_pair): Rename calls to match new patterns.
3646 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3647 (load_pair<DREG:mode><DREG2:mode>): ... This.
3648 (store_pair<mode>): Rename to...
3649 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3650 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3652 (V_INT_EQUIV): Handle SImode.
3653 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3656 2018-05-22 Martin Sebor <msebor@redhat.com>
3659 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3660 or ajust the presumed bound on an operation to avoid unnecessary
3663 2018-05-22 Martin Sebor <msebor@redhat.com>
3665 PR tree-optimization/85826
3666 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3667 assuming that a DECL necesarily has a constant size.
3669 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3672 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3674 2018-05-22 Richard Biener <rguenther@suse.de>
3676 PR tree-optimization/85834
3677 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3678 non-constant and non-zero memset arguments.
3680 2018-05-22 Martin Liska <mliska@suse.cz>
3683 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3685 2018-05-22 Richard Biener <rguenther@suse.de>
3687 PR tree-optimization/85863
3688 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3689 comparisons when vectype is specified.
3690 (vectorizable_condition): Do not specify vectype for
3691 vect_is_simple_cond when SLP vectorizing.
3693 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3696 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3697 define __ibm128 as long double.
3698 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3699 as a distinct type when IEEE 128-bit support is enabled.
3700 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3701 128-bit types to use the correct functions.
3702 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3703 convert between 128-bit floating point types that have different
3704 modes but the same representation, instead of using gen_lowpart to
3706 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3708 (IFKF_reg): New attributes to give the register constraints for
3710 (extend<mode>tf2_internal): New insns to mark an explicit
3711 conversion between 128-bit floating point types that have a
3712 different mode but share the same representation.
3714 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3716 PR tree-optimization/85814
3717 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3718 a null return from get_strinfo when unsharing the next
3719 strinfo in the chain.
3721 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3724 * varasm.c (weak_finish): Clean up weak_decls.
3726 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3728 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3729 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3730 UNSPEC_UADALP values.
3731 * config/aarch64/iterators.md (ABAL): New int iterator.
3734 (sur): Add mappings for the above.
3735 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3737 (aarch64_<sur>abal<mode>_4): Likewise.
3738 (aarch64_<sur>adalp<mode>_3): Likewise.
3739 (<sur>sadv16qi): New define_expand.
3741 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3743 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3744 (*movdf_internal): Ditto.
3745 (*rcpsf2_sse): Ditto.
3746 (*rsqrtsf2_sse): Ditto.
3747 (*sqrt<mode>2_sse): Ditto.
3749 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3751 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3753 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3754 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3755 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3756 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3758 * config/aarch64/arm_neon.h: Likewise.
3759 * config/aarch64/iterators.md (VQ_I): New.
3761 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3763 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3765 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3768 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3770 * config/nds32/constraints.md (S): New constraint.
3771 * config/nds32/nds32.md (call_internal): Use constraint S.
3772 (call_value_internal): Likewise.
3773 (sibcall_internal): Likewise.
3774 (sibcall_value_internal): Likewise.
3776 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3777 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3782 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3783 Chung-Ju Wu <jasonwucj@gmail.com>
3785 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3786 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3787 (nds32_rtx_costs_impl): Simplify.
3788 (nds32_address_cost_impl): Simplify.
3789 (nds32_init_rtx_costs): New function.
3790 (nds32_rtx_costs_speed_prefer): Likewise.
3791 (nds32_rtx_costs_size_prefer): Likewise.
3792 (nds32_address_cost_speed_prefer): Likewise.
3793 (nds32_address_cost_speed_fwprop): Likewise.
3794 (nds32_address_cost_size_prefer): Likewise.
3795 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3796 * config/nds32/nds32.c (nds32_option_override): Use
3797 nds32_init_rtx_costs function.
3799 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3801 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3802 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3803 (TARGET_PIPELINE_N8): Likewise.
3804 (TARGET_PIPELINE_N10): Likewise.
3805 (TARGET_PIPELINE_N13): Likewise.
3806 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3808 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3810 * config/nds32/nds32-fpu.md: Update copyright year.
3812 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3814 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3816 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3818 * config/nds32/nds32.c
3819 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3820 * config/nds32/nds32.opt (minline-asm-r15): New option.
3822 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3824 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3826 * config/nds32/nds32.md (abssi2): New pattern.
3828 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3830 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3831 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3832 Merge insn pattern from sse_cvtsi2ss<round_name> and
3833 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3834 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3835 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3836 using SWI48 mode iterator.
3837 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3838 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3839 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3840 pattern from sse_cvttss2si<round_saeonly_name>
3841 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3842 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3843 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3844 using SWI48 mode iterator.
3845 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3846 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3847 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3848 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3849 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3850 using SWI48 mode iterator.
3851 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3852 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3853 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3854 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3855 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3856 SWI48 mode iterator.
3857 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3858 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3859 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3860 pattern from sse_cvttsd2si<round_saeonly_name>
3861 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3863 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3865 * config/nds32/nds32-md-auxiliary.c
3866 (nds32_valid_smw_lwm_base_p): Refine.
3867 (nds32_output_smw_single_word): Refine.
3868 (nds32_output_smw_double_word): New.
3869 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3871 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3873 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3874 (nds32_output_stack_pop): Refine.
3875 (nds32_expand_unaligned_load): Refine.
3876 (nds32_expand_unaligned_store): Refine.
3878 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3879 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/constants.md: Add TP_REGNUM constant.
3882 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3883 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3885 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3886 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3887 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3888 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3889 and PIC code generation.
3890 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3892 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3894 * config/nds32/nds32.md: Support TLS and PIC.
3895 * config/nds32/nds32.c: Support TLS and PIC.
3896 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3897 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3900 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3903 mode with E_ prefix.
3905 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3906 Chung-Ju Wu <jasonwucj@gmail.com>
3908 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3909 * config/nds32/nds32-md-auxiliary.c
3910 (symbolic_reference_mentioned_p): New.
3911 (nds32_legitimize_ict_address): New.
3912 (nds32_expand_ict_move): New.
3913 (nds32_indirect_call_referenced_p): New.
3914 (nds32_symbol_binds_local_p): Delete.
3915 (nds32_long_call_p): Modify.
3916 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3917 * config/nds32/nds32-protos.h
3918 (symbolic_reference_mentioned_p): Declare.
3919 (nds32_legitimize_ict_address): Declare.
3920 (nds32_expand_ict_move): Declare.
3921 (nds32_indirect_call_referenced_p): Declare.
3922 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3923 (nds32_relax_group): Use nds32_ict_const_p as condition.
3924 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3925 (nds32_asm_file_start): Output ict_model directive in asm code.
3926 (nds32_legitimate_address_p): Consider indirect call.
3927 (nds32_print_operand): Consider indirect call.
3928 (nds32_print_operand_address): Consider indirect call.
3929 (nds32_insert_attributes): Handle "indirect_call" attribute.
3930 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3931 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3932 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3933 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3934 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3935 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3936 (TARGET_ICT_MODEL_SMALL): Define.
3937 (TARGET_ICT_MODEL_LARGE): Define.
3938 * config/nds32/nds32.md (movsi): Consider ict model.
3939 (call, call_value): Consider ict model.
3940 (sibcall, sibcall_value): Consider ict model.
3941 * config/nds32/nds32.opt (mict-model): New option.
3942 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3945 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3946 Monk Chiang <sh.chiang04@gmail.com>
3947 Jim Wilson <jimw@sifive.com>
3949 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3950 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3951 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3952 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3953 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3954 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3955 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3956 compute save_libcall_adjustment properly.
3957 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3958 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3959 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3960 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3961 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3962 (ABI_SPEC): Handle mabi=ilp32e.
3963 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3964 (RVE): Add RVE mask.
3965 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3966 <-march>: Add rv32e as an example.
3968 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3971 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3972 (intra_create_variable_infos): Handle C++ constructors.
3974 2018-05-18 Martin Liska <mliska@suse.cz>
3976 * passes.def: Remove a redundant pass.
3978 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3981 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3983 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3985 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3987 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3989 (armv2, armv2a, armv3, armv3m): Delete architectures.
3990 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3991 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3992 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3994 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3995 (*mulsidi3adddi): Likewise.
3996 (mulsidi3): Likewise.
3997 (*mulsidi3_nov6): Likewise.
3998 (umulsidi3): Likewise.
3999 (umulsidi3_nov6): Likewise.
4000 (umaddsidi4): Likewise.
4001 (*umulsidi3adddi): Likewise.
4002 (smulsi3_highpart): Likewise.
4003 (*smulsi3_highpart_nov6): Likewise.
4004 (umulsi3_highpart): Likewise.
4005 (*umulsi3_highpart_nov6): Likewise.
4006 * config/arm/arm.h (arm_arch3m): Delete.
4007 * config/arm/arm.c (arm_arch3m): Delete.
4008 (arm_option_override_internal): Update armv3-related comment.
4009 (arm_configure_build_target): Delete use of isa_bit_mode32.
4010 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4011 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4012 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4014 (mulusa3): Likewise.
4015 * config/arm/arm-protos.h (arm_arch3m): Delete.
4016 * config/arm/arm-tables.opt: Regenerate.
4017 * config/arm/arm-tune.md: Likewise.
4018 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4019 deleted architectures.
4021 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4023 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4024 (armv5t, armv5te): New features.
4025 (ARMv5, ARMv5e): Delete fgroups.
4026 (ARMv5t, ARMv5te): Adjust for above changes.
4028 (armv5, armv5e): Delete arches.
4029 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4031 (*call_reg_arm): Likewise.
4032 (*call_value_reg_armv5): Likewise.
4033 (*call_value_reg_arm): Likewise.
4034 (*call_symbol): Likewise.
4035 (*call_value_symbol): Likewise.
4036 (*sibcall_insn): Likewise.
4037 (*sibcall_value_insn): Likewise.
4039 (prefetch): Likewise.
4040 (define_split and define_peephole2 dependent on arm_arch5):
4042 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4044 (TARGET_ARM_QBIT): Likewise.
4045 (TARGET_DSP_MULTIPLY): Likewise.
4046 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4047 (arm_arch5, arm_arch5e): Delete.
4048 (arm_arch5t, arm_arch5te): Declare.
4049 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4050 (arm_arch5t): Declare.
4051 (arm_option_reconfigure_globals): Update for the above.
4052 (arm_options_perform_arch_sanity_checks): Update comment, replace
4053 use of arm_arch5 with arm_arch5t.
4054 (use_return_insn): Likewise.
4055 (arm_emit_call_insn): Likewise.
4056 (output_return_instruction): Likewise.
4057 (arm_final_prescan_insn): Likewise.
4058 (arm_coproc_builtin_available): Likewise.
4059 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4060 arm_arch5e with arm_arch5t and arm_arch5te.
4061 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4062 (arm_arch5t, arm_arch5te): Declare.
4063 * config/arm/arm-tables.opt: Regenerate.
4064 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4065 * config/arm/t-multilib: Likewise.
4066 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4067 instead of arm_arch5.
4068 (*call_reg_thumb1): Likewise.
4069 (*call_value_reg_thumb1_v5): Likewise.
4070 (*call_value_reg_thumb1): Likewise.
4071 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4073 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4075 2018-05-18 Martin Liska <mliska@suse.cz>
4077 PR gcov-profile/84846
4078 * doc/gcov.texi: Document -t option of gcov tool.
4080 2018-05-18 Martin Liska <mliska@suse.cz>
4082 PR gcov-profile/84846
4083 * gcov.c (print_usage): Add new -t option.
4084 (process_args): Handle the option.
4085 (generate_results): Use stdout as output when requested by
4088 2018-05-18 Martin Liska <mliska@suse.cz>
4090 PR gcov-profile/84846
4091 * coverage.c (coverage_init): Write PWD to .gcno file.
4092 * doc/gcov.texi: Document how working directory is printed.
4093 * gcov-dump.c (dump_gcov_file): Print PWD.
4094 * gcov.c (output_intermediate_file): Likewise.
4095 (read_graph_file): Read PWD string.
4096 (output_lines): Print PWD.
4098 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4101 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4102 for retval and return false if all args to phi are zero.
4104 2018-05-18 Richard Biener <rguenther@suse.de>
4106 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4108 (evrp_dom_walker::before_dom_children): Call it.
4110 2018-05-18 Richard Biener <rguenther@suse.de>
4112 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4113 results when processing array refs with variable index.
4115 2018-05-18 Toon Moene <toon@moene.org>
4117 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4118 directly after that of -floop-interchange. Indicate that both
4119 options are enabled by default when specifying -O3.
4121 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4123 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4124 iterator. Delete separate integer-mode vec_set<mode> expander.
4125 (aarch64_simd_vec_setv2di): Delete.
4126 (vec_setv2di): Delete.
4127 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4128 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4129 the "w, r" alternative.
4131 2018-05-18 Martin Liska <mliska@suse.cz>
4133 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4134 * tree-pass.h (make_pass_lower_switch_O0): New function.
4135 * tree-switch-conversion.c (node_has_low_bound): Remove.
4136 (node_has_high_bound): Likewise.
4137 (node_is_bounded): Likewise.
4138 (class pass_lower_switch): Make it a template type and create
4140 (pass_lower_switch::execute): Add template argument.
4141 (make_pass_lower_switch): New function.
4142 (make_pass_lower_switch_O0): New function.
4143 (do_jump_if_equal): Remove.
4144 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4145 all the hard work to tree optimization passes.
4147 2018-05-18 Martin Liska <mliska@suse.cz>
4149 * dbgcnt.c (limit_low): Renamed from limit.
4150 (limit_high): New variable.
4151 (dbg_cnt_is_enabled): Check for upper limit.
4152 (dbg_cnt): Adjust dumping.
4153 (dbg_cnt_set_limit_by_index): Add new argument for high
4155 (dbg_cnt_set_limit_by_name): Likewise.
4156 (dbg_cnt_process_single_pair): Parse new format.
4157 (dbg_cnt_process_opt): Use strtok.
4158 (dbg_cnt_list_all_counters): Remove 'value' and add
4160 * doc/invoke.texi: Document changes.
4162 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4164 * doc/sourcebuild.texi (scalar_all_fma): Document.
4165 * tree.def (FMA_EXPR): Delete.
4166 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4167 * internal-fn.c (ternary_direct): New macro.
4168 (expand_ternary_optab_fn): Likewise.
4169 (direct_ternary_optab_supported_p): Likewise.
4170 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4171 * builtins.c (fold_builtin_fma): Delete.
4172 (fold_builtin_3): Don't call it.
4173 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4174 * expr.c (expand_expr_real_2): Likewise.
4175 * fold-const.c (operand_equal_p): Likewise.
4176 (fold_ternary_loc): Likewise.
4177 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4178 * gimple.c (DEFTREECODE): Likewise.
4179 * gimplify.c (gimplify_expr): Likewise.
4180 * optabs-tree.c (optab_for_tree_code): Likewise.
4181 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4182 * tree-eh.c (operation_could_trap_p): Likewise.
4183 (stmt_could_throw_1_p): Likewise.
4184 * tree-inline.c (estimate_operator_cost): Likewise.
4185 * tree-pretty-print.c (dump_generic_node): Likewise.
4186 (op_code_prio): Likewise.
4187 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4188 * tree-ssa-operands.c (get_expr_operands): Likewise.
4189 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4190 * fold-const-call.h (fold_fma): Delete.
4191 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4192 CFN_FNMA and CFN_FNMS.
4194 * genmatch.c (combined_fn): New enum.
4195 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4196 (commutative_op): New function.
4197 (commutate): Use it. Handle more than 2 operands.
4198 (dt_operand::gen_gimple_expr): Use commutative_op.
4199 (parser::parse_expr): Allow :c to be used with non-binary
4200 operators if the commutative operand is known.
4201 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4202 CFN_FMS, CFN_FNMA and CFN_FNMS.
4203 (backprop::process_assign_use): Remove FMA_EXPR handling.
4204 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4205 (gen_hsa_fma): New function.
4206 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4207 IFN_FNMA and IFN_FNMS.
4208 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4209 * gimple-fold.h (follow_all_ssa_edges): Declare.
4210 * gimple-fold.c (follow_all_ssa_edges): New function.
4211 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4212 gimple_build interface and use follow_all_ssa_edges to fold the result.
4213 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4214 instead of checking for optabs directly.
4215 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4216 rather than FMA_EXPRs.
4217 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4218 call to IFN_FMA instead of an FMA_EXPR.
4220 2018-05-17 Jim Wilson <jimw@sifive.com>
4222 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4223 sign extended promoted subreg, and the range does not have the sign bit
4224 set, then do a sign extend.
4226 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4227 test, check for sign extended subreg and/or constant operands, and
4228 do a sign extend in that case.
4230 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4232 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4233 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4235 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4236 Change logics_shift_reg to logics_shift_imm.
4237 (thunderx2t99_fp_loadpair_basic): Delete.
4238 (thunderx2t99_fp_storepair_basic): Delete.
4239 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4240 (thunderx2t99_asimd_polynomial): Delete.
4241 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4242 and neon_fp_mul_d_scalar_q.
4243 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4244 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4245 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4246 (thunderx2t99_asimd_lut): Add missing tbl types.
4247 (thunderx2t99_asimd_ext): Delete.
4248 (thunderx2t99_asimd_load1_1_mult): Delete.
4249 (thunderx2t99_asimd_load1_2_mult): Delete.
4250 (thunderx2t99_asimd_load1_ldp): New.
4251 (thunderx2t99_asimd_load1): New.
4252 (thunderx2t99_asimd_load2): Add missing *load2* types.
4253 (thunderx2t99_asimd_load3): New.
4254 (thunderx2t99_asimd_load4): New.
4255 (thunderx2t99_asimd_store1_1_mult): Delete.
4256 (thunderx2t99_asimd_store1_2_mult): Delete.
4257 (thunderx2t99_asimd_store2_mult): Delete.
4258 (thunderx2t99_asimd_store2_onelane): Delete.
4259 (thunderx2t99_asimd_store_stp): New.
4260 (thunderx2t99_asimd_store1): New.
4261 (thunderx2t99_asimd_store2): New.
4262 (thunderx2t99_asimd_store3): New.
4263 (thunderx2t99_asimd_store4): New.
4265 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4267 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4268 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4270 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4271 Segher Boessenkool <segher@kernel.crashing.org>
4274 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4277 2018-05-17 Richard Biener <rguenther@suse.de>
4279 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4280 for pruning loop and prune defs feeding only already visited PHIs.
4282 2018-05-17 Richard Biener <rguenther@suse.de>
4284 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4286 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4287 Richard Biener <rguenther@suse.de>
4289 PR tree-optimization/85793
4290 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4291 for VMAT_ELEMENTWISE.
4293 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4295 * internal-fn.h (lookup_internal_fn): Declare
4296 * internal-fn.c (lookup_internal_fn): New function.
4297 * gimple.c (gimple_build_call_from_tree): Handle calls to
4299 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4300 internal function names.
4301 * tree-pretty-print.c (dump_generic_node): Likewise.
4302 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4304 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4306 * gimple-fold.h (gimple_build): Make the function forms take
4307 combined_fn rather than built_in_function.
4308 (gimple_simplify): Likewise.
4309 * gimple-match-head.c (gimple_simplify): Likewise.
4310 * gimple-fold.c (gimple_build): Likewise.
4311 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4312 rather than gimple_build_call_internal.
4313 (get_initial_defs_for_reduction): Likewise.
4314 (vect_create_epilog_for_reduction): Likewise.
4315 (vectorizable_live_operation): Likewise.
4317 2018-05-17 Martin Liska <mliska@suse.cz>
4319 * gimple-ssa-sprintf.c (format_directive): Do not use
4320 space in between 'G_' and '('.
4322 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4325 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4326 even if the mask is not all ones.
4329 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4331 (ix86_gimple_fold_builtin): Likewise.
4334 * config/i386/i386.c: Include tree-vector-builder.h.
4335 (ix86_vector_shift_count): New function.
4336 (ix86_fold_builtin): Fold shift builtins by scalar count.
4337 (ix86_gimple_fold_builtin): Likewise.
4339 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4340 _mm512_setzero): New intrinsics.
4342 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4343 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4345 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4346 code generation for cases where splatting a value is not useful.
4347 * simplify-rtx.c (simplify_ternary_operation): Simplify
4348 vec_merge across a vec_duplicate and a paradoxical subreg forming
4349 a vector mode to a vec_concat.
4351 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4353 * config.gcc: Support "goldmont-plus".
4354 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4356 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4357 PROCESSOR_GOLDMONT_PLUS.
4358 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4359 (processor_target_table): Add "goldmont-plus".
4360 (PTA_GOLDMONT_PLUS): Define.
4361 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4362 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4363 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4364 (fold_builtin_cpu): Add "goldmont-plus".
4365 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4366 (ix86_option_override_internal): Add "goldmont-plus".
4367 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4368 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4369 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4370 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4372 2018-05-17 Richard Biener <rguenther@suse.de>
4374 PR tree-optimization/85757
4375 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4376 remove defs that only feed that PHI from further processing.
4378 2018-05-16 Jim Wilson <jimw@sifive.com>
4380 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4382 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4384 2018-05-16 Mark Wielaard <mark@klomp.org>
4386 * dwarf2out.c (count_index_strings): New function.
4387 (output_indirect_strings): Call count_index_strings and generate
4388 header for dwarf_version >= 5.
4390 2018-05-16 Mark Wielaard <mark@klomp.org>
4392 * dwarf2out.c (dwarf_FORM): New function.
4393 (set_indirect_string): Use dwarf_FORM.
4394 (reset_indirect_string): Likewise.
4395 (size_of_die): Likewise.
4396 (value_format): Likewise.
4397 (output_die): Likewise.
4398 (add_skeleton_AT_string): Likewise.
4399 (output_macinfo_op): Likewise.
4400 (index_string): Likewise.
4401 (output_index_string_offset): Likewise.
4402 (output_index_string): Likewise.
4403 (count_index_strings): Likewise.
4405 2018-05-16 Carl Love <cel@us.ibm.com>
4407 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4408 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4410 2018-05-16 Martin Jambor <mjambor@suse.cz>
4412 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4413 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4415 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4417 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4418 (fnma<mode>4): Likewise.
4419 (fms<mode>4): Likewise.
4420 (fnms<mode>4): Likewise.
4421 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4422 (aarch64_fnma<mode>4): Likewise.
4423 (aarch64_fms<mode>4): Likewise.
4424 (aarch64_fnms<mode>4): Likewise.
4425 (aarch64_fnmadd<mode>4): Likewise.
4427 2018-05-16 Jason Merrill <jason@redhat.com>
4429 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4431 2018-05-16 Richard Biener <rguenther@suse.de>
4433 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4434 (dump_stmt_cost): Declare.
4435 (add_stmt_cost): Dump cost we add.
4436 (add_stmt_costs): New function.
4437 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4439 (vect_analyze_stmt): Adjust prototype.
4440 (vectorizable_condition): Likewise.
4441 (vectorizable_live_operation): Likewise.
4442 (vectorizable_reduction): Likewise.
4443 (vectorizable_induction): Likewise.
4444 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4445 cost vector to pass to vectorizable_ and record afterwards.
4446 (vect_model_reduction_cost): Take cost vector argument and adjust.
4447 (vect_model_induction_cost): Likewise.
4448 (vectorizable_reduction): Likewise.
4449 (vectorizable_induction): Likewise.
4450 (vectorizable_live_operation): Likewise.
4451 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4452 SLP_TREE_NUMBER_OF_VEC_STMTS.
4453 (vect_analyze_slp_cost_1): Remove.
4454 (vect_analyze_slp_cost): Likewise.
4455 (vect_slp_analyze_node_operations): Take visited args and
4456 a target cost vector. Avoid processing already visited stmt sets.
4457 (vect_slp_analyze_operations): Use a local cost vector to gather
4458 costs and register those of non-discarded instances.
4459 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4460 (vect_schedule_slp_instance): Remove copying of
4461 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4463 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4464 adding cost. Record cost entry location.
4465 (vect_prologue_cost_for_slp_op): Function to compute cost of
4466 a constant or invariant generated for SLP vect in the prologue,
4467 split out from vect_analyze_slp_cost_1.
4468 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4469 (vect_model_promotion_demotion_cost): Likewise.
4470 (vect_model_store_cost): Likewise, make static.
4471 (vect_model_load_cost): Likewise.
4472 (vectorizable_bswap): Add cost vector arg and adjust.
4473 (vectorizable_call): Likewise.
4474 (vectorizable_simd_clone_call): Likewise.
4475 (vectorizable_conversion): Likewise.
4476 (vectorizable_assignment): Likewise.
4477 (vectorizable_shift): Likewise.
4478 (vectorizable_operation): Likewise.
4479 (vectorizable_store): Likewise.
4480 (vectorizable_load): Likewise.
4481 (vectorizable_condition): Likewise.
4482 (vectorizable_comparison): Likewise.
4483 (can_vectorize_live_stmts): Likewise.
4484 (vect_analyze_stmt): Likewise.
4485 (vect_transform_stmt): Adjust calls to vectorizable_*.
4486 * tree-vectorizer.c: Include gimple-pretty-print.h.
4487 (dump_stmt_cost): New function.
4489 2018-05-16 Richard Biener <rguenther@suse.de>
4491 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4492 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4493 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4494 (check_name): New callback.
4495 (dse_classify_store): Track cycles via a visited bitmap of PHI
4496 defs and simplify handling of in-loop and across loop dead stores
4497 and properly fail for loop-variant refs. Handle byte-tracking with
4498 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4501 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4503 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4504 (vect_get_mask_type_for_stmt): Likewise.
4505 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4507 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4508 to determine the statement's vector type and the vector type that
4509 should be used for calculating nunits. Deal with cases in which
4510 the type has to be deferred.
4511 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4512 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4513 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4514 (vect_determine_vf_for_stmt): New functions, split out from...
4515 (vect_determine_vectorization_factor): ...here.
4516 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4517 (vect_get_mask_type_for_stmt): New functions, split out from
4518 vect_determine_vectorization_factor.
4520 2018-05-16 Richard Biener <rguenther@suse.de>
4522 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4523 verify the [VEC_]COND_EXPR embedded comparison.
4525 2018-05-15 Martin Sebor <msebor@redhat.com>
4527 PR tree-optimization/85753
4528 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4529 RECORD_TYPE in addition to ARRAY_TYPE.
4531 2018-05-15 Martin Sebor <msebor@redhat.com>
4534 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4536 2018-05-15 Richard Biener <rguenther@suse.de>
4538 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4539 add by_clobber_p one. Change algorithm to collect all defs
4540 representing uses we need to walk and try reducing them to
4541 a single one before failing.
4542 (dse_dom_walker::dse_optimize_stmt): Adjust.
4544 2018-05-13 Mark Wielaard <mark@klomp.org>
4546 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4547 (size_of_loc_descr): Likewise.
4548 (output_loc_operands): Likewise.
4549 (output_loc_operands_raw): Likewise.
4550 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4551 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4552 (hash_loc_operands): Likewise.
4553 (compare_loc_operands): Likewise.
4555 2018-05-14 Mark Wielaard <mark@klomp.org>
4557 * dwarf2out.c (count_index_addrs): New function.
4558 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4560 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4562 PR tree-optimization/83648
4563 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4564 return value as malloc candidate.
4566 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4569 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4570 param as true in call to suggest_attribute.
4572 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4574 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4575 -mreadonly-in-sdata.
4577 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4579 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4581 (aarch64_crypto_aesd_fused): Likewise.
4583 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4585 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4586 (movsi_aarch64): Likewise.
4587 (load_pairsi): Likewise.
4588 (load_pairdi): Likewise.
4589 (store_pairsi): Likewise.
4590 (store_pairdi): Likewise.
4591 (load_pairsf): Likewise.
4592 (load_pairdf): Likewise.
4593 (store_pairsf): Likewise.
4594 (store_pairdf): Likewise.
4595 (zero_extend): Likewise.
4596 (trunc): Swap alternatives.
4597 (fcvt_target): Add '?' to prefer w over r.
4599 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4602 * config/i386/i386.md: Disallow non-commutative arithmetics in
4603 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4604 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4605 in the peephole2 before it.
4607 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4609 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4610 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4611 (ix86_handle_option): Handle -mcldemote.
4612 * config.gcc: New header.
4613 * config/i386/cldemoteintrin.h: New file.
4614 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4615 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4617 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4618 OPTION_MASK_ISA_CLDEMOTE.
4619 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4620 (ix86_valid_target_attribute_inner_p): Ditto.
4621 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4622 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4623 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4624 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4625 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4627 * config/i386/i386.opt: Add -mcldemote.
4628 * config/i386/x86intrin.h: New header.
4629 * doc/invoke.texi: Add -mcldemote.
4631 2018-05-14 Richard Biener <rguenther@suse.de>
4633 * doc/match-and-simplify.texi: Adjust :s documentation.
4635 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4637 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4638 intended memcpy size.
4639 (REORDER_45): Likewise.
4641 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4643 * sort.cc: New file.
4644 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4645 * vec.c (qsort_chk): Use gcc_qsort.
4646 * Makefile.in (OBJS-libcommon): Add sort.o.
4647 (build/sort.o): New target. Use it...
4648 (BUILD_RTL): ... here, and...
4649 (build/gencfn-macros): ... here, and...
4650 (build/genmatch): ... here.
4652 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4653 Chung-Ju Wu <jasonwucj@gmail.com>
4655 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4656 * config/nds32/nds32-graywolf.md: New file.
4657 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4658 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4660 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4661 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4662 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4663 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4664 * config/nds32/pipelines.md: Include n15 settings.
4666 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4667 Chung-Ju Wu <jasonwucj@gmail.com>
4669 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4670 * config/nds32/nds32-n13.md: New file.
4671 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4672 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4674 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4675 * config/nds32/nds32.md (pipeline_model): Add n13.
4676 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4677 * config/nds32/pipelines.md: Include n13 settings.
4679 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4680 Chung-Ju Wu <jasonwucj@gmail.com>
4682 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4683 * config/nds32/nds32-n10.md: New file.
4684 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4685 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4687 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4688 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4689 * config/nds32/nds32.md (pipeline_model): Add n10.
4690 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4691 * config/nds32/pipelines.md: Include n10 settings.
4693 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4694 Kito Cheng <kito.cheng@gmail.com>
4695 Chung-Ju Wu <jasonwucj@gmail.com>
4697 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4698 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4699 Add enum values for DSP extension instructions.
4700 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4702 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4703 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4705 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4706 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4707 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4708 * config/nds32/nds32-intrinsic.md: Likewise.
4709 * config/nds32/nds32_intrinsic.h: Likewise.
4710 * config/nds32/nds32-md-auxiliary.c: Likewise.
4711 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4712 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4713 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4714 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4715 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4716 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4717 TYPE_DMAC in switch statement.
4718 * config/nds32/nds32.c: New checking and implementation for DSP
4719 extension instructions.
4720 * config/nds32/nds32.h: Likewise.
4721 * config/nds32/nds32.md: Likewise.
4722 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4723 * config/nds32/predicates.md: Implement new predicates for DSP
4726 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4728 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4729 Reformat alternatives and attributes so it is easier to identify
4730 which constraints/attributes go with which instruction.
4731 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4732 (mov<mode>_softfloat32, FMOVE64): Likewise.
4733 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4734 (mov<mode>_softfloat64, FMOVE64): Likewise.
4736 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4738 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4740 (Basic PowerPC Built-in Functions): The new name of the
4741 subsection previously known as "PowerPC Built-in Functions".
4742 (Basic PowerPC Built-in Functions Available on all Configurations):
4744 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4745 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4746 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4747 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4749 2018-05-11 Martin Jambor <mjambor@suse.cz>
4752 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4755 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4758 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4760 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4762 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4763 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4764 (ix86_handle_option): Handle -mwaitpkg.
4765 * config.gcc: New header.
4766 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4767 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4768 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4770 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4771 OPTION_MASK_ISA_WAITPKG.
4772 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4773 (ix86_option_override_internal): Add PTA_WAITPKG.
4774 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4775 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4776 IX86_BUILTIN_TPAUSE.
4777 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4778 __builtin_ia32_umwait and __builtin_ia32_tpause.
4779 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4780 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4781 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4782 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4783 UNSPECV_TPAUSE): New.
4784 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4785 * config/i386/i386.opt: Add -mwaitpkg.
4786 * config/i386/waitpkgintrin.h: New file.
4787 * config/i386/x86intrin.h: New header.
4788 * doc/invoke.texi: Add -mwaitpkg.
4790 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4793 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4795 (cortex-m0): Use armv6s-m isa.
4796 (cortex-m0plus): Likewise.
4797 (cortex-m1): Likewise.
4798 (cortex-m0.small-multiply): Likewise.
4799 (cortex-m0plus.small-multiply): Likewise.
4800 (cortex-m1.small-multiply): Likewise.
4802 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4803 Jakub Jelinek <jakub@redhat.com>
4805 PR tree-optimization/85692
4806 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4807 source permute as well.
4809 2018-05-11 Martin Liska <mliska@suse.cz>
4812 * doc/extend.texi: Document LLVM style format for no_sanitize
4815 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4817 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4818 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4819 (mode_supports_vsx_dform_quad): Likewise.
4820 (mode_supports_vmx_dform): Move these functions to be next to the
4821 other mode_supports functions.
4822 (mode_supports_dq_form): Likewise.
4823 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4824 mode_supports_dq_form.
4825 (reg_offset_addressing_ok_p): Likewise.
4826 (offsettable_ok_by_alignment): Likewise.
4827 (rs6000_legitimate_offset_address_p): Likewise.
4828 (legitimate_lo_sum_address_p): Likewise.
4829 (rs6000_legitimize_address): Likewise.
4830 (rs6000_legitimize_reload_address): Likewise.
4831 (rs6000_secondary_reload_inner): Likewise.
4832 (rs6000_preferred_reload_class): Likewise.
4833 (rs6000_output_move_128bit): Likewise.
4835 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4837 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4838 Generate SImode target register for null target.
4839 <case IX86_BUILTIN_XGETBV>: Ditto.
4840 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4841 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4843 2018-05-10 Carl Love <cel@us.ibm.com>
4845 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4846 dcbtt and dcbtstt if operands[2] is 0.
4848 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4851 * config/i386/sse.md (usadv64qi): New expander.
4853 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4855 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4856 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4857 -maltivec=be support.
4858 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4859 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4860 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4861 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4862 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4863 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4864 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4865 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4866 altivec_vsumsws): Adjust.
4867 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4868 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4869 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4871 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4872 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4873 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4874 (altivec_lve<VI_char>x): Delete expand.
4875 (*altivec_lve<VI_char>x_internal): Rename to...
4876 (altivec_lve<VI_char>x): ... this.
4877 (altivec_lvxl_<mode>): Delete expand.
4878 (*altivec_lvxl_<mode>_internal): Rename to ...
4879 (altivec_lvxl_<mode>): ... this.
4880 (altivec_stvxl_<mode>): Delete expand.
4881 (*altivec_stvxl_<mode>_internal): Rename to ...
4882 (altivec_stvxl_<mode>): ... this.
4883 (altivec_stve<VI_char>x): Delete expand.
4884 (*altivec_stve<VI_char>x_internal): Rename to ...
4885 (altivec_stve<VI_char>x): ... this.
4886 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4887 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4888 reduc_plus_scal_<mode>): Adjust.
4889 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4891 (rs6000_cpu_cpp_builtins): Adjust.
4892 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4893 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4894 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4895 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4896 -maltivec=be support.
4897 (rs6000_split_vec_extract_var): Adjust.
4898 (rs6000_split_v4si_init): Adjust.
4899 (swap_selector_for_mode): Delete.
4900 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4901 altivec_expand_stvex_be): Delete.
4902 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4903 -maltivec=be support.
4904 (rs6000_gimple_fold_builtin): Ditto.
4905 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4907 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4908 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4909 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4910 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4911 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4912 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4913 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4914 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4915 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4916 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4917 anonymous split): Adjust.
4918 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4919 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4921 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4923 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4924 when --with-gxx-include-dir is also specified.
4925 * configure: Regenerate.
4927 2018-05-09 Jim Wilson <jimw@sifive.com>
4930 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4931 * config/riscv/t-withmultilib: New.
4932 * config/riscv/withmultilib.h: New.
4933 * doc/install.texi: Document RISC-V --with-multilib-list support.
4935 2018-05-09 Richard Biener <rguenther@suse.de>
4937 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4939 (vect_bb_vectorization_profitable_p): Adjust. Compute
4940 actual scalar cost using the cost vector and the add_stmt_cost
4943 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4945 PR rtl-optimization/85645
4946 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4947 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4949 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4951 PR rtl-optimization/85645
4952 * shrink-wrap.c (spread_components): Return a boolean saying if
4953 anything was changed.
4954 (try_shrink_wrapping_separate): Iterate spread_components until
4955 nothing changes anymore.
4957 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4959 PR rtl-optimization/85645
4960 * regrename.c (build_def_use): Also kill the chains that include the
4961 destination of a REG_CFA_REGISTER note.
4963 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4965 PR rtl-optimization/85645
4966 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4967 insn that has a REG_CFA_REGISTER note.
4969 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4971 * cfgexpand.c (expand_clobber): New function.
4972 (expand_gimple_stmt_1): Use it.
4973 * tree-vect-stmts.c (vect_clobber_variable): New function,
4975 (vectorizable_simd_clone_call): ...here.
4976 (vectorizable_store): Emit a clobber either side of an
4977 IFN_STORE_LANES sequence.
4978 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4980 2018-05-09 Tom de Vries <tom@codesourcery.com>
4983 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4984 (define_insn "trap_if_false"): Add exit after trap.
4986 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4988 PR rtl-optimization/85638
4989 * bb-reorder.c: Include common/common-target.h.
4990 (create_forwarder_block): New function extracted from...
4991 (fix_up_crossing_landing_pad): ...here. Rename into...
4992 (dw2_fix_up_crossing_landing_pad): ...this.
4993 (sjlj_fix_up_crossing_landing_pad): New function.
4994 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4995 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4996 from both partitions and exit the loop after one iteration.
4998 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5001 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5003 (Basic PowerPC Built-in Functions): The new name of the
5004 subsection previously known as "PowerPC Built-in Functions".
5005 (Basic PowerPC Built-in Functions Available on all Configurations):
5007 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5009 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5011 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5013 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5016 2018-05-08 Jim Wilson <jimw@sifive.com>
5018 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5019 (LD_EMUL_SUFFIX): New.
5020 (LINK_SPEC): Use it.
5022 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5024 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5026 (Basic PowerPC Built-in Functions): The new name of the
5027 subsection previously known as "PowerPC Built-in Functions".
5028 (Basic PowerPC Built-in Functions Available on all Configurations):
5030 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5032 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5034 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5036 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5039 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5042 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5043 after cmpelim optimization.
5045 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5047 * config.gcc: Support "goldmont".
5048 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5049 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5051 * config/i386/i386.c (m_GOLDMONT): Define.
5052 (processor_target_table): Add "goldmont".
5053 (PTA_GOLDMONT): Define.
5054 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5055 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5056 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5057 (fold_builtin_cpu): Add "goldmont".
5058 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5059 (ix86_option_override_internal): Add "goldmont".
5060 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5061 (processor_type): Add PROCESSOR_GOLDMONT.
5062 * config/i386/i386.md: Add CPU "glm".
5063 * config/i386/glm.md: New file.
5064 * config/i386/x86-tune.def: Add m_GOLDMONT.
5065 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5067 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5070 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5072 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5073 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5074 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5077 * config/i386/i386.c (ix86_fold_builtin): Handle
5078 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5081 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5082 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5084 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5087 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5088 (check_arch): Likewise.
5089 (check_fpu): Return the result rather than printing it.
5090 (end arch): Fix operator precedence.
5091 (end cpu): Likewise.
5092 (END): Print the result from check_fpu.
5094 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5095 Alan Hayward <alan.hayward@arm.com>
5096 David Sherwood <david.sherwood@arm.com>
5098 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5099 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5100 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5101 (*fcmuo<mode>_and): New patterns.
5103 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5105 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5106 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5107 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5108 (cmp_op, sve_imm_con): New code attributes.
5109 (SVE_COND_INT_CMP, imm_con): Delete.
5110 (cmp_op): Remove above unspecs from int attribute.
5111 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5113 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5114 comparison-specific unspecs.
5115 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5116 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5117 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5118 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5119 (*vec_fcm<cmp_op><mode>): Rename to...
5120 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5121 (*vec_fcmuo<mode>): Rename to...
5122 (*fcmuo<mode>): ...this and adjust likewise.
5123 (*pred_fcm<cmp_op><mode>): New pattern.
5124 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5125 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5127 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5129 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5130 (aarch64_emit_sve_predicated_cond): New function.
5131 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5132 (aarch64_emit_unspec_cond_or): Replace with...
5133 (aarch64_emit_sve_or_conds): ...this new function. Use
5134 aarch64_emit_sve_ptrue_op for the individual comparisons and
5135 aarch64_emit_binop to OR them together.
5136 (aarch64_emit_inverted_unspec_cond): Replace with...
5137 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5138 aarch64_emit_sve_ptrue_op for the comparison and
5139 aarch64_emit_unop to invert the result.
5140 (aarch64_expand_sve_vec_cmp_float): Update after the above
5141 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5143 2018-05-07 Nathan Sidwell <nathan@acm.org>
5145 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5146 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5147 (Backwards Compatibility): Likewise.
5149 2018-05-07 Luis Machado <luis.machado@linaro.org>
5153 2018-05-07 Luis Machado <luis.machado@linaro.org>
5155 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5156 <prefetch_dynamic_strides>: New const bool field.
5157 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5158 prefetch_dynamic_strides.
5159 (exynosm1_prefetch_tune): Likewise.
5160 (thunderxt88_prefetch_tune): Likewise.
5161 (thunderx_prefetch_tune): Likewise.
5162 (thunderx2t99_prefetch_tune): Likewise.
5163 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5165 (aarch64_override_options_internal): Update to set
5166 PARAM_PREFETCH_DYNAMIC_STRIDES.
5167 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5168 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5169 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5170 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5171 prefetch-dynamic-strides setting.
5173 2018-05-07 Luis Machado <luis.machado@linaro.org>
5175 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5176 <minimum_stride>: New const int field.
5177 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5178 minimum_stride field.
5179 (exynosm1_prefetch_tune): Likewise.
5180 (thunderxt88_prefetch_tune): Likewise.
5181 (thunderx_prefetch_tune): Likewise.
5182 (thunderx2t99_prefetch_tune): Likewise.
5183 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5184 (aarch64_override_options_internal): Update to set
5185 PARAM_PREFETCH_MINIMUM_STRIDE.
5186 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5187 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5188 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5189 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5190 stride is constant and is below the minimum stride threshold.
5192 2018-05-07 Luis Machado <luis.machado@linaro.org>
5194 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5197 2018-05-07 Luis Machado <luis.machado@linaro.org>
5199 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5200 <prefetch_dynamic_strides>: New const bool field.
5201 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5202 prefetch_dynamic_strides.
5203 (exynosm1_prefetch_tune): Likewise.
5204 (thunderxt88_prefetch_tune): Likewise.
5205 (thunderx_prefetch_tune): Likewise.
5206 (thunderx2t99_prefetch_tune): Likewise.
5207 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5209 (aarch64_override_options_internal): Update to set
5210 PARAM_PREFETCH_DYNAMIC_STRIDES.
5211 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5212 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5213 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5214 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5215 prefetch-dynamic-strides setting.
5217 2018-05-07 Luis Machado <luis.machado@linaro.org>
5219 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5220 <minimum_stride>: New const int field.
5221 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5222 minimum_stride field.
5223 (exynosm1_prefetch_tune): Likewise.
5224 (thunderxt88_prefetch_tune): Likewise.
5225 (thunderx_prefetch_tune): Likewise.
5226 (thunderx2t99_prefetch_tune): Likewise.
5227 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5228 (aarch64_override_options_internal): Update to set
5229 PARAM_PREFETCH_MINIMUM_STRIDE.
5230 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5231 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5232 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5233 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5234 stride is constant and is below the minimum stride threshold.
5236 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5239 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5240 the type is addressable. Don't force op into register if it has
5243 2018-05-05 Roland McGrath <mcgrathr@google.com>
5246 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5247 any section for which we don't know a specific type it should have,
5248 regardless of name. Previously this was done only for the exact
5249 names ".init_array", ".fini_array", and ".preinit_array".
5250 (default_elf_asm_named_section): Add comment about
5251 relationship with default_section_type_flags and SECTION_NOTYPE.
5252 (get_section): Don't consider it a type conflict if one side has
5253 SECTION_NOTYPE and the other doesn't, as long as neither has the
5254 SECTION_BSS et al used in the default_section_type_flags logic.
5256 2018-05-05 Tom de Vries <tom@codesourcery.com>
5259 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5260 (workaround_barsyncs): New function.
5261 (nvptx_reorg): Use workaround_barsyncs.
5262 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5263 (define_expand "nvptx_membar_cta"): New define_expand.
5264 (define_insn "*nvptx_membar_cta"): New insn.
5266 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5268 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5269 To improve optimization opportunities.
5270 * builtin-types.def: The new needed builtin types for the above.
5272 2018-05-04 Richard Biener <rguenther@suse.de>
5274 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5275 * gimple-ssa-store-merging.c
5276 (imm_store_chain_info::output_merged_store): Remove redundant create,
5277 release split_store vector contents on failure.
5278 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5279 scalar stmt vector on cache hit.
5281 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5283 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5285 * config.gcc (powerpc-xilinx-eabi*): Remove.
5286 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5288 (fusion_addis_mem_combo_load): Ditto.
5289 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5291 (rs6000_cpu_cpp_builtins): Ditto.
5292 * config/rs6000/rs6000-linux.c
5293 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5294 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5295 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5297 (rs6000_setup_reg_addr_masks): Ditto.
5298 (rs6000_init_hard_regno_mode_ok): Ditto.
5299 (rs6000_option_override_internal): Ditto.
5300 (legitimate_lo_sum_address_p): Ditto.
5301 (rs6000_legitimize_address): Ditto.
5302 (rs6000_legitimize_reload_address): Ditto.
5303 (rs6000_legitimate_address_p): Ditto.
5304 (abi_v4_pass_in_fpr): Ditto.
5305 (setup_incoming_varargs): Ditto.
5306 (rs6000_gimplify_va_arg): Ditto.
5307 (rs6000_split_multireg_move): Ditto.
5308 (rs6000_savres_strategy): Ditto.
5309 (rs6000_emit_prologue_components): Ditto.
5310 (rs6000_emit_epilogue_components): Ditto.
5311 (rs6000_emit_prologue): Ditto.
5312 (rs6000_emit_epilogue): Ditto.
5313 (rs6000_elf_file_end): Ditto.
5314 (rs6000_function_value): Ditto.
5315 (rs6000_libcall_value): Ditto.
5316 * config/rs6000/rs6000.h: Ditto.
5317 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5318 (TARGET_MINMAX): ... this. New.
5319 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5320 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5321 (*movsi_internal1_single): Delete.
5322 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5323 mfpu=, mxilinx-fpu): Delete.
5324 * config/rs6000/singlefp.h: Delete.
5325 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5326 * config/rs6000/t-rs6000: Ditto.
5327 * config/rs6000/t-xilinx: Delete.
5328 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5329 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5330 (VStype_simple): Delete.
5331 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5332 * config/rs6000/xfpu.h: Delete.
5333 * config/rs6000/xfpu.md: Delete.
5334 * config/rs6000/xilinx.h: Delete.
5335 * config/rs6000/xilinx.opt: Delete.
5336 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5337 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5339 2018-05-04 Tom de Vries <tom@codesourcery.com>
5342 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5345 2018-05-04 Richard Biener <rguenther@suse.de>
5348 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5349 (expand_complex_div_wide): Likewise.
5350 (expand_complex_operations_1): Likewise.
5351 (expand_complex_libcall): Preserve EH info of the original stmt.
5352 (tree_lower_complex): Handle removed blocks.
5353 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5354 on complex multiplication and division libcall builtins.
5356 2018-05-04 Richard Biener <rguenther@suse.de>
5359 * fold-const.c (negate_expr_p): Restrict negation of operand
5360 zero of a division to when we know that can happen without
5362 (fold_negate_expr_1): Likewise.
5364 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5367 * real.h (real_nextafter): Declare.
5368 * real.c (real_nextafter): New function.
5369 * fold-const-call.c (fold_const_nextafter): New function.
5370 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5371 CASE_CFN_NEXTTOWARD.
5372 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5373 even when arg1_mode is different from arg0_mode.
5375 2018-05-03 Nathan Sidwell <nathan@acm.org>
5377 * doc/extend.texi (Deprecated Features): Remove
5379 (Backwards Compatibility): Likewise.
5380 * doc/invoke.texi (C++ Language Options): Likewise.
5381 (C++ Dialect Options): Likewise.
5383 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5386 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5387 _mm512_mask_mullox_epi64): New intrinsics.
5389 2018-05-03 Tom de Vries <tom@codesourcery.com>
5392 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5393 dump files): Add offload-tree.
5395 2018-05-03 Richard Biener <rguenther@suse.de>
5397 PR tree-optimization/85615
5398 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5399 to loops not nested in BBs loop father to avoid creating multi-entry
5402 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5404 PR tree-optimization/70291
5405 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5406 arguments. Change return type to tree. Emit libcall as a new
5407 statement rather than replacing existing one when inplace_p is true.
5408 (expand_complex_multiplication_components): New function.
5409 (expand_complex_multiplication): Expand floating-point complex
5410 multiplication using the above.
5411 (expand_complex_division): Rename inner_type parameter to type.
5412 Update expand_complex_libcall call-site.
5413 (expand_complex_operations_1): Update expand_complex_multiplication
5414 and expand_complex_division call-sites.
5416 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5419 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5420 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5421 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5422 the highest significant bit of the shift count mask is clear. In
5423 check whether and[sq]i3 is needed verify that all significant bits
5424 of the shift count other than the highest are set.
5426 2018-05-02 Tom de Vries <tom@codesourcery.com>
5429 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5430 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5431 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5432 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5433 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5434 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5435 __builtin_goacc_parlevel_size.
5437 2018-05-02 Richard Biener <rguenther@suse.de>
5439 PR tree-optimization/85597
5440 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5441 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5444 2018-05-02 Tom de Vries <tom@codesourcery.com>
5447 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5448 dump files): Add ltrans-tree.
5450 2018-05-02 Tom de Vries <tom@codesourcery.com>
5453 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5454 dump files): Add wpa-ipa.
5456 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5458 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5459 powerpc*-*-linux*paired* target.
5460 * config/rs6000/750cl.h: Delete.
5461 * config/rs6000/paired.h: Delete.
5462 * config/rs6000/paired.md: Delete.
5463 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5465 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5466 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5467 comment. Remove paired float support.
5468 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5469 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5471 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5472 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5474 * config/rs6000/rs6000.c: Remove paired float support.
5475 (paired_expand_vector_init, paired_expand_vector_move,
5476 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5477 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5478 paired_expand_builtin, paired_expand_predicate_builtin,
5479 paired_init_builtins): Delete.
5480 * config/rs6000/rs6000.h: Remove paired float support.
5481 * config/rs6000/rs6000.md: Remove paired float support.
5482 (move_from_CR_ov_bit): Delete.
5483 * config/rs6000/rs6000.opt (mpaired): Delete.
5484 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5485 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5487 2018-05-02 Richard Biener <rguenther@suse.de>
5490 * gimplify.c (gimplify_save_expr): When in SSA form allow
5491 SAVE_EXPRs to compute to SSA vars.
5493 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5496 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5497 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5498 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5499 clobber operands[2], instead use a new pseudo. Formatting fixes.
5501 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5503 PR tree-optimization/85586
5504 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5505 exit early for statements in the same group if the accesses are
5508 2018-05-02 Tom de Vries <tom@codesourcery.com>
5511 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5514 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5516 PR tree-optimization/85143
5517 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5519 2018-05-01 Tom de Vries <tom@codesourcery.com>
5522 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5523 not found" error message.
5525 2018-05-01 Tom de Vries <tom@codesourcery.com>
5528 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5529 * vec.c (test_ordered_remove_if): New function.
5530 (vec_c_tests): Call test_ordered_remove_if.
5531 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5532 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5533 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5534 VEC_ORDERED_REMOVE_IF.
5536 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5538 PR tree-optimization/82665
5539 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5540 pointer subtraction where arguments come from a memchr call.
5542 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5544 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5545 --push-state --as-needed and --pop-state instead of --as-needed and
5546 --no-as-needed if ld supports it.
5547 * configure: Regenerated.
5550 * doc/install.texi2html: Replace _002d with - and _002a with * in
5551 generated html files using sed.
5553 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5556 * gcc-rich-location.c (blank_line_before_p): New function.
5557 (use_new_line): New function.
5558 (gcc_rich_location::add_fixit_insert_formatted): New function.
5559 * gcc-rich-location.h
5560 (gcc_rich_location::add_fixit_insert_formatted): New function.
5562 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5564 * selftest.c (assert_streq): Rename "expected" and "actual" to
5565 "val1" and "val2". Extend NULL-handling to cover both inputs
5566 symmetrically, while still requiring both to be non-NULL for a pass.
5567 * selftest.h (assert_streq): Rename "expected" and "actual" to
5569 (ASSERT_EQ): Likewise.
5570 (ASSERT_EQ_AT): Likewise.
5571 (ASSERT_KNOWN_EQ): Likewise.
5572 (ASSERT_KNOWN_EQ_AT): Likewise.
5573 (ASSERT_NE): Likewise.
5574 (ASSERT_MAYBE_NE): Likewise.
5575 (ASSERT_MAYBE_NE_AT): Likewise.
5576 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5577 the assertion to pass.
5578 (ASSERT_STREQ_AT): Likewise.
5580 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5582 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5583 interaction with -pie.
5585 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5587 * selftest.h: Fix alphabetization of per-source-file selftest
5590 2018-04-30 Jason Merrill <jason@redhat.com>
5592 PR c++/61982 - dead stores to destroyed objects.
5593 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5596 2018-04-30 Jason Merrill <jason@redhat.com>
5598 * tree.c (build_clobber): New.
5599 * tree.h: Declare it.
5600 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5602 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5604 * diagnostic-show-locus.c (layout::layout): Update for
5605 location_get_source_line returning a char_span.
5606 (struct char_span): Move to input.h.
5607 (struct correction): Update for fields in char_span becoming
5609 (struct source_line): Update for location_get_source_line
5610 returning a char_span.
5611 (layout::print_line): Likewise.
5612 * edit-context.c (edited_file::print_content): Likewise.
5613 (edited_file::print_diff_hunk): Likewise.
5614 (edited_file::print_run_of_changed_lines): Likewise.
5615 (edited_file::get_num_lines): Likewise.
5616 (edited_line::edited_line): Likewise.
5617 * final.c (asm_show_source): Likewise.
5618 * input.c (location_get_source_line): Convert return type
5619 from const char * to char_span, losing the final "line_len"
5621 (dump_location_info): Update for the above.
5622 (get_substring_ranges_for_loc): Likewise. Use a char_span
5623 when handling the literal within the line.
5624 (test_reading_source_line): Update for location_get_source_line
5625 returning a char_span.
5626 * input.h (class char_span): Move here from
5627 diagnostic-show-locus.c, converting from a struct to a class.
5628 Make data members private.
5629 (char_span::operator bool): New.
5630 (char_span::length): New.
5631 (char_span::get_buffer): New.
5632 (char_span::operator[]): New.
5633 (char_span::subspan): Make const.
5634 (char_span::xstrdup): New.
5635 (location_get_source_line): Convert return type from const char *
5636 to char_span, losing the final "line_size" param.
5638 2018-04-30 Jan Hubicka <jh@suse.cz>
5640 * lto-wrapper.c (ltrans_priorities): New static var.
5641 (cmp_priority): New.
5642 (run_gcc): Read priorities and if doing parallel build order
5643 the Makefile by them.
5645 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5647 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5649 2018-04-30 Richard Biener <rguenther@suse.de>
5651 * tree-cfg.c (verify_address): Remove base argument, add
5652 flag whether to check TREE_ADDRESSABLE and do that.
5653 (verify_expr): Remove.
5654 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5655 (verify_gimple_assign_single): Likewise.
5656 (verify_gimple_switch): Likewise.
5657 (verify_expr_location_1): Dereference tp once. Add (disabled)
5658 piece from verify_expr.
5659 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5661 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5663 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5665 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5667 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5668 (small_data_pattern): Likewise.
5669 (arc_rewrite_small_data): Likewise.
5670 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5671 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5672 (get_symbol_alignment): New function.
5673 (legitimate_small_data_address_p): Likewise.
5674 (legitimate_scaled_address): Update, call
5675 legitimate_small_data_address_p.
5676 (output_sdata): New static variable.
5677 (arc_print_operand): Update how we handle small data operands.
5678 (arc_print_operand_address): Likewise.
5679 (arc_legitimate_address_p): Update, use
5680 legitimate_small_data_address_p.
5681 (arc_rewrite_small_data_p): Remove.
5682 (arc_rewrite_small_data_1): Likewise.
5683 (arc_rewrite_small_data): Likewise.
5684 (small_data_pattern): Likewise.
5685 (compact_sda_memory_operand): Update to use
5686 legitimate_small_data_address_p and get_symbol_alignment.
5687 (prepare_move_operands): Don't rewite sdata pattern.
5688 (prepare_extend_operands): Remove.
5689 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5691 (zero_extendqisi2): Likewise.
5692 (zero_extendhisi2): Likewise.
5693 (extendqihi2): Likewise.
5694 (extendqisi2): Likewise.
5695 (extendhisi2): Likewise.
5699 * config/arc/constraints.md (Usd): Change it to memory constraint.
5701 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5703 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5704 as source of std instructions.
5705 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5706 allow 6-bit constants as source for store instructions.
5707 (movdi_insn): Update instruction pattern to allow 6-bit constants
5708 as source for store instructions.
5710 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5712 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5714 2018-04-30 Nathan Sidwell <nathan@acm.org>
5715 Sandra Loosemore <sandra@codesourcery.com>
5717 * dumpfile.c (dump_open): Allow '-' for stdout.
5718 * doc/invoke.texi (Developer Options): Document dump filename
5719 determination early. Document stdin/stdout selection.
5721 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5723 Microblaze Target: PIC data text relative
5725 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5726 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5728 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5729 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5730 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5731 New addressing mode for data-text relative position indepenedent code.
5732 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5733 'ADDRESS_SYMBOLIC_TXT_REL'.
5734 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5735 (microblaze_legitimate_pic_operand): Exclude function calls from
5736 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5737 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5739 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5740 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5741 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5742 for 'address + offset'.
5743 (microblaze_expand_prologue): Add new function prologue call for
5745 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5746 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5747 table in case of TARGET_PIC_DATA_TEXT_REL.
5748 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5749 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5750 Add new macros 'UNSPEC_TEXT',
5751 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5752 + exclude function calls from 'UNSPEC_PLT' in case of data text
5754 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5755 new target hook for generating address diff vector tables in case of
5757 * doc/tm.texi : Regenerate.
5758 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5759 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5760 of addr diff vector generation.
5761 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5762 target hook definition.
5763 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5764 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5765 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5766 option for data text relative.
5768 2018-04-30 Richard Biener <rguenther@suse.de>
5770 * tree-chrec.h (evolution_function_is_constant_p): Remove
5772 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5774 2018-04-30 Richard Biener <rguenther@suse.de>
5777 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5779 2018-04-30 Richard Biener <rguenther@suse.de>
5781 PR tree-optimization/28364
5782 PR tree-optimization/85275
5783 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5784 copying first exit test.
5786 2018-04-28 Mark Wielaard <mark@klomp.org>
5788 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5790 (dwarf_AT): Handle DW_AT_addr_base.
5791 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5793 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5796 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5797 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5798 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5799 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5801 2018-04-28 Richard Biener <rguenther@suse.de>
5803 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5804 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5805 to reflect use. Only add interesting stmts.
5807 2018-04-27 Martin Jambor <mjambor@suse.cz>
5810 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5811 the jump function allows for passing through aggregate values.
5813 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5815 * input.h (in_system_header_at): Convert from macro to inline
5817 (from_macro_expansion_at): Likewise.
5818 (from_macro_definition_at): Likewise.
5820 2018-04-27 Jeff Law <law@redhat.com>
5822 * config.gcc: Mark tile* targets as deprecated/obsolete.
5824 2018-04-27 Richard Biener <rguenther@suse.de>
5826 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5829 2018-04-27 Richard Biener <rguenther@suse.de>
5831 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5833 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5835 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5836 with Yd constraint. Set "preferred_for_speed" attribute from
5837 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5839 (*movdi_internal): Ditto.
5840 (movti_interunit splitters): Remove
5841 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5842 (movdi_interunit splitters): Ditto.
5843 * config/i386/constraints.md (Ye): Remove.
5844 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5846 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5849 * config/aarch64/constraints.md (Usg): Limit to 31.
5852 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5854 PR tree-optimization/85529
5855 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5856 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5857 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5858 zero extension or masking of the MSB bit.
5859 (optimize_range_tests): Add FIRST_BB argument, pass it through
5860 to optimize_range_tests_var_bound.
5861 (maybe_optimize_range_tests, reassociate_bb): Adjust
5862 optimize_range_tests callers.
5864 2018-04-26 Richard Biener <rguenther@suse.de>
5865 Jakub Jelinek <jakub@redhat.com>
5867 * cgraph.h (symbol_table): Just declare debug method here.
5868 * symtab.c (symbol_table::debug): Define.
5870 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5872 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5874 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5876 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5877 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5878 (*movdi_internal): Substitute Yi and Yj constraint with x
5879 and Ym and Yn constraint with y constraint. Update "isa"
5880 attribute and set "preferred_for_speed" attribute from
5881 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5882 (*movsi_internal): Ditto.
5883 (*movdf_internal): Ditto.
5884 (*movsf_internal): Ditto.
5885 (*zero_extendsidi2): Ditto.
5886 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5887 (sse2_loadld): Ditto.
5888 (*vec_extract<ssevecmodelower>_0): Ditto.
5889 (*vec_extractv4si_0_zext_sse4): Ditto.
5890 (vec_concatv2di): Ditto.
5891 (*vec_dup<mode>): Ditto.
5892 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5893 * config/i386/constraints.md (Yi): Remove.
5898 2018-04-26 Nathan Sidwell <nathan@acm.org>
5900 * dumpfile.c (dump_open): New.
5901 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5902 (dump_finish): Detect stdio/stderr by value not name.
5904 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5906 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5908 2018-04-26 Tom de Vries <tom@codesourcery.com>
5911 * config/nvptx/nvptx.c (verify_neutering_jumps)
5912 (verify_neutering_labels): New function
5913 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5915 2018-04-26 Tom de Vries <tom@codesourcery.com>
5918 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5919 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5922 2018-04-26 Richard Biener <rguenther@suse.de>
5923 Tom de Vries <tom@codesourcery.com>
5926 * lto-streamer-out.c (output_function): Fixup loops if required to match
5927 discovery done in the reader.
5929 2018-04-26 Richard Biener <rguenther@suse.de>
5931 PR tree-optimization/85116
5932 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5933 have a loop exit from the single latch predecessor. Remove
5934 case of header with just condition.
5935 (ch_base::copy_headers): Exclude infinite loops from any
5937 (pass_ch::execute): Record exits.
5939 2018-04-26 Richard Biener <rguenther@suse.de>
5941 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5942 prologue cost vector and pass it to vect_get_load_cost.
5943 (vect_get_peeling_costs_all_drs): Likewise.
5944 (vect_peeling_hash_get_lowest_cost): Likewise.
5945 (vect_enhance_data_refs_alignment): Likewise.
5947 2018-04-26 Richard Biener <rguenther@suse.de>
5950 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5951 checking of integer<->pointer conversions.
5952 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5953 sign-/zero-extending pointer types.
5954 (expand_omp_for_static_chunk): Likewise.
5956 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5957 Jean Lee <xiaoyur347@gmail.com>
5959 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5960 (TARGET_ASAN_SHADOW_OFFSET): Define.
5961 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5962 true for -fsanitize=address.
5964 2018-04-25 Mark Wielaard <mark@klomp.org>
5966 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5969 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5971 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5972 than "alu", remove explicit "memory" and "imm_disp" attributes.
5973 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5976 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5977 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5978 gen_lowpart_no_emit.
5980 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5983 * config/i386/i386.c (ix86_expand_builtin): Change memory
5984 operand to XI, extend p0 to Pmode.
5985 * config/i386/i386.md: Change unspec volatile and operand
5986 1 mode to XI, change operand 0 mode to P.
5988 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5990 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5991 GET_MODE_MASK before any checking.
5992 (nds32_can_use_bset_p): Likewise.
5993 (nds32_can_use_btgl_p): Likewise.
5995 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5997 * config/nds32/nds32-doubleword.md: New define_split pattern for
5998 illegal register number.
6000 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6002 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6004 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6006 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6008 2018-04-25 Richard Biener <rguenther@suse.de>
6010 * lto-streamer.h (LTO_major_version): Bump to 8.
6012 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6014 * BASE-VER: Set to 9.0.0.
6016 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6018 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6019 in __abskf2 and __powikf2.
6021 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6024 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6025 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6026 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6027 Use the above on operand 2. Reindent.
6028 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6030 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6033 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6035 * config/i386/i386.opt (mcet): Removed.
6036 * doc/install.texi: Remove -mcet documentation.
6037 * doc/invoke.texi: Likewise.
6039 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6042 * doc/install.texi: Remove -mcet from bootstrap-cet.
6044 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6047 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6048 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6052 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6053 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6054 containing a CONST_VECTOR.
6056 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6058 * doc/install.texi: Update newlib dependency for nvptx.
6060 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6063 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6064 instead of INTVAL when shifting x left.
6066 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6068 PR tree-optimization/85478
6069 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6070 vect_grouped_store_supported for single element vectors.
6072 2018-04-24 Richard Biener <rguenther@suse.de>
6075 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6076 load cost increase to the case of non-constant step.
6078 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6081 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6082 destination if any_malformed_asm.
6084 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6087 * expr.c (store_field): In the bitfield case, if the value comes from
6088 a function call and is returned in registers by means of a PARALLEL,
6089 do not change the mode of the temporary unless BLKmode and VOIDmode.
6091 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6093 PR rtl-optimization/85423
6094 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6095 dependencies to debug insns when the previous insn is non-debug.
6097 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6099 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6100 enums into a single definition.
6101 (fls): Fix predicates and printing.
6104 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6106 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6107 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6108 and short u6 immediate.
6109 (check_if_valid_sleep_operand): Remove.
6110 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6112 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6114 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6115 flag_always_save_lp condition.
6116 * config/nds32/nds32.opt (malways-save-lp): New option.
6118 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6120 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6121 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6122 * config/nds32/nds32.h
6123 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6124 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6126 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6128 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6129 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6131 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6132 Chung-Ju Wu <jasonwucj@gmail.com>
6134 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6136 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6137 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6139 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6141 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6143 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6145 * config/nds32/nds32-protos.h (nds32_data_alignment,
6146 nds32_local_alignment): Declare.
6147 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6148 nds32_local_alignment): New functions.
6149 (TARGET_CONSTANT_ALIGNMENT): Define.
6150 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6152 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6154 * config/nds32/nds32.c
6155 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6156 (TARGET_MODES_TIEABLE_P): Likewise.
6158 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6160 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6163 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6164 Chung-Ju Wu <jasonwucj@gmail.com>
6166 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6167 for unaligned access.
6168 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6170 * config/nds32/nds32-intrinsic.md: Likewise.
6171 * config/nds32/nds32_intrinsic.h: Likewise.
6172 * config/nds32/nds32.h (nds32_builtins): Likewise.
6173 * config/nds32/nds32.opt (munaligned-access): New option.
6174 * config/nds32/nds32.c (nds32_asm_file_start): Display
6175 flag_unaligned_access status.
6177 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6179 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6180 -mno-relax is present.
6181 * config/riscv/linux.h (LINK_SPEC): Ditto.
6183 2018-04-20 Martin Sebor <msebor@redhat.com>
6186 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6188 (gimple_fold_builtin_stxcpy_chk): Same.
6189 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6191 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6194 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6195 __powikf2 when long double is IEEE 128-bit.
6197 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6199 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6200 step to make sure stack always aligned.
6202 2018-04-20 Carl Love <cel@us.ibm.com>
6205 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6206 size check for arg0.
6208 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6209 Tom de Vries <tom@codesourcery.com>
6212 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6213 Emit insns for calls too.
6214 (nvptx_find_par): Always look for worker-level predecessor insn.
6215 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6217 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6218 (nvptx_process_pars): Propagate frames for calls.
6220 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6223 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6225 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6226 (ix86_handle_option): Don't handle OPT_mibt.
6227 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6229 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6231 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6232 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6233 (ix86_target_macros): Define __CET__ with flag_cf_protection
6234 for -fcf-protection.
6235 * config/i386/i386.c (isa2_opts): Remove -mibt.
6236 * config/i386/i386.h (TARGET_IBT): Removed.
6237 (TARGET_IBT_P): Likewise.
6238 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6239 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6240 * config/i386/i386.opt (mcet): Update help message.
6243 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6244 -mcet as an alias for -mshstk.
6246 2018-04-20 Richard Biener <rguenther@suse.de>
6249 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6250 complexity by forcing a single use of the multiply operand.
6252 2018-04-20 Martin Jambor <mjambor@suse.cz>
6255 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6256 recursion dependency to only apply to non-clones.
6258 2018-04-20 Martin Jambor <mjambor@suse.cz>
6261 * ipa-cp.c (create_specialized_node): Check that clones of
6262 self-recursive edges exist during IPA-CP.
6264 2018-04-19 Toon Moene <toon@moene.org>
6266 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6269 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6271 PR tree-optimization/85467
6272 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6273 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6274 VECTOR_CST element to type.
6276 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6279 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6280 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6281 (builtin_longjmp): Likewise.
6282 (save_stack_nonlocal): New pattern.
6283 (restore_stack_nonlocal): Likewise.
6285 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6288 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6289 Replace ASM_OUTPUT_LABEL with fprintf.
6291 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6294 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6295 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6296 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6297 define __IBT__ and __SHSTK__ for -fcf-protection.
6298 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6300 (ix86_trampoline_init): Likewise.
6301 (x86_output_mi_thunk): Likewise.
6302 (ix86_notrack_prefixed_insn_p): Likewise.
6303 (ix86_option_override_internal): Don't disallow -fcf-protection.
6304 * config/i386/i386.md (rdssp<mode>): Also enable for
6306 (incssp<mode>): Likewise.
6307 (nop_endbr): Likewise.
6308 * config/i386/i386.opt (mcet): Change help message to built-in
6312 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6313 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6314 enable CET built-in functions.
6316 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6318 * common/config/i386/i386-common.c
6319 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6320 OPTION_MASK_ISA_MOVDIRI_UNSET,
6321 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6322 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6323 * config.gcc (movdirintrin.h): New header.
6324 * config/i386/cpuid.h (bit_MOVDIRI,
6325 bit_MOVDIR64B): New bits.
6326 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6328 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6329 (VOID, PVOID, PCVOID)): New function types.
6330 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6331 __builtin_ia32_directstoreu_u64,
6332 __builtin_ia32_movdir64b): New builtins.
6333 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6334 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6336 (ix86_valid_target_attribute_inner_p): Ditto.
6337 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6338 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6339 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6340 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6341 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6342 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6343 (movdiri<mode>, movdir64b_<mode>): New.
6344 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6345 * config/i386/immintrin.h: Include movdirintrin.h.
6346 * config/i386/movdirintrin.h: New file.
6347 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6349 2018-04-19 Richard Biener <rguenther@suse.de>
6352 * cfg.c (clear_bb_flags): When loop state says we have
6353 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6355 2018-04-19 Richard Biener <rguenther@suse.de>
6357 PR tree-optimization/84737
6358 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6359 copying restrict info.
6360 (vect_setup_realignment): Use it.
6361 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6362 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6363 the first DR to all generated stores.
6364 (vectorizable_load): Likewise for loads.
6366 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6368 PR tree-optimization/85446
6369 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6370 the integral and pointer types to have the same precision.
6372 * doc/install.texi: Document --disable-cet being the default and
6375 2018-04-18 Martin Liska <mliska@suse.cz>
6377 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6380 2018-04-18 Martin Liska <mliska@suse.cz>
6383 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6386 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6387 arguments if they are comparable.
6389 2018-04-18 Martin Liska <mliska@suse.cz>
6392 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6395 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6398 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6401 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6402 ENDBR after calling __morestack.
6404 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6407 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6408 by using gcc_base_ver to generate a gcc_driver_version, and use
6409 it when generating GCC_DRIVER_NAME.
6410 * configure: Regenerate.
6412 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6415 * config.gcc: Obsolete powerpc*-*-*spe*.
6417 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6420 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6421 (stabstr_D): Change type of unum from unsigned int to
6422 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6425 2018-04-17 Jim Wilson <jimw@sifive.com>
6428 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6429 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6430 Set arg_pointer_offset after using pretend_args_size.
6432 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6434 PR rtl-optimization/85431
6435 * dse.c (record_store): Ignore zero width stores.
6438 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6439 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6440 __builtin_stack_restore rather than after it.
6441 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6442 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6443 argument instead of virtual_dynamic_stack_rtx.
6445 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6447 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6449 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6450 Add note to error message to explain internal mapping of overloaded
6451 built-in function name to non-overloaded built-in function name.
6452 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6455 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6458 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6459 where the inputs overlap with the output.
6461 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6463 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6464 (=v, v) alternative and explicit "memory" attribute.
6465 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6466 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6468 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6469 "sselog1" type instead of "sselog".
6470 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6471 "sselog". Remove explicit "memory" attribute.
6472 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6473 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6475 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6476 "sselog1" type instead of "sselog". Remove explicit "memory"
6478 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6479 use "sselog1" type instead of "sselog". Remove explicit "memory"
6481 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6482 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6484 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6485 "sselog1" type instead of "sselog". Remove explicit "memory"
6487 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6488 use "sselog1" type instead of "sselog". Remove explicit "memory"
6492 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6495 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6498 2018-04-17 Martin Jambor <mjambor@suse.cz>
6501 * ipa-cp.c (create_specialized_node): Call
6502 expand_all_artificial_thunks if necessary.
6504 2018-04-17 Martin Liska <mliska@suse.cz>
6507 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6508 in message, remote space in between '_G' and '('.
6510 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6513 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6514 avx512f_vmcmp<mode>3<round_saeonly_name>,
6515 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6516 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6517 avx512f_rndscale<mode><round_saeonly_name>,
6518 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6519 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6520 Use %<iptr>2 instead of %2 for -masm=intel.
6521 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6522 avx512f_vcvttss2usi<round_saeonly_name>,
6523 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6525 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6526 avx512f_vcvttsd2usi<round_saeonly_name>,
6527 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6528 Use %q1 instead of %1 for -masm=intel.
6529 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6530 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6531 of %3 for -masm=intel.
6532 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6534 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6536 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6538 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6539 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6541 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6543 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6544 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6545 %g1 and one with %0 and %1.
6546 (avx512er_vmrcp28<mode><round_saeonly_name>,
6547 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6549 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6550 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6551 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6552 of %0 and %{%4%} for -masm=intel.
6553 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6554 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6555 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6556 order of %0 and %{%5%}%{z%} for -masm=intel.
6558 2018-04-17 Jan Hubicka <jh@suse.cz>
6561 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6563 2018-04-17 Martin Liska <mliska@suse.cz>
6566 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6567 for target_clone error message. Make default implementation
6568 clone to be a local declaration.
6569 (separate_attrs): Add new argument and check for an empty
6571 (expand_target_clones): Handle it.
6572 (ipa_target_clone): Make redirection just for target_clones
6575 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6576 Tom de Vries <tom@codesourcery.com>
6579 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6580 tiled basic blocks without omp continue statements.
6582 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6585 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6586 vec_extract expression as having side effects to make sure it gets
6589 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6592 * config/i386/i386.c (get_builtin_code_for_version): Check
6595 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6598 * gcc/config.gcc: Support "skylake".
6599 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6601 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6602 (processor_target_table): Add "skylake".
6603 (ix86_option_override_internal): Add "skylake".
6604 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6605 PROCESSOR_CANNONLAKE.
6606 (get_builtin_code_for_version): Fix priority for
6607 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6608 PROCESSOR_SKYLAKE-AVX512.
6609 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6610 (processor_type): Add PROCESSOR_SKYLAKE.
6612 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6613 Jason Merrill <jason@redhat.com>
6616 * convert.c (convert_to_integer_1): Use direct recursion for
6617 enumeral types and types with a precision less than the number
6618 of bits in their mode.
6620 2018-04-16 Julia Koval <julia.koval@intel.com>
6623 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6624 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6626 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6629 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6630 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6631 and -mno-direct-move.
6633 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6636 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6637 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6638 Ensure negative shifts result in {0}.
6640 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6642 PR rtl-optimization/79916
6643 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6644 regs (if any) to define how to gnerate SD moves when LRA is in
6647 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6649 PR rtl-optimization/85393
6650 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6651 * except.c (expand_dw2_landing_pad_for_region): Make static.
6652 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6653 a label and unconditional jump to old_bb, rather than
6654 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6657 PR rtl-optimization/85376
6658 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6659 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6660 instead of a specific value.
6662 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6663 Bin Cheng <bin.cheng@arm.com>
6665 PR tree-optimization/82965
6666 PR tree-optimization/83991
6667 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6668 by_profile_only parameter.
6669 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6670 information if the loop was predicted to iterate too many times.
6671 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6673 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6676 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6679 2018-04-13 Martin Liska <mliska@suse.cz>
6680 Jakub Jelinek <jakub@redhat.com>
6683 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6684 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6685 * builtins.c (expand_builtin_memory_copy_args): Use
6686 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6687 handle dest_addr == pc_rtx.
6689 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6692 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6693 asked to not generate direct moves.
6694 (fix_trunc<mode>si2_stfiwx): Similar.
6695 (fix_trunc<mode>si2_internal): Similar.
6697 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6700 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6701 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6702 lookup if dest in some wider mode is known to be const0_rtx and
6703 if so, record permanent equivalence for it to be ZERO_EXTEND of
6704 the narrower mode destination.
6706 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6708 * lto-streamer-out.c (output_function): Revert 259346.
6709 * omp-expand.c (expand_oacc_for): Likewise.
6711 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6713 PR rtl-optimization/85354
6714 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6715 * sel-sched.c (sel_global_init): ... here.
6717 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6720 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6721 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6722 mode for PE-COFF targets.
6723 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6724 (i386_pe_asm_lto_end): Likewise.
6725 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6726 (TARGET_ASM_LTO_END): Likewise.
6727 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6728 (i386_pe_asm_lto_start): New function.
6729 (i386_pe_asm_lto_end): Likewise.
6731 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6732 Richard Biener <rguenther@suse.de>
6735 * lto-streamer-out.c (output_function): Fix CFG loop state before
6737 * omp-expand.c (expand_oacc_for): Handle calls to internal
6738 functions like regular functions.
6740 2018-04-12 Richard Biener <rguenther@suse.de>
6743 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6744 for the early LTO debug to properly generate references to it
6745 during DIE emission. Do not re-use that for the skeleton for
6747 (dwarf2out_early_finish): Likewise.
6749 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6752 * config/i386/sse.md
6753 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6754 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6755 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6756 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6757 and output is a reg, avoid creating invalid lowpart subreg, but
6758 instead split into a 512-bit move. Don't split if not AVX512VL,
6759 input is xmm16+ reg and output is a mem.
6760 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6761 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6762 xmm16+ reg and output is a mem.
6764 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6766 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6767 also for flag_dwarf2_cfi_asm.
6769 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6771 PR rtl-optimization/85342
6772 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6773 a bool scalar var inside of the loop instead. Don't try to update
6774 recog_data.operand after failed apply_change_group.
6776 2018-04-12 Tom de Vries <tom@codesourcery.com>
6779 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6780 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6781 array with flexible array member as array without given dimension.
6782 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6783 argument for undefined param to true.
6785 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6788 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6789 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6790 from PowerPC section.
6791 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6792 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6793 help text that is too long.
6794 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6795 help text that is too long.
6796 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6797 help text that is too long.
6799 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6801 * config/alpha/alpha.md (stack_probe_internal): Rename
6802 from "probe_stack". Update all callers.
6804 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6806 PR rtl-optimization/84566
6807 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6808 sched_macro_fuse_insns.
6810 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6813 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6814 (compute_block_dependences): ... from here.
6816 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6818 PR tree-optimization/85331
6819 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6820 from int to HOST_WIDE_INT.
6822 2018-04-11 Martin Jambor <mjambor@suse.cz>
6825 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6826 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6827 not the same as the source val.
6828 (cgraph_edge_brings_value_p): New parameter.
6829 (gather_edges_for_value): Pass destination value to
6830 cgraph_edge_brings_value_p.
6831 (perhaps_add_new_callers): Likewise.
6832 (get_info_about_necessary_edges): Likewise and exclude values brought
6833 only by self-recursive edges.
6834 (create_specialized_node): Redirect only clones of self-calling edges.
6835 (+self_recursive_pass_through_p): New function.
6836 (find_more_scalar_values_for_callers_subset): Use it.
6837 (find_aggregate_values_for_callers_subset): Likewise.
6838 (known_aggs_to_agg_replacement_list): Removed.
6839 (decide_whether_version_node): Re-calculate known constants for all
6840 remaining context clones.
6842 2018-04-11 Richard Biener <rguenther@suse.de>
6845 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6846 from early DWARF output.
6847 (dwarf2out_early_finish): Output line info unconditionally into
6848 early DWARF and add reference to it.
6850 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6853 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6854 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6855 other than V2DFmode using iptr mode attribute.
6856 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6858 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6860 PR rtl-optimization/84659
6861 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6863 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6866 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6868 (output_loc_list): Pass address of a dummy size variable even in the
6869 locview handling loop.
6870 (index_location_lists): Add comment on why skip_loc_list_entry can't
6873 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6876 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6879 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6882 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6883 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6884 and -mstring-compare-inline-limit.
6886 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6889 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6890 for stack clash protection in a register whenever we need it to be in
6893 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6895 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6896 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6898 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6901 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6903 (mlong-double-): Ditto.
6904 * config/rs6000/sysv4.opt (msdata=): Ditto.
6905 (mtls-size=): Ditto.
6907 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6909 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6910 erroneous entries for
6911 "vector int vec_ldl (int, long int *)", and
6912 "vector unsigned int vec_ldl (int, unsigned long int *)".
6913 Add comments and entries for
6914 "vector bool char vec_ldl (int, bool char *)",
6915 "vector bool short vec_ldl (int, bool short *)",
6916 "vector bool int vec_ldl (int, bool int *)",
6917 "vector bool long long vec_ldl (int, bool long long *)",
6918 "vector pixel vec_ldl (int, pixel *)",
6919 "vector long long vec_ldl (int, long long *)",
6920 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6921 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6922 type tree bool_long_long_type_node and correct definition of
6923 bool_V2DI_type_node to make reference to this new type tree.
6924 (rs6000_mangle_type): Replace erroneous reference to
6925 bool_long_type_node with bool_long_long_type_node.
6926 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6927 comments to emphasize sign distinctions for char and int types and
6928 replace RS6000_BTI_bool_long constant with
6929 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6930 use of RS6000_BTI_pixel.
6931 (bool_long_type_node): Remove this macro definition.
6932 (bool_long_long_type_node): New macro definition
6934 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6936 PR rtl-optimization/85300
6937 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6938 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6939 simplify_unary_operation fails.
6941 2018-04-10 Martin Liska <mliska@suse.cz>
6943 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6944 cgraph_edge and ipa_ref.
6946 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6950 * config/i386/sse.md
6951 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6952 computation of the VEC_MERGE selector from mask.
6953 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6954 Fix decoding of the VEC_MERGE selector into mask.
6956 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6958 PR tree-optimization/85286
6959 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6961 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6963 * final.c (final_1): Set insn_last_address as well as
6964 insn_current_address.
6966 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6969 * explow.c (emit_stack_probe): Call validize_mem on memory location
6970 before passing it to gen_probe_stack. Create address operand and
6971 legitimize it for the probe_stack_address case.
6973 2018-04-09 Jan Hubicka <jh@suse.cz>
6976 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6977 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6978 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6980 (free_lang_data): Rebuild type inheritance graph.
6982 2018-04-09 Martin Sebor <msebor@redhat.com>
6984 * invoke.texi (-finline-small-functions): Mention other optimization
6986 (-findirect-inlining, -fpartial-inlining): Same.
6987 (-finline-functions-called-once): Same.
6988 (-freorder-blocks-and-partition): Same.
6990 2018-04-09 Jan Hubicka <jh@suse.cz>
6993 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6994 jumps; choose last target that matches the criteria (i.e.
6995 no partition changes for non-crossing jumps).
6996 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6997 support for redirecting crossing jumps to non-crossing.
6999 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7001 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7002 also for naked functions.
7004 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7006 * config/arc/arc.md (add_shift): New pattern.
7007 (add_shift2): Likewise.
7008 (sub_shift): Likewise.
7009 (sub_shift_cmp0_noout): Likewise.
7010 (compare_si_ashiftsi): Likewise.
7011 (xbfu_cmp0_noout): New combine pattern.
7012 (xbfu_cmp0"): Likewise.
7013 (movsi_set_cc_insn): Place the predicable variant first.
7014 (commutative_binary_cmp0_noout): Remove clobber.
7015 (commutative_binary_cmp0): New pattern.
7016 (noncommutative_binary_cmp0): Likewise.
7017 (noncommutative_binary_cmp0_noout): Likewise.
7018 (noncommutative_binary_comparison_result_used): Removed.
7019 (rsub_cmp0): New pattern.
7020 (rsub_cmp0_noout): Likewise.
7021 (extzvsi): Changed, keep only meaningful variants.
7022 (SQH, SEZ): New iterators.
7023 (SQH_postfix): New mode attribute.
7024 (SEZ_prefix): New code attribute.
7025 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7026 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7027 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7029 (noncommutative_operator): Check the availability of barrel
7032 2018-04-09 Richard Biener <rguenther@suse.de>
7034 PR tree-optimization/85284
7035 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7036 Only use the niter constraining form of simple_iv when the exit
7039 2018-04-09 Tom de Vries <tom@codesourcery.com>
7042 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7043 (define_expand "*memory_barrier"): New define_expand.
7044 (define_insn "memory_barrier"): New insn.
7046 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7048 PR rtl-optimization/80463
7049 PR rtl-optimization/83972
7050 PR rtl-optimization/83480
7052 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7053 correct producer for the insn.
7054 (tidy_control_flow): Fixup seqnos in case of debug insns.
7056 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7058 PR rtl-optimization/83913
7060 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7061 different sched-times when merging exprs.
7063 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7065 PR rtl-optimization/83962
7067 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7068 tidy_fallthru_edge and tidy_control_flow.
7070 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7072 PR rtl-optimization/83530
7074 * sel-sched.c (force_next_insn): New global variable.
7075 (remove_insn_for_debug): When force_next_insn is true, also leave only
7076 next insn in the ready list.
7077 (sel_sched_region): When the region wasn't scheduled, make another pass
7078 over it with force_next_insn set to 1.
7080 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7082 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7084 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7085 for interrupt control.
7086 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7087 functions for interrupt control.
7088 * config/nds32/nds32-intrinsic.md: Likewise.
7089 * config/nds32/nds32_intrinsic.h: Likewise.
7090 * config/nds32/nds32.h (nds32_builtins): Likewise.
7092 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7094 * config/nds32/nds32.c (nds32_init_machine_status,
7095 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7096 strict_aligned_p field.
7097 (nds32_expand_to_rtl_hook): New function.
7098 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7099 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7101 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7102 Chung-Ju Wu <jasonwucj@gmail.com>
7104 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7105 * config/nds32/nds32-n7.md: New file.
7106 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7107 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7109 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7110 * config/nds32/nds32.md (pipeline_model): Add n7.
7111 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7112 * config/nds32/pipelines.md: Include n7 settings.
7114 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7115 Chung-Ju Wu <jasonwucj@gmail.com>
7117 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7118 * config/nds32/nds32-e8.md: New file.
7119 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7120 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7122 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7123 * config/nds32/nds32.md (pipeline_model): Add e8.
7124 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7125 * config/nds32/pipelines.md: Include e8 settings.
7127 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7128 Chung-Ju Wu <jasonwucj@gmail.com>
7130 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7131 * config/nds32/nds32-n8.md: New file.
7132 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7133 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7135 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7136 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7137 * config/nds32/nds32.md (pipeline_model): Add n8.
7138 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7139 * config/nds32/pipelines.md: Include n8 settings.
7141 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7142 Chung-Ju Wu <jasonwucj@gmail.com>
7144 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7145 * config/nds32/nds32-n9-2r1w.md: New file.
7146 * config/nds32/nds32-n9-3r2w.md: New file.
7147 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7148 nds32_register_ports): New or modify for cpu n9.
7149 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7151 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7152 * config/nds32/nds32-utils.c: New file.
7153 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7154 TARGET_MUL_SLOW): Define.
7155 * config/nds32/nds32.md (pipeline_model): New attribute.
7156 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7157 New options that support cpu n9.
7158 * config/nds32/pipelines.md: Include n9 settings.
7159 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7161 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7163 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7164 information if necessary.
7165 (output_cond_branch_compare_zero): Likewise.
7166 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7167 (nds32_target_alignment): Refine for alignment.
7168 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7169 (FUNCTION_BOUNDARY): Modify.
7170 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7172 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7174 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7176 * config/nds32/constants.md (unspec_volatile_element): Add values for
7177 TLB operation and data prefetch.
7178 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7179 functions for TLB operation and data prefetch.
7180 * config/nds32/nds32-intrinsic.md: Likewise.
7181 * config/nds32/nds32_intrinsic.h: Likewise.
7182 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7183 (nds32_print_operand): Likewise.
7184 * config/nds32/nds32.h (nds32_builtins): Likewise.
7186 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7187 Andrew Pinski <pinsika@gcc.gnu.org>
7190 * match.pd: Use constant_boolean_node of correct type instead of
7191 boolean_true_node or boolean_false_node for simplifying
7192 pointer comparisons to zero.
7194 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7196 PR tree-optimization/80021
7197 * tree.c (verify_type_variant): Make error call in verify_variant_match
7198 translatable and remove final full stop.
7200 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7202 * config/nds32/constants.md (unspec_volatile_element): Add
7203 UNSPEC_VOLATILE_EH_RETURN.
7204 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7205 nds32_output_stack_pop): Support dwarf exception handling process.
7206 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7207 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7208 exception handling process.
7209 (nds32_compute_stack_frame): Likewise.
7210 (nds32_return_addr_rtx): Likewise.
7211 (nds32_initial_elimination_offset): Likewise.
7212 (nds32_expand_prologue): Likewise.
7213 (nds32_expand_epilogue): Likewise.
7214 (nds32_dynamic_chain_address): New function.
7215 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7217 (DYNAMIC_CHAIN_ADDRESS): Define.
7218 (EH_RETURN_DATA_REGNO): Define.
7219 (EH_RETURN_STACKADJ_RTX): Define.
7220 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7221 patterns for dwarf exception handling.
7223 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7225 * config/nds32/nds32.h: Clean up obsolete macros.
7227 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7229 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7230 Add enum values for particular instructions.
7231 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7232 particular intrinsic functions.
7233 * config/nds32/nds32-intrinsic.md: Likewise.
7234 * config/nds32/nds32_intrinsic.h: Likewise.
7235 * config/nds32/nds32.h (nds32_builtins): Likewise.
7236 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7237 (btst, ave): New patterns for particular instructions.
7239 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7241 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7242 Add enum values for atomic load/store and memory sync.
7243 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7245 * config/nds32/nds32-intrinsic.md: Likewise.
7246 * config/nds32/nds32_intrinsic.h: Likewise.
7247 * config/nds32/nds32.h (nds32_builtins): Likewise.
7249 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7251 PR tree-optimization/85257
7252 * fold-const.c (native_encode_vector): If not all elts could fit
7253 and off is -1, return 0 rather than offset.
7254 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7255 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7256 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7257 adjust buffer in native_interpret_expr call.
7259 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7261 * config/nds32/constants.md (unspec_volatile_element): Add cache
7262 control enum values.
7263 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7264 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7265 * config/nds32/nds32.c (nds32_cctl_names): New.
7266 (nds32_print_operand): Handle cache control register names.
7267 * config/nds32/nds32.h (nds32_builtins): New enum values.
7268 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7270 * config/nds32/nds32.md (type): Add mmu.
7271 * config/nds32/pipelines.md (simple_insn): Add mmu.
7273 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7275 * config/nds32/nds32.md (type): Remove call.
7276 * config/nds32/pipelines.md (simple_insn): Likewise.
7278 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7280 * config/nds32/constants.md (unspec_volatile_element): Add
7281 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7282 UNSPEC_VOLATILE_FMFCFG.
7283 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7284 description for fmfcfg and fmfcsr.
7285 (bdesc_1arg): Add fmtcsr.
7286 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7287 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7288 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7289 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7290 unspec_fmfcfg): New patterns.
7291 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7292 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7293 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7294 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7295 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7296 __nds32__fmfcfg): Define.
7298 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7300 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7301 intrinsic register names.
7302 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7303 intrinsic register enum values and macros.
7305 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7307 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7308 for load/store addressing form.
7309 (nds32_print_operand_address): Likewise.
7311 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7314 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7315 based on LABEL_REF. Remove useless assertion.
7316 (pic_address_needs_scratch): Fix formatting.
7317 (sparc_legitimize_pic_address): Minor tweaks.
7318 (sparc_delegitimize_address): Adjust assertion accordingly.
7319 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7320 into symbolic_operand.
7321 (movsi_high_pic_label_ref): Likewise.
7322 (movsi_lo_sum_pic_label_ref): Likewise.
7323 (movdi_pic_label_ref): Likewise.
7324 (movdi_high_pic_label_ref): Likewise.
7325 (movdi_lo_sum_pic_label_ref): Likewise.
7327 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7329 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7330 custom LIB_SPEC setup.
7332 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7333 Kito Cheng <kito.cheng@gmail.com>
7335 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7336 * config/riscv/freebsd.h: New.
7338 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7340 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7341 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7344 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7345 Kito Cheng <kito.cheng@gmail.com>
7347 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7348 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7349 * config/nds32/nds32-protos.h (nds32_output_call,
7350 nds32_output_return): Declare.
7351 * config/nds32/nds32.md: Refine all the call and return patterns.
7353 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7356 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7357 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7359 PR rtl-optimization/84872
7360 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7361 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7364 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7366 * expr.c (copy_blkmode_to_reg): Revert 254862.
7367 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7369 2018-04-06 Richard Biener <rguenther@suse.de>
7372 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7373 after seeing a component reference with an adjacent field. Treat
7374 refs to arrays at struct end of external decls similar to
7375 refs to unconstrained commons.
7377 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7380 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7381 look through SAVE_EXPRs with non-side-effects argument. Adjust
7383 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7386 2018-04-06 Richard Biener <rguenther@suse.de>
7389 * alias.c (find_base_term): New wrapper around find_base_term
7390 unwinding CSELIB_VAL_PTR changes.
7391 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7394 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7396 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7398 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7399 constant definitions.
7400 ("nop"): lr 0,0 -> nopr r0
7401 ("nop_lr0", "nop_lr1"): New insn definitions.
7403 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7405 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7406 NDS32_V3PUSH_AVAILABLE_P macro.
7408 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7409 Chung-Ju Wu <jasonwucj@gmail.com>
7411 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7412 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7413 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7414 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7415 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7416 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7417 * config/nds32/constraints.md: New constraints and checking for hard
7418 float configuration.
7419 * config/nds32/iterators.md: New mode iterator and attribute for hard
7420 float configuration.
7421 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7423 * config/nds32/nds32-fpu.md: New file.
7424 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7425 deal with hard float code generation.
7426 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7428 (abi_type, float_reg_number): New enum type.
7429 * config/nds32/nds32-predicates.c: New predicates for hard float.
7430 * config/nds32/nds32-protos.h: Declare functions for hard float.
7431 * config/nds32/nds32.c: Implementation for hard float configuration.
7432 * config/nds32/nds32.h: Definitions for hard float configuration.
7433 * config/nds32/nds32.md: Include hard float machine description and
7434 modify patterns for hard float configuration.
7435 * config/nds32/nds32.opt: New options for hard float configuration.
7436 * config/nds32/predicates.md: New predicates for hard float
7439 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7441 * common/config/nds32/nds32-common.c
7442 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7444 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7447 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7448 CONSTRUCTOR_ELT (ctor, ...)->value.
7450 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7453 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7455 2018-04-05 Tom de Vries <tom@codesourcery.com>
7458 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7461 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7462 Kito Cheng <kito.cheng@gmail.com>
7464 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7465 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7466 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7467 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7469 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7470 Kito Cheng <kito.cheng@gmail.com>
7472 * config/nds32/constraints.md (Ufe): New memory constraint.
7473 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7474 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7475 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7477 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7478 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7480 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7482 * config/nds32/nds32.md: Use optimize_size in the condition for
7483 alu-shift instructions.
7485 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7487 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7489 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7491 * config/nds32/nds32.md (negsi2): Refine pattern.
7493 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7494 Chung-Ju Wu <jasonwucj@gmail.com>
7496 * config/nds32/iterators.md (shift_rotate): New code iterator.
7497 (shift): New code attribute.
7498 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7499 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7500 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7501 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7502 bit-wise operations.
7503 (andsi3, *andsi3): Ditto.
7504 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7505 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7506 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7507 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7508 nds32_ior_operand, nds32_xor_operand): New predicates.
7510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7512 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7513 (addsi3, subsi3): ... this.
7515 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7517 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7519 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7521 * config/nds32/nds32.md: Adjust indention.
7523 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7525 * config/nds32/nds32.md (feature): New attribute.
7527 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7529 * config/nds32/nds32.md (subtype): New attribute.
7531 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7534 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7535 expansion to perform a bitwise AND of the argument followed by a
7536 boolean negation of the result.
7538 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7540 PR rtl-optimization/84878
7541 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7542 the basic block. Assert the use reference is not artificial and that
7543 it has an associated insn.
7545 2018-04-04 Michael Matz <matz@suse.de>
7547 * builtins.c (compute_objsize): Pass correct operand
7548 to array_at_struct_end_p.
7550 2018-04-04 Richard Biener <rguenther@suse.de>
7553 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7554 from contexts for DINFO_LEVEL_TERSE and below.
7556 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7558 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7559 resiter_operand condition.
7560 * config/nds32/nds32.md (*move<mode>): Ditto.
7562 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7563 Monk Chiang <sh.chiang04@gmail.com>
7565 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7567 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7569 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7571 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7572 Kito Cheng <kito.cheng@gmail.com>
7574 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7575 nds32_cond_code_str, output_cond_branch,
7576 output_cond_branch_compare_zero, nds32_expand_cbranch,
7577 nds32_expand_cstore, nds32_expand_movcc,
7578 nds32_output_cbranchsi4_equality_zero,
7579 nds32_output_cbranchsi4_equality_reg,
7580 nds32_output_cbranchsi4_equality_reg_or_const_int,
7581 nds32_output_cbranchsi4_greater_less_zero: New functions.
7582 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7583 nds32_expand_cstore, nds32_expand_movcc,
7584 nds32_output_cbranchsi4_equality_zero,
7585 nds32_output_cbranchsi4_equality_reg,
7586 nds32_output_cbranchsi4_equality_reg_or_const_int,
7587 nds32_output_cbranchsi4_greater_less_zero): Declare.
7588 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7589 nds32_rimm11s_operand): New predicates.
7590 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7591 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7594 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7596 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7597 * config/nds32/nds32.md: Ditto.
7598 * config/nds32/pipelines.md: Ditto.
7600 2018-04-04 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/85168
7603 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7604 propagating abnormals.
7606 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7608 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7610 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7611 Kito Cheng <kito.cheng@gmail.com>
7613 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7614 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7615 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7616 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7617 * config/nds32/nds32.md (sibcall_internal): New.
7618 (sibcall_register): Remove.
7619 (sibcall_immediate): Remove.
7620 (sibcall_value_internal): New.
7621 (sibcall_value_register): Remove.
7622 (sibcall_value_immediate): Remove.
7623 * config/nds32/predicates.md (nds32_general_register_operand): New.
7624 (nds32_call_address_operand): New.
7626 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7628 PR rtl-optimization/85167
7629 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7630 bb_defs if *split_p, instead preinitialize it to NULL.
7632 PR tree-optimization/85156
7633 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7634 evaluating the argument multiple times.
7636 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7638 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7640 (_mm_cvtpd_ps): Likewise.
7641 (_mm_cvttpd_epi32): Likewise.
7642 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7643 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7644 vector, pixel, and bool following altivec.h include.
7646 2018-04-03 Martin Sebor <msebor@redhat.com>
7648 * doc/extend.texi (Common Function Attributes): Clarify.
7649 (const attribute): Likewise.
7650 (pure attribute): Likewise.
7652 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7655 * config/i386/i386.c (ix86_expand_vector_set): Use
7656 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7658 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7660 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7661 instructions when changing rounding bits to preserve precision bits
7662 in the x87 control word.
7664 2018-04-03 Martin Liska <mliska@suse.cz>
7666 PR tree-optimization/82491
7667 * rtl.h (strip_offset_and_add): Replace += suboffset with
7668 poly_uint64 () + suboffset.
7670 2018-03-29 Martin Liska <mliska@suse.cz>
7671 Martin Jambor <mjambor@suse.cz>
7674 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7675 param_type is not an integral or pointer type.
7677 2018-04-03 Richard Biener <rguenther@suse.de>
7679 * sese.h (recompute_all_dominators): Remove.
7681 2018-04-02 Martin Sebor <msebor@redhat.com>
7683 * doc/invoke.texi (-Wrestrict): Fix typos.
7685 2018-04-02 Jim Wilson <jimw@sifive.com>
7687 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7688 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7689 (<optab>di3, <optab>si3_extend): Likewise.
7690 (<optab>si3_mask, <optab>si3_mask_1): New.
7691 (<optab>di3_mask, <optab>di3_mask_1): New.
7692 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7693 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7694 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7696 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7698 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7701 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7703 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7704 (nds32_canonicalize_comparison): New function.
7706 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7707 Kito Cheng <kito.cheng@gmail.com>
7708 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7710 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7711 * config/nds32/constants.md (unspec_volatile_element): Add
7712 UNSPEC_VOLATILE_RELAX_GROUP.
7713 * config/nds32/nds32-relax-opt.c: New file.
7714 * config/nds32/nds32-predicates.c
7715 (nds32_symbol_load_store_p): New function.
7716 * config/nds32/nds32-protos.h
7717 (nds32_symbol_load_store_p): Declare function.
7718 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7719 * config/nds32/nds32.c
7720 (nds32_register_pass): New function to register pass.
7721 (nds32_register_passes): New function to register passes.
7722 * config/nds32/nds32.md (relax_group): New pattern.
7723 * config/nds32/nds32.opt (mrelax-hint): New option.
7724 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7726 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7728 * config/nds32/t-nds32: Modify files dependency.
7730 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7732 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7733 (PROFILE_HOOK): Define its implementation.
7735 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7737 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7738 type and 32-bit size.
7740 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7743 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7744 (V_128_256): New mode iterator.
7745 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7746 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7747 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7749 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7750 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7752 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7755 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7756 NaN inputs correctly.
7758 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7761 * config/rs6000/vsx.md (??r): New mode attribute.
7762 (*vsx_mov<mode>_64bit): Use it.
7763 (*vsx_mov<mode>_32bit): Likewise.
7765 2018-03-30 Martin Sebor <msebor@redhat.com>
7767 PR tree-optimization/84818
7768 * builtins.c (check_access): Use warning_n.
7770 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7773 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7775 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7778 2018-03-30 Julia Koval <julia.koval@intel.com>
7781 * x86-tune.def (movx, partial_reg_dependency): Enable for
7784 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7787 * lra-constraints.c (process_alt_operands): Move setting
7788 this_alternative_matches below.
7790 2018-03-29 Martin Liska <mliska@suse.cz>
7793 * doc/invoke.texi: Document how LTO works with debug info.
7794 Describe auto-load support of binutils. Mention 'x86-64'
7795 as valid option value of -march option.
7797 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7799 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7802 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7803 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7804 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7807 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7810 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7811 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7812 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7813 for RS6000_BTM_POWERPC64.
7814 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7815 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7816 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7821 2018-03-28 Carl Love <cel@us.ibm.com>
7824 2017-09-27 Carl Love <cel@us.ibm.com>
7826 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7827 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7828 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7831 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7833 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7834 instead of __vector bool.
7835 (_mm_max_pu8): Likewise.
7836 (_mm_min_pi16): Likewise.
7838 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7841 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7842 (DIVWEUO): Likewise.
7844 (DIVDEUO): Likewise.
7845 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7846 DIVWEUO and DIVDEUO.
7847 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7848 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7849 (div_extend): Likewise.
7850 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7852 (__builtin_divweuo): Likewise.
7853 (__builtin_divdeo): Likewise.
7854 (__builtin_divdeuo): Likewise.
7856 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7859 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7860 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7862 PR tree-optimization/82004
7863 * gimple-match-head.c (optimize_pow_to_exp): New function.
7864 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7865 Don't fold to exp if optimize_pow_to_exp is false.
7867 2018-03-28 Martin Liska <mliska@suse.cz>
7870 * calls.c (initialize_argument_information): Fix trailing space.
7871 * common.opt: Fix typo and provide better explanation for
7872 -fsanitize-coverage option.
7873 * config/i386/i386.opt: Fix typo.
7875 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7876 Martin Liska <mliska@suse.cz>
7879 * gimplify.c (asan_poison_variable): Don't do the check for
7880 gimplify_omp_ctxp here.
7881 (gimplify_decl_expr): Do it here.
7882 (gimplify_target_expr): Likewise.
7884 2018-03-28 Martin Liska <mliska@suse.cz>
7887 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7888 chkp_type_bounds_count if MPX is not enabled.
7890 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7892 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7894 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7897 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7898 function to create the function decl for complex long double
7899 multiply and divide for -mabi=ieeelongdouble.
7900 (init_float128_ieee): Call it.
7902 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7905 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7906 -fcf-protection=branch -mibt.
7907 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7909 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7912 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7913 arm_word_relocations.
7915 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7918 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7919 extern array declarations.
7921 2018-03-27 Richard Biener <rguenther@suse.de>
7924 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7925 explicit single_use checks.
7927 2018-03-27 Richard Biener <rguenther@suse.de>
7929 PR tree-optimization/85082
7930 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7933 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7935 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7936 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7937 Turn on fasynchronous-unwind-tables and funwind-tables.
7939 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7942 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7943 (*bmi_blsr_<mode>_ccz): Ditto.
7945 2018-03-26 Tom de Vries <tom@codesourcery.com>
7947 PR tree-optimization/85063
7948 * omp-general.c (offloading_function_p): New function. Factor out
7950 * omp-offload.c (pass_omp_target_link::gate): ... here.
7951 * omp-general.h (offloading_function_p): Declare.
7952 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7953 with attribute omp declare target for offloading functions.
7955 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7957 PR tree-optimization/84005
7958 * tree-data-ref.h (get_base_for_alignment): Declare.
7959 * tree-data-ref.c (get_base_for_alignment_1): New function.
7960 (get_base_for_alignment): Likewise.
7961 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7962 get_base_for_alignment to find a suitable base object, instead
7963 of always using drb->base_address.
7965 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7968 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7969 known size by default.
7971 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7974 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7975 and non BLKmode operands.
7977 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7980 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7981 Clean up attributes.
7983 2018-03-23 Richard Biener <rguenther@suse.de>
7986 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7987 we are going to emit early debug for LTO.
7989 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7992 * function.c (match_asm_constraints_1): Don't optimize if input
7993 doesn't satisfy general_operand predicate for output's mode.
7996 * alias.c (write_dependence_p): Don't require for x_canonicalized
7997 non-VOIDmode if x has VOIDmode.
8000 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8001 just don't try to optimize it rather than assert it never happens.
8003 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8005 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8006 macro expansions for definition of ST_INTERNAL_<mode> and
8007 LD_INTERNAL_<mode> builtins.
8008 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8010 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8012 (altivec_expand_st_builtin): Likewise.
8013 (altivec_expand_builtin): Remove calls to deleted functions.
8014 (rs6000_address_for_altivec): Delete this function.
8015 * config/rs6000/vector.md: Remove expands for
8016 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8018 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8021 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8022 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8023 re-computing once computed.
8024 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8025 (arm_init_machine_status): Initialize
8026 machine->static_chain_stack_bytes.
8028 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8031 * doc/extend.texi: Add four new prototypes for vec_ld.
8032 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8033 definitions for more logical presentation.
8034 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8035 entries for V1TI variants of __builtin_altivec_ld builtin.
8036 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8037 handling of V1TI variant of LVX icode pattern.
8038 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8039 (rs6000_gimple_fold_builtin): Likewise.
8040 (altivec_init_builtins): Add code to define
8041 __builtin_altivec_lvx_v1ti function.
8043 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8046 * function.c (match_asm_constraints_1): Don't do the optimization
8047 if input isn't a REG, SUBREG, MEM or constant.
8049 2018-03-22 Tom de Vries <tom@codesourcery.com>
8051 PR tree-optimization/84956
8052 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8053 bb_has_abnormal_pred.
8055 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8058 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8059 DECL_INITIAL (decl) to decl at the end.
8060 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8063 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8065 * doc/extend.texi (__builtin_tgmath): Document when complex
8066 integer types are treated as _Complex _Float64.
8068 2018-03-21 Tom de Vries <tom@codesourcery.com>
8070 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8072 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8074 PR tree-optimization/84960
8075 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8076 if it is ENTRY block, move them into single succ of ENTRY in that case.
8078 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8080 PR tree-optimization/84811
8081 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8082 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8083 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8084 (known_subrange_p): Update accordingly. Cast each value involved
8085 in the size comparison, rather than casting the result of the
8088 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8090 PR tree-optimization/84982
8091 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8092 by flipping the least significant bit rather than all bits from
8093 bitpos to bitpos + bitsize - 1.
8095 2018-03-21 Nathan Sidwell <nathan@acm.org>
8097 * doc/extend.texi (Deprecated Features): Remove mention of
8098 long-deleted deprecations.
8100 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8103 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8104 * configure: Regenerate.
8106 2018-03-21 Tom de Vries <tom@codesourcery.com>
8108 PR tree-optimization/83126
8109 * tree-parloops.c (num_phis): New function.
8110 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8112 2018-03-21 Nathan Sidwell <nathan@acm.org>
8114 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8115 mention anon-struct/union members and trailing attributes.
8117 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8119 PR tree-optimization/84969
8120 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8121 builtin memset partitions if they set different rhs values.
8123 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8125 PR rtl-optimization/84989
8126 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8127 VEC_DUPLICATE with scalar result mode.
8129 2018-03-21 Martin Liska <mliska@suse.cz>
8132 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8133 not intended return statement.
8135 2018-03-21 Martin Liska <mliska@suse.cz>
8138 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8139 (chkp_find_bound_slots_1): Limit number of iterations.
8141 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8144 * Minor grammar fixes for x86 options.
8146 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8149 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8150 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8152 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8155 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8156 (altivec_lvx_<mode>_1op): Likewise.
8157 (altivec_stvx_<mode>_2op): Likewise.
8158 (altivec_stvx_<mode>_1op): Likewise.
8159 (altivec_lvx_<VM2:mode>): New define_expand.
8160 (altivec_stvx_<VM2:mode>): Likewise.
8161 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8162 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8163 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8164 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8165 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8166 (rs6000_gen_lvx): Likewise.
8167 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8168 (altivec_expand_stv_builtin): Likewise.
8169 (altivec_expand_builtin): Likewise.
8170 * config/rs6000/vector.md: Likewise.
8172 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8175 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8178 2018-03-20 Richard Biener <rguenther@suse.de>
8181 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8182 sign-conversions as zero, fall back to standard scalar_stmt
8185 2018-03-20 Martin Liska <mliska@suse.cz>
8188 * predict.c (rebuild_frequencies): Handle case when we have
8189 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8191 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8194 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8195 flag_section_anchors.
8196 * varasm.c (use_blocks_for_decl_p): Remove hack for
8197 dw2_force_const_mem.
8200 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8202 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8203 be created, use lowpart_subreg of operands[0] rather than operands[0]
8205 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8206 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8207 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8208 and n constraint instead of aarch64_shift_imm_di and Usd.
8209 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8210 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8212 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8215 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8216 to favor GPR over NEON registers.
8217 (<shift>di3_neon): Likewise.
8219 2018-03-20 Tom de Vries <tom@codesourcery.com>
8222 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8223 (nvptx_process_pars): Emit bar.sync asap and alap.
8225 2018-03-20 Tom de Vries <tom@codesourcery.com>
8228 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8229 seen_label if seen_label is already set.
8231 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8234 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8235 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8236 Use 1U instead of 1. Formatting fixes.
8239 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8240 instead of TREE_TYPE (s1) for the return value.
8242 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8244 PR tree-optimization/84946
8245 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8246 bitsize + bitsize in poly_uint64 rather than poly_int64.
8249 * dwarf2asm.c: Include fold-const.c.
8250 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8251 of decl rather than decl itself.
8253 PR rtl-optimization/84643
8254 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8256 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8259 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8260 calling assemble_variable.
8262 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8265 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8266 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8268 2018-03-19 Jim Wilson <jimw@sifive.com>
8271 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8272 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8273 (riscv_first_stack_step): Likewise.
8274 (riscv_option_override): Use STACK_BOUNDARY instead of
8276 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8278 (BIGGEST_ALIGNMENT): Set to 128.
8279 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8280 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8283 2018-03-19 Richard Biener <rguenther@suse.de>
8285 PR tree-optimization/84933
8286 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8287 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8289 2018-03-19 Richard Biener <rguenther@suse.de>
8291 PR tree-optimization/84859
8292 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8293 (cond_if_else_store_replacement): Perform sinking operation on
8294 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8295 Generalize what a BB with a single eligible store is.
8297 2018-03-19 Richard Biener <rguenther@suse.de>
8299 PR tree-optimization/84929
8300 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8301 chrec_is_positive against non-chrec arg.
8303 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8306 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8308 2018-03-18 Martin Liska <mliska@suse.cz>
8310 PR rtl-optimization/84635
8311 * regrename.c (build_def_use): Use matches_mode only when
8314 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8316 PR tree-optimization/84913
8317 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8318 vectorize chains of COND_EXPRs.
8320 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8322 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8324 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8326 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8328 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8330 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8332 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8333 Kito Cheng <kito.cheng@gmail.com>
8335 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8336 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8337 (nds32_adjust_reg_alloc_order): New function.
8338 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8340 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8342 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8343 nds32_print_operand, nds32_print_operand_address): Use
8344 HOST_WIDE_INT_PRINT_DEC instead.
8346 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8348 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8350 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8353 * config/i386/i386.c (initial_ix86_tune_features,
8354 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8356 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8357 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8358 rather than 1u << ix86_tune. Formatting fix.
8359 (ix86_option_override_internal): Change ix86_arch_mask from
8360 unsigned int to unsigned HOST_WIDE_INT, initialize to
8361 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8362 (ix86_function_specific_restore): Likewise.
8364 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8367 * postreload.c (reload_combine_recognize_pattern): Perform
8368 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8369 truncate_int_for_mode the result for the destination's mode.
8372 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8373 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8375 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8378 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8381 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8382 regno_allocno_class_array and sorted_pseudos.
8383 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8384 insns where regno is used.
8386 2018-03-16 Martin Liska <mliska@suse.cz>
8389 * multiple_target.c (create_dispatcher_calls): Redirect
8390 reference in the symbol table.
8392 2018-03-16 Martin Liska <mliska@suse.cz>
8395 * multiple_target.c (create_dispatcher_calls): Redirect also
8398 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8402 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8403 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8404 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8406 2018-03-16 Julia Koval <julia.koval@intel.com>
8408 * doc/invoke.texi (Skylake Server): Add CLWB.
8409 Cannonlake): Remove CLWB.
8411 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8413 PR tree-optimization/84841
8414 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8416 (FLOAT_ONE_CONST_TYPE): Define.
8417 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8418 (sort_by_operand_rank): Put entries with higher constant_type last
8419 rather than first to match comments.
8421 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8423 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8426 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8429 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8430 check and use address of target always.
8432 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8435 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8436 (indirect_thunk_bnd_needed): Likewise.
8437 (indirect_thunks_used): Likewise.
8438 (indirect_thunks_bnd_used): Likewise.
8439 (indirect_return_needed): New.
8440 (indirect_return_bnd_needed): Likewise.
8441 (output_indirect_thunk_function): Add a bool argument for
8443 (output_indirect_thunk_function): Don't generate alias for
8444 function return thunk.
8445 (ix86_code_end): Call output_indirect_thunk_function to generate
8446 function return thunks.
8447 (ix86_output_function_return): Set indirect_return_bnd_needed
8448 and indirect_return_needed instead of indirect_thunk_bnd_needed
8449 and indirect_thunk_needed.
8451 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8453 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8454 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8455 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8457 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8458 Paul Hua <paul.hua.gm@gmail.com>
8461 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8463 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8465 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8466 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8469 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8472 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8473 instead of GET_MODE_SIZE when comparing Units.
8475 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8478 * varasm.c (hash_section): Return an unchangeble hash value
8479 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8480 Return !aarch64_can_use_per_function_literal_pools_p ().
8482 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8485 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8486 rather than address of cmode as last argument to prepare_cmp_insn.
8488 2018-03-15 Julia Koval <julia.koval@intel.com>
8490 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8491 F_AVX512VNNI, F_AVX512BITALG): New.
8493 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8496 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8497 insn for floating-point loads and stores.
8499 2018-03-14 Carl Love <cel@us.ibm.com>
8501 * config/rs6000/rs6000-c.c: Add macro definitions for
8502 ALTIVEC_BUILTIN_VEC_PERMXOR.
8503 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8504 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8505 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8506 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8508 * config/doc/extend.texi: Add prototypes for vec_permxor.
8510 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8513 * diagnostic-show-locus.c (class layout_point): Convert m_line
8514 from int to linenum_type.
8515 (line_span::comparator): Use linenum "compare" function when
8516 comparing line numbers.
8517 (test_line_span): New function.
8518 (layout_range::contains_point): Convert param "row" from int to
8520 (layout_range::intersects_line_p): Likewise.
8521 (layout::will_show_line_p): Likewise.
8522 (layout::print_source_line): Likewise.
8523 (layout::should_print_annotation_line_p): Likewise.
8524 (layout::print_annotation_line): Likewise.
8525 (layout::print_leading_fixits): Likewise.
8526 (layout::annotation_line_showed_range_p): Likewise.
8527 (struct line_corrections): Likewise for field m_row.
8528 (line_corrections::line_corrections): Likewise for param "row".
8529 (layout::print_trailing_fixits): Likewise.
8530 (layout::get_state_at_point): Likewise.
8531 (layout::get_x_bound_for_row): Likewise.
8532 (layout::print_line): Likewise.
8533 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8534 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8535 * input.c (selftest::test_linenum_comparisons): New function.
8536 (selftest::input_c_tests): Call it.
8537 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8538 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8539 * selftest.h (ASSERT_GT): New macro.
8540 (ASSERT_GT_AT): New macro.
8541 (ASSERT_LT): New macro.
8542 (ASSERT_LT_AT): New macro.
8544 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8546 PR rtl-optimization/84780
8547 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8549 2018-03-14 Martin Liska <mliska@suse.cz>
8551 * tree.c (record_node_allocation_statistics): Use
8552 get_stats_node_kind.
8553 (get_stats_node_kind): New function extracted from
8554 record_node_allocation_statistics.
8555 (free_node): Use get_stats_node_kind.
8557 2018-03-14 Richard Biener <rguenther@suse.de>
8559 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8560 that the value-set of ANTIC_IN doesn't grow.
8563 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8565 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8566 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8568 2018-03-14 Julia Koval <julia.koval@intel.com>
8570 * config.gcc (icelake-client, icelake-server): New.
8572 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8573 (initial_ix86_arch_features): Ditto.
8574 (PTA_SKYLAKE): Add SGX.
8575 (PTA_ICELAKE): Remove.
8576 (PTA_ICELAKE_CLIENT): New.
8577 (PTA_ICELAKE_SERVER): New.
8578 (ix86_option_override_internal): Split up icelake on icelake client and
8580 (get_builtin_code_for_version): Ditto.
8581 (fold_builtin_cpu): Ditto.
8582 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8583 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8584 * config/i386/i386.h (processor_type): Ditto.
8585 * doc/invoke.texi: Ditto.
8587 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8590 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8591 INTEGER_CST offset, add it together with bitpos / 8 and
8592 sign extend based on POINTER_SIZE.
8596 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8599 * config/i386/constraints.md (Yc): New register constraint.
8600 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8601 Use Yc constraint for alternative 2 of operand 0. Remove
8602 preferred_for_speed attribute.
8604 2018-03-14 Richard Biener <rguenther@suse.de>
8606 PR tree-optimization/84830
8607 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8608 with the old one to avoid oscillations.
8610 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8613 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8615 (assign_by_spills): Return a flag of reload assignment failure.
8616 Do not process the reload assignment failures. Do not spill other
8617 reload pseudos if they has the same reg class. Update n if
8619 (lra_assign): Add a return arg. Set up from the result of
8620 assign_by_spills call.
8621 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8622 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8623 usage_insns if it is not NULL.
8624 (spill_hard_reg_in_range): New function.
8625 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8626 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8627 function prototypes.
8628 (lra_assign): Change prototype.
8629 * lra.c (lra): Add code to deal with fails by splitting hard reg
8632 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8634 * config/riscv/riscv.opt (mrelax): New option.
8635 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8636 "norelax" when riscv_mrelax is disabled.
8637 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8639 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8642 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8643 reassociation for int modes.
8645 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8647 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8648 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8650 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8651 * config/aarch64/aarch64-sve.md
8652 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8653 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8654 (*extend<mode><Vwide>2): Rename to...
8655 (aarch64_sve_extend<mode><Vwide>2): ...this.
8656 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8657 renaming the old pattern to...
8658 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8660 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8661 define_expand, renaming the old pattern to...
8662 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8663 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8664 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8665 account when deciding which SVE instruction the optab should use.
8666 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8668 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8670 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8671 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8672 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8673 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8674 (tlsdesc_small_<mode>): Turn a define_expand and use
8675 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8676 (tlsdesc_small_advsimd_<mode>): ...this.
8677 (tlsdesc_small_sve_<mode>): New pattern.
8679 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8681 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8682 (UNSPEC_UMUL_HIGHPART): New constants.
8683 (MUL_HIGHPART): New int iteraor.
8684 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8685 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8687 (*<su>mul<mode>3_highpart): New define_insn.
8689 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8692 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8695 2018-03-13 Martin Liska <mliska@suse.cz>
8698 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8700 (sem_item_optimizer::~sem_item_optimizer): Release it.
8701 (sem_item_optimizer::merge_classes): Register variable aliases.
8702 (sem_item_optimizer::fixup_pt_set): New function.
8703 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8704 * ipa-icf.h: Declare new variables and functions.
8706 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8709 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8710 integer_pow2p@2 and test integer_pow2p in condition.
8711 (A < 0 ? C : 0): Similarly for @1.
8714 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8715 characters starting at p contain '\0' character, don't look beyond
8719 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8720 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8723 * reg-stack.c (change_stack): Change update_end var from int to
8724 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8725 also call set_block_for_insn on the newly added insns and rescan.
8728 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8729 on the last operand.
8732 * tree.c (stabilize_reference_1): Return save_expr (e) for
8733 STATEMENT_LIST even if it doesn't have side-effects.
8735 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8737 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8739 2018-03-12 Renlin Li <renlin.li@arm.com>
8741 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8742 aarch64_output_scalar_simd_mov_immediate.
8744 2018-03-12 Martin Sebor <msebor@redhat.com>
8746 PR tree-optimization/83456
8747 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8748 for perfectly overlapping calls to memcpy.
8749 (gimple_fold_builtin_memory_chk): Same.
8750 (gimple_fold_builtin_strcpy): Handle no-warning.
8751 (gimple_fold_builtin_stxcpy_chk): Same.
8752 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8754 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8756 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8757 parameter. Use it for SFmode.
8758 (rs6000_function_arg_advance_1): Adjust.
8759 (rs6000_function_arg): Adjust.
8760 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8762 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8764 PR rtl-optimization/84169
8765 PR rtl-optimization/84780
8766 * combine.c (can_combine_p): Check for a 2-insn combination whether
8767 the destination register is used between the two insns, too.
8769 2018-03-12 Richard Biener <rguenther@suse.de>
8771 PR tree-optimization/84803
8772 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8773 for refs DR analysis didn't process.
8775 2018-03-12 Richard Biener <rguenther@suse.de>
8777 PR tree-optimization/84777
8778 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8779 force-vectorize loops ignore whether we are optimizing for size.
8781 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8783 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8784 (TARGET_MD_ASM_ADJUST): Define.
8786 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8787 Kito Cheng <kito.cheng@gmail.com>
8788 Chung-Ju Wu <jasonwucj@gmail.com>
8790 * config/nds32/nds32.c (nds32_compute_stack_frame,
8791 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8792 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8793 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8794 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8795 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8796 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8797 * config/nds32/nds32.md (prologue, epilogue): Use macro
8798 NDS32_V3PUSH_AVAILABLE_P to do checking.
8800 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8803 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8804 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8805 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8806 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8807 addition of most attributes on !orig_type_die or the attribute not
8808 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8810 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8811 Chung-Ju Wu <jasonwucj@gmail.com>
8813 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8815 * config/nds32/nds32.opt (mvh): New option.
8817 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8818 Chung-Ju Wu <jasonwucj@gmail.com>
8820 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8822 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8823 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8826 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8827 Chung-Ju Wu <jasonwucj@gmail.com>
8829 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8831 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8832 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8834 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8835 Kito Cheng <kito.cheng@gmail.com>
8836 Chung-Ju Wu <jasonwucj@gmail.com>
8838 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8839 UNSPEC_FFMISM and UNSPEC_FLMISM.
8840 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8841 for ffb, ffmism and flmism.
8842 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8843 (unspec_ffmism): Ditto.
8844 (unspec_flmism): Ditto.
8845 (nds32_expand_builtin_impl): Check if string extension is available.
8846 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8847 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8849 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8852 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8855 * lra-assigns.c (assign_by_spills): Return a flag of reload
8856 assignment failure. Do not process the reload assignment
8857 failures. Do not spill other reload pseudos if they has the same
8859 (lra_assign): Add a return arg. Set up from the result of
8860 assign_by_spills call.
8861 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8862 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8863 usage_insns if it is not NULL.
8864 (spill_hard_reg_in_range): New function.
8865 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8866 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8867 function prototypes.
8868 (lra_assign): Change prototype.
8869 * lra.c (lra): Add code to deal with fails by splitting hard reg
8872 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8875 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8877 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8880 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8881 (dw_val_node): Add val_symbolic_view.
8882 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8883 (symview_upper_bound): New.
8884 (new_line_info_table): Initialize symviews_since_reset.
8885 (dwarf2out_source_line): Count symviews_since_reset and set
8886 symview_upper_bound.
8887 (dw_val_equal_p): Handle symview.
8888 (add_AT_symview): New.
8889 (print_dw_val): Handle symview.
8890 (attr_checksum, attr_checksum_ordered): Likewise.
8891 (same_dw_val_p, size_of_die): Likewise.
8892 (value_format, output_die): Likewise.
8893 (add_high_low_attributes): Use add_AT_symview for entry_view.
8894 (dwarf2out_finish): Reset symview_upper_bound, clear
8897 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8900 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8901 Add strict argument and use it.
8902 (rs6000_split_multireg_move): Update for new strict argument.
8903 (mem_operand_gpr): Disallow all non-offsettable addresses.
8904 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8906 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8909 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8910 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8911 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8914 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8915 decl, use remap_type if we want to use the type.
8917 2018-03-09 Martin Sebor <msebor@redhat.com>
8919 PR tree-optimization/84526
8920 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8922 (builtin_access::generic_overlap): Be prepared to handle non-array
8925 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8927 PR rtl-optimization/84682
8928 * lra-constraints.c (process_address_1): Check is_address flag
8929 for address constraints.
8930 (process_alt_operands): Likewise.
8931 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8932 preprocess_constraints.
8933 * recog.h (preprocess_constraints): Add oploc parameter.
8935 * recog.c (preprocess_constraints): Test address_operand for
8936 CT_ADDRESS constraints.
8938 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8941 * lra-assigns.c (assign_by_spills): Return a flag of reload
8942 assignment failure. Do not process the reload assignment
8943 failures. Do not spill other reload pseudos if they has the same
8945 (lra_assign): Add a return arg. Set up from the result of
8946 assign_by_spills call.
8947 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8948 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8949 usage_insns if it is not NULL.
8950 (spill_hard_reg_in_range): New function.
8951 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8952 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8953 function prototypes.
8954 (lra_assign): Change prototype.
8955 * lra.c (lra): Add code to deal with fails by splitting hard reg
8958 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8961 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8962 Accept complain bool parameter. Only emit errors if it is true.
8963 (arm_parse_cpu_option_name): Likewise.
8964 (arm_target_thumb_only): Adjust callers of the above.
8965 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8966 prototype to take a default true bool parameter.
8967 (arm_parse_arch_option_name): Likewise.
8969 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8970 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8974 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8975 * configure: Regenerate.
8976 * configure.ac ("linker --version-script option"): New.
8977 ("linker soname option"): New.
8979 2018-03-09 Richard Biener <rguenther@suse.de>
8981 PR tree-optimization/84775
8982 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8983 immediate uses of predicate stmts and mark them modified.
8986 PR tree-optimization/84178
8987 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8989 (version_loop_for_if_conversion): Delay update_ssa call.
8990 (tree_if_conversion): Delay update_ssa until after predicate
8993 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8996 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8997 when the function accesses prior frames.
8999 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9002 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9003 gen_llsym, otherwise call maybe_gen_llsym.
9006 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9007 has ',' character inside of it.
9009 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9012 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9013 as clobbering CC_REGNUM.
9015 2018-03-08 Richard Biener <rguenther@suse.de>
9018 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9019 (follow_copies_to_constant): Do not follow SSA names registered
9022 2018-03-08 Richard Biener <rguenther@suse.de>
9024 PR tree-optimization/84178
9025 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9027 (version_loop_for_if_conversion): Delay update_ssa call.
9028 (tree_if_conversion): Delay update_ssa until after predicate
9031 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9033 PR tree-optimization/84178
9034 * tree-if-conv.c (release_bb_predicate): Remove the
9035 the assertion that the stmts have NULL use_ops.
9036 Discard the statements, asserting that they haven't
9037 yet been added to a BB.
9039 2018-03-08 Richard Biener <rguenther@suse.de>
9041 PR tree-optimization/84746
9042 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9043 (phi_translate): Pass in destination ANTIC_OUT set.
9044 (phi_translate_1): Likewise. For a simplified result lookup
9045 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9046 (phi_translate_set): Adjust.
9047 (do_pre_regular_insertion): Likewise.
9048 (do_pre_partial_partial_insertion): Likewise.
9050 2018-03-08 Martin Liska <mliska@suse.cz>
9052 PR gcov-profile/84735
9053 * doc/gcov.texi: Document usage of profile files.
9054 * gcov-io.h: Document changes in the format.
9056 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9060 * dwarf2out.c (struct dw_line_info_table): Update comments for
9062 (FORCE_RESET_NEXT_VIEW): New.
9063 (FORCE_RESETTING_VIEW_P): New.
9064 (RESETTING_VIEW_P): Check for -1 too.
9065 (ZERO_VIEW_P): Likewise.
9066 (new_line_info_table): Force-reset next view.
9067 (dwarf2out_begin_function): Likewise.
9068 (dwarf2out_source_line): Simplify zero_view_p initialization.
9069 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9070 view directly. Omit view when omitting .loc at line 0.
9072 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9074 PR tree-optimization/84740
9075 * tree-switch-conversion.c (process_switch): Call build_constructors
9076 only if info.phi_count is non-zero.
9078 PR tree-optimization/84739
9079 * tree-tailcall.c (find_tail_calls): Check call arguments against
9080 DECL_ARGUMENTS (current_function_decl) rather than
9081 DECL_ARGUMENTS (func) when checking for tail recursion.
9083 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9085 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9086 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9087 Volker Reichelt's entry and add entries for people that perform
9088 GCC fuzzy testing and report numerous bugs.
9090 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9093 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9094 readonly data in sdata, if that is disabled.
9095 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9096 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9097 -mreadonly-in-sdata option.
9099 2018-03-07 Martin Sebor <msebor@redhat.com>
9101 PR tree-optimization/84468
9102 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9103 basic block when looking for nul assignment.
9105 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9108 * except.h (output_function_exception_table): Adjust prototype.
9109 * except.c (output_function_exception_table): Remove FNNAME parameter
9110 and add SECTION parameter. Ouput one part of the table at a time.
9111 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9112 the first part of the exception table and emit unwind directives.
9113 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9114 (i386_pe_seh_cold_init): Likewise.
9115 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9116 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9117 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9118 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9119 * config/i386/winnt.c: Include except.h.
9120 (struct seh_frame_state): Add reg_offset, after_prologue and
9121 in_cold_section fields.
9122 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9123 (i386_pe_seh_cold_init): New function.
9124 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9125 to seh->in_cold_section.
9126 (seh_emit_push): Record the offset of the push.
9127 (seh_emit_save): Record the offet of the save.
9128 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9129 Test seh->after_prologue to disregard the epilogue.
9130 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9131 (i386_pe_end_cold_function): New function.
9133 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9136 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9137 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9140 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9142 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9143 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9146 * multiple_target.c: Include tree-inline.h and intl.h.
9147 (expand_target_clones): Diagnose and fail if node->definition and
9148 !tree_versionable_function_p (node->decl).
9150 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9152 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9154 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9155 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9156 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9158 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9161 * combine.c (try_combine): Use reg_or_subregno instead of handling
9162 just paradoxical SUBREGs and REGs.
9164 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9166 * config/arc/arc.c (arc_finalize_pic): Remove function.
9167 (arc_must_save_register): We use single base PIC register, remove
9168 checks to save/restore the PIC register.
9169 (arc_expand_prologue): Likewise.
9170 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9172 (arc_verify_short): Likewise.
9173 (arc_attr_type): Likewise.
9174 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9175 (walk_stores): Likewise.
9176 (arc_address_cost): Make it static.
9177 (arc_verify_short): Likewise.
9178 (branch_dest): Likewise.
9179 (arc_attr_type): Likewise.
9180 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9181 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9182 (arc_final_prescan_insn): Remove inserting the nops due to
9183 hardware hazards. It is done in reorg step.
9184 (insn_length_variant_t): Remove.
9185 (insn_length_parameters_t): Likewise.
9186 (arc_insn_length_parameters): Likewise.
9187 (arc_get_insn_variants): Likewise.
9188 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9190 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9193 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9196 PR tree-optimization/84687
9197 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9199 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9201 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9203 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9204 Rename to ppc_speculation_barrier.
9205 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9206 __builtin_ppc_speculation_barrier.
9208 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9211 * combine.c (combine_simplify_rtx): Don't try to simplify if
9212 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9215 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9217 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9218 to 32 bytes when compiling for POWER9.
9220 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9223 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9224 regparm >= 3 with no arg reg available also for calls with
9225 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9228 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9230 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9232 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9235 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9237 2018-03-05 Richard Biener <rguenther@suse.de>
9239 PR tree-optimization/84486
9240 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9241 When inserting a __builtin_assume_aligned call set the LHS
9242 SSA name alignment info accordingly.
9244 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9246 PR tree-optimization/84114
9247 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9248 Avoid reassociation of FLOAT_MODE addition.
9250 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9252 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9253 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9254 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9255 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9256 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9257 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9258 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9260 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9261 __builtin_ia32_wbinvd): New builtins.
9262 (SPECIAL_ARGS2): New.
9263 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9264 (SPECIAL_ARGS2): New.
9265 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9266 (ix86_valid_target_attribute_inner_p): Ditto.
9267 (ix86_init_mmx_sse_builtins): Add special_args2.
9268 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9269 TARGET_WBNOINVD_P): New.
9270 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9271 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9272 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9273 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9274 * config/i386/pconfigintrin.h: New file.
9275 * config/i386/wbnoinvdintrin.h: Ditto.
9276 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9278 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9280 2018-03-05 Richard Biener <rguenther@suse.de>
9282 PR tree-optimization/84670
9283 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9285 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9286 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9287 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9288 doesn't grow if all successors (recursively) were visited at least
9291 2018-03-05 Richard Biener <rguenther@suse.de>
9293 PR tree-optimization/84650
9294 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9295 if executed in the loop pipeline.
9297 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9299 * doc/configfiles.texi (Configuration Files): Move info about
9300 conditionalizing $target-protos.h to...
9301 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9302 differs from $target-protos.h.
9304 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9305 Chung-Ju Wu <jasonwucj@gmail.com>
9307 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9308 * config/nds32/nds32-multiple.md (setmemsi): Define.
9309 * config/nds32/nds32-memory-manipulation.c
9310 (nds32_gen_dup_4_byte_to_word_value): New.
9311 (emit_setmem_word_loop): New.
9312 (emit_setmem_byte_loop): New.
9313 (nds32_expand_setmem_loop): New.
9314 (nds32_expand_setmem_loop_v3m): New.
9315 (nds32_expand_setmem_unroll): New.
9316 (nds32_expand_setmem): New.
9318 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9319 Chung-Ju Wu <jasonwucj@gmail.com>
9321 * config/nds32/nds32-memory-manipulation.c
9322 (nds32_emit_load_store): New.
9323 (nds32_emit_post_inc_load_store): New.
9324 (nds32_emit_mem_move): New.
9325 (nds32_emit_mem_move_block): New.
9326 (nds32_expand_movmemsi_loop_unknown_size): New.
9327 (nds32_expand_movmemsi_loop_known_size): New.
9328 (nds32_expand_movmemsi_loop): New.
9329 (nds32_expand_movmemsi_unroll): New.
9330 (nds32_expand_movmemqi): Rename ...
9331 (nds32_expand_movmemsi): ... to this.
9332 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9333 (movmemsi): ... to this.
9334 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9335 (nds32_expand_movmemsi): ... to this.
9337 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9338 Monk Chiang <sh.chiang04@gmail.com>
9339 Chung-Ju Wu <jasonwucj@gmail.com>
9341 * config/nds32/nds32-protos.h
9342 (nds32_expand_load_multiple): New arguments.
9343 (nds32_expand_store_multiple): Ditto.
9344 (nds32_valid_multiple_load_store): Rename ...
9345 (nds32_valid_multiple_load_store_p): ... to this.
9346 * config/nds32/nds32-memory-manipulation.c
9347 (nds32_expand_load_multiple): Refine implementation.
9348 (nds32_expand_store_multiple): Ditto.
9349 * config/nds32/nds32-multiple.md
9350 (load_multiple): Update nds32_expand_load_multiple interface.
9351 (store_multiple): Update nds32_expand_store_multiple interface.
9352 * config/nds32/nds32-predicates.c
9353 (nds32_valid_multiple_load_store): Rename ...
9354 (nds32_valid_multiple_load_store_p): ... to this and refine
9356 * config/nds32/predicates.md
9357 (nds32_load_multiple_and_update_address_operation): New predicate.
9358 (nds32_store_multiple_and_update_address_operation): New predicate.
9360 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9361 Chung-Ju Wu <jasonwucj@gmail.com>
9363 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9364 (combo): New attribute.
9365 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9367 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9369 * config/nds32/nds32.opt: Change -mcmodel= default value.
9371 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9372 Monk Chiang <sh.chiang04@gmail.com>
9373 Chung-Ju Wu <jasonwucj@gmail.com>
9375 * config/nds32/constants.md (unspec_element): New enum.
9376 * config/nds32/constraints.md (Umw): New constraint.
9377 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9378 * config/nds32/nds32-intrinsic.md: Likewise.
9379 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9380 (nds32_valid_smw_lwm_base_p): New.
9381 (nds32_output_smw_single_word): New.
9382 (nds32_output_lmw_single_word): New.
9383 (nds32_expand_unaligned_load): New.
9384 (nds32_expand_unaligned_store): New.
9385 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9386 (nds32_output_smw_single_word): Declare.
9387 (nds32_output_lmw_single_word): Declare.
9388 (nds32_expand_unaligned_load): Declare.
9389 (nds32_expand_unaligned_store): Declare.
9390 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9391 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9392 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9393 NDS32_BUILTIN_UASTORE_DW.
9394 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9397 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9398 Kito Cheng <kito.cheng@gmail.com>
9399 Chung-Ju Wu <jasonwucj@gmail.com>
9401 * config/nds32/nds32-intrinsic.c
9402 (nds32_expand_builtin_null_ftype_reg): Delete.
9403 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9404 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9405 (nds32_read_argument): New.
9406 (nds32_legitimize_target): Ditto.
9407 (nds32_legitimize_argument): Ditto.
9408 (nds32_check_constant_argument): Ditto.
9409 (nds32_expand_unop_builtin): Ditto.
9410 (nds32_expand_unopimm_builtin): Ditto.
9411 (nds32_expand_binop_builtin): Ditto.
9412 (nds32_builtin_decl_impl): Ditto.
9413 (builtin_description): Ditto.
9414 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9415 (nds32_init_builtins_impl): Ditto.
9416 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9417 (nds32_builtin_decl): New.
9418 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9419 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9421 2018-03-02 Jeff Law <law@redhat.com>
9423 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9424 (redundant_insn, fill_simple_delay_slots): Likewise.
9425 (fill_slots_from_thread): Likewise.
9426 * resource.c (mark_referenced_resources): Likewise.
9427 (mark_set_resources, find_dead_or_set_registers): Likewise.
9429 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9431 * substring-locations.h (format_warning_va): Formatting fix for
9433 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9435 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9436 * substring-locations.c: Include intl.h.
9437 (format_warning_va): Turned into small wrapper around
9438 format_warning_n_va, renamed to ...
9439 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9440 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9442 (format_warning_at_substring_n): New function.
9443 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9444 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9445 format_warning_at_substring with just a shorter name instead of
9446 const function pointer.
9447 (fmtwarn_n): New function.
9448 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9449 appropriate, get rid of all the fmtstr temporaries, move conditionals
9450 with G_() wrapped string literals directly into fmtwarn arguments,
9451 cast dir.len to (int), formatting fixes.
9453 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9455 * doc/invoke.texi: Remove "Cilk Plus" references.
9457 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9458 Richard Biener <rguenther@suse.de>
9461 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9462 for error or warning attributes if CALL_FROM_THUNK_P is set.
9465 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9468 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9469 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9472 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9473 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9475 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9476 instead of -1U in last predictors element's probability member.
9478 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9481 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9482 arguments if they are comparable.
9484 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9486 PR tree-optimization/84634
9487 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9488 masks and masked_loop_p with a single loop_masks, making sure it's
9489 null for bb vectorization.
9491 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9493 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9494 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9495 loop->force_vectorize to check whether there is no alias.
9497 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9500 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9502 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9504 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9505 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9506 instead of a loop around prev_real_insn.
9507 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9511 * config/i386/i386.c (ix86_print_operand): Use conditional
9512 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9515 2018-03-02 Richard Biener <rguenther@suse.de>
9517 PR tree-optimization/84427
9518 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9519 (bitmap_set_subtract_values): Rewrite to handle multiple
9522 (prune_clobbered_mems): Likewise.
9523 (phi_translate): Take edge instead of pred/phiblock.
9524 (phi_translate_1): Likewise.
9525 (phi_translate_set): Likewise. Insert all translated
9526 exprs for a value into the set, keeping possibly multiple
9527 expressions per value.
9528 (compute_antic_aux): Adjust for phi_translate changes.
9529 When intersecting union the expressions and prune those
9530 not in the final value set, keeping possibly multiple
9531 expressions per value. Do not use value-insertion
9532 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9533 all expressions. Add verification that the value-sets
9534 only shrink during iteration.
9535 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9536 (do_pre_regular_insertion): Likewise.
9537 (do_pre_partial_partial_insertion): Likewise.
9539 2018-03-02 Richard Biener <rguenther@suse.de>
9542 * config/darwin.c (saved_debug_info_level): New static global.
9543 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9544 (darwin_asm_lto_end): Restore debug info generation settings.
9546 2018-03-01 Martin Liska <mliska@suse.cz>
9549 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9552 2018-03-01 Richard Biener <rguenther@suse.de>
9555 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9556 type annotation with LTO when debug was disabled at compile-time.
9558 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9560 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9562 (mips_final_postscan_insn): Likewise.
9564 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9566 PR rtl-optimization/84528
9567 * alias.c (init_alias_target): Add commentary.
9568 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9569 a unique base value if the frame pointer is not eliminated
9570 to the stack pointer.
9572 2018-03-01 Tom de Vries <tom@codesourcery.com>
9574 PR rtl-optimization/83327
9575 * lra-int.h (hard_regs_spilled_into): Declare.
9576 * lra.c (hard_regs_spilled_into): Define.
9577 (init_reg_info): Init hard_regs_spilled_into.
9578 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9579 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9580 (process_bb_lives): Handle hard_regs_spilled_into.
9581 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9582 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9584 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9586 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9587 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9588 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9589 * config/rs6000/aix72.h: New file.
9591 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9593 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9594 instead of warning_at with conditional singular and plural messages
9598 * stor-layout.c (update_alignment_for_field): For
9599 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9600 && !DECL_PACKED (field), do the alignment update, just use
9601 only desired_align instead of MAX (type_align, desired_align)
9603 (place_field): Don't do known_align < desired_align handling
9604 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9605 is non-NULL, instead do it after rli->prev_field handling and
9606 only if not within a bitfield word. For DECL_PACKED (field)
9607 use type_align of BITS_PER_UNIT.
9609 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9611 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9612 superfluous parentheses and trailing spaces.
9614 2018-02-28 Richard Biener <rguenther@suse.de>
9616 PR tree-optimization/84584
9617 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9618 SCoPs with fake exit edge.
9620 2018-02-28 Martin Liska <mliska@suse.cz>
9623 * timevar.c (timer::print): Fix format to properly print 100%
9626 2018-02-28 Richard Biener <rguenther@suse.de>
9629 * genmatch.c (capture_info::walk_match): Do not mark
9630 captured expressions without operands as expr_p given
9631 they act more like predicates and should be subject to
9632 "lost tail" side-effect preserving.
9634 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9636 PR rtl-optimization/81611
9637 * auto-inc-dec.c (attempt_change): Move dead note from
9638 mem_insn if it's the next use of regno
9639 (find_address): Take address use of reg holding
9640 non-incremented value. Add parm to limit search to the named
9642 (merge_in_block): Attempt to use a mem insn that is the next
9643 use of the original regno.
9645 2018-02-27 Martin Sebor <msebor@redhat.com>
9648 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9649 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9651 2018-02-27 Martin Sebor <msebor@redhat.com>
9653 PR translation/84207
9654 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9655 n argument to unsigned HOST_WIDE_INT.
9656 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9657 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9658 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9659 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9661 2018-02-27 Richard Biener <rguenther@suse.de>
9663 PR tree-optimization/84512
9664 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9665 Do not use the estimate returned from record_stmt_cost for
9666 the scalar iteration cost but sum properly using add_stmt_cost.
9668 2018-02-27 Richard Biener <rguenther@suse.de>
9670 PR tree-optimization/84466
9671 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9672 Adjust last change to less strictly validate use operands.
9674 2018-02-27 Martin Liska <mliska@suse.cz>
9676 PR gcov-profile/84548
9677 * gcov.c (process_file): Allow partial overlap and consider it
9678 also as group functions.
9679 (output_lines): Properly calculate range of lines for a group.
9681 2018-02-27 Martin Liska <mliska@suse.cz>
9683 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9684 'ggc' suffixes. Change first column width.
9685 (timer::print): Fix formatting of the column.
9687 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9689 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9690 preserve inline entry blocks for the sake of debug inline
9691 entry point markers alone.
9692 (remove_unused_locals): Suggest in comments a better place to
9693 force the preservation of inline entry blocks that are
9694 otherwise unused, but do not preserve them.
9696 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9698 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9700 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9703 * config/i386/constraints.md (Bs): Replace
9704 ix86_indirect_branch_register with
9705 TARGET_INDIRECT_BRANCH_REGISTER.
9707 * config/i386/i386.md (indirect_jump): Likewise.
9708 (tablejump): Likewise.
9709 (*sibcall_memory): Likewise.
9710 (*sibcall_value_memory): Likewise.
9711 Peepholes of indirect call and jump via memory: Likewise.
9712 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9713 (*sibcall_value_GOT_32): Likewise.
9714 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9715 (GOT_memory_operand): Likewise.
9716 (call_insn_operand): Likewise.
9717 (sibcall_insn_operand): Likewise.
9718 (GOT32_symbol_operand): Likewise.
9719 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9721 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9723 PR rtl-optimization/83496
9724 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9725 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9726 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9727 redundant insn, if any.
9728 (relax_delay_slots): Likewise.
9729 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9731 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9733 PR tree-optimization/83965
9734 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9735 that grouped statements are part of a reduction chain. Return
9736 true if the statement is not marked as a reduction itself but
9738 (vect_recog_dot_prod_pattern): Don't check whether the statement
9739 is part of a group here.
9740 (vect_recog_sad_pattern): Likewise.
9741 (vect_recog_widen_sum_pattern): Likewise.
9743 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9746 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9748 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9751 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9753 (ix86_output_indirect_function_return): New prototype.
9754 (ix86_split_simple_return_pop_internal): Likewise.
9755 * config/i386/i386.c (indirect_return_via_cx): New.
9756 (indirect_return_via_cx_bnd): Likewise.
9757 (indirect_thunk_name): Handle return va CX_REG.
9758 (output_indirect_thunk_function): Create alias for
9759 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9760 (ix86_output_indirect_jmp): Remove the bool argument.
9761 (ix86_output_indirect_function_return): New function.
9762 (ix86_split_simple_return_pop_internal): Likewise.
9763 * config/i386/i386.md (*indirect_jump): Don't pass false
9764 to ix86_output_indirect_jmp.
9765 (*tablejump_1): Likewise.
9766 (simple_return_pop_internal): Change it to define_insn_and_split.
9767 Call ix86_split_simple_return_pop_internal to split it for
9769 (simple_return_indirect_internal): Call
9770 ix86_output_indirect_function_return instead of
9771 ix86_output_indirect_jmp.
9773 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9776 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9777 memset and value initialization afterwards.
9779 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9781 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9783 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9786 * common/config/aarch64/aarch64-common.c
9787 (aarch_option_optimization_table[]): Switch
9788 off fomit-frame-pointer
9790 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9791 Chung-Ju Wu <jasonwucj@gmail.com>
9793 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9795 (store_multiple): Ditto.
9797 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9799 * config.gcc: Add --with-cpu support for nds32 target.
9800 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9801 * config/nds32/nds32.opt: Add -mcpu= option.
9803 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9805 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9806 isel=yes): Warn for these deprecated options.
9808 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9810 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9811 ISA_2_5_MASKS_EMBEDDED.
9813 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9815 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9816 p->max as pointers rather than using iterative_hash_expr.
9818 2018-02-23 Carl Love <cel@us.ibm.com>
9820 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9821 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9823 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9824 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9825 P8V_BUILTIN_VEC_VUNSIGNED2.
9827 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9830 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9831 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9833 * lra-constraints.c (curr_insn_transform): Set up
9834 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9836 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9837 * lra-eliminations.c (spill_pseudos): Ditto.
9838 (process_insn_for_elimination): Ditto.
9839 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9840 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9841 LRA_NON_CLOBBERED_ALT.
9843 2018-02-22 Martin Sebor <msebor@redhat.com>
9845 PR tree-optimization/84480
9846 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9847 to maybe_diag_stxncpy_trunc. Call it.
9848 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9849 from gimple_fold_builtin_strcpy. Print inlining stack.
9850 (handle_builtin_stxncpy): Print inlining stack.
9851 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9853 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9856 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9857 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9858 and -fcheck-pointer-bounds are used together.
9859 (indirect_thunk_prefix): New enum.
9860 (indirect_thunk_need_prefix): New function.
9861 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9862 "_nt" instead of "_bnd" for NOTRACK prefix.
9863 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9864 (output_indirect_thunk_function): Likewise.
9866 (ix86_code_end): Update output_indirect_thunk_function calls.
9867 (ix86_output_indirect_branch_via_reg): Replace
9868 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9869 (ix86_output_indirect_branch_via_push): Likewise.
9870 (ix86_output_function_return): Likewise.
9871 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9872 incompatible with -fcf-protection=branch and
9873 -fcheck-pointer-bounds.
9875 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9878 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9879 Change gcc_assert call to output_operand_lossage.
9881 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9883 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9885 2018-02-22 DJ Delorie <dj@redhat.com>
9886 Sebastian Perta <sebastian.perta@renesas.com>
9887 Oleg Endo <olegendo@gcc.gnu.org>
9889 * config/rx/rx.c (rx_rtx_costs): New function.
9890 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9892 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9894 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9896 2018-02-22 Martin Liska <mliska@suse.cz>
9899 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9900 Add "native" as a possible value.
9902 2018-02-22 Martin Liska <mliska@suse.cz>
9905 * config/i386/i386.c (ix86_option_override_internal):
9906 Add "native" as a possible value for -march and -mtune.
9908 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9911 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9912 to all type variants.
9914 PR tree-optimization/84503
9915 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9916 width as info->bitpos + info->bitsize - start.
9917 (merged_store_group::merge_overlapping): Simplify width computation.
9918 (check_no_overlap): New function.
9919 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9920 start + width and last_order of the group, fail if check_no_overlap
9922 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9923 to group if check_no_overlap fails.
9925 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9927 * config/rs6000/altivec.md: Delete contraint arguments to
9928 define_expand, define_split, and define_peephole2, and in
9929 define_insn_and_split if always unused.
9930 * config/rs6000/darwin.md: Ditto.
9931 * config/rs6000/dfp.md: Ditto.
9932 * config/rs6000/rs6000.md: Ditto.
9933 * config/rs6000/sync.md: Ditto.
9934 * config/rs6000/vector.md: Ditto.
9935 * config/rs6000/vsx.md: Ditto.
9937 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9939 * config/rs6000/altivec.md: Write output control strings as braced
9940 blocks instead of double-quoted strings.
9941 * config/rs6000/darwin.md: Ditto.
9942 * config/rs6000/rs6000.md: Ditto.
9943 * config/rs6000/vector.md: Ditto.
9944 * config/rs6000/vsx.md: Ditto.
9946 2018-02-21 Jason Merrill <jason@redhat.com>
9948 PR c++/84314 - ICE with templates and fastcall attribute.
9949 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9951 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9953 * ipa-cp.c (determine_versionability): Fix comment typos.
9955 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9958 * ipa-cp.c (determine_versionability): Do not version functions caling
9961 2018-02-21 Martin Liska <mliska@suse.cz>
9964 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9965 Add "native" as a possible value.
9966 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9967 the macro when native cpu detection is available.
9969 2018-02-21 Martin Liska <mliska@suse.cz>
9972 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9973 Add "native" as a possible value.
9974 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9975 when native cpu detection is available.
9977 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9978 Martin Sebor <msebor@redhat.com>
9980 PR tree-optimization/84478
9981 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9983 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9984 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9985 support which is conservatively correct, for 2 only stay conservative
9986 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9987 argument to the 2 argument get_range_strlen, adjust 6 arg
9988 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9990 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9991 (gimple_fold_builtin_strlen): Pass true as last argument to
9994 2018-02-20 Martin Sebor <msebor@redhat.com>
9997 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9998 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9999 (builtin_memref::builtin_memref): Factor out parts into
10000 set_base_and_offset and call it.
10002 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10004 PR middle-end/84406
10005 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10006 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10007 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10008 search at the associated MODE_INT.
10010 2018-02-20 Jeff Law <law@redhat.com>
10012 PR middle-end/82123
10013 PR tree-optimization/81592
10014 PR middle-end/79257
10015 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10016 for range data rather than using global data.
10017 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10018 range data rather than using global data.
10019 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10020 pass it to children as needed.
10021 (struct directive::fmtresult): Similarly.
10022 (struct directive::set_width): Similarly.
10023 (struct directive::set_precision): Similarly.
10024 (format_integer, format_directive, parse_directive): Similarly.
10025 (format_none): Accept unnamed vr_values parameter.
10026 (format_percent, format_floating, format_character): Similarly.
10027 (format_string, format_plain): Similarly.
10028 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10029 the EVRP range analyzer for range data rather than using global data.
10030 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10031 gimple-ssa-evrp-analyze.h
10032 (class sprintf_dom_walker): Add after_dom_children member function.
10033 Add evrp_range_analyzer member.
10034 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10035 range analyzer as needed.
10036 (sprintf_dom_walker::after_dom_children): New member function.
10037 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10039 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10040 (evrp_range_analyzer::pop_to_marker): Likewise.
10042 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10044 PR tree-optimization/84419
10045 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10046 with the required type if its current type is compatible but
10049 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10051 PR middle-end/82004
10052 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10053 after vectorization.
10055 2018-02-20 Martin Liska <mliska@suse.cz>
10058 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10059 possible values if we don't have a hint.
10061 2018-02-20 Martin Liska <mliska@suse.cz>
10065 * final.c (shorten_branches): Build align_tab array with one
10067 * opts.c (finish_options): Add alignment option limit check.
10068 (MAX_CODE_ALIGN): Likewise.
10069 (MAX_CODE_ALIGN_VALUE): Likewise.
10070 * doc/invoke.texi: Document maximum allowed option value for
10071 all -falign-* options.
10073 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10076 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10077 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10078 * var-tracking.c (emit_note_insn_var_location): Remove all references
10079 to NOTE_INSN_CALL_ARG_LOCATION.
10080 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10081 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10082 Use copy_rtx_if_shared.
10083 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10084 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10085 (dwarf2out_var_location): Remove handling of
10086 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10088 * final.c (final_scan_insn): Remove all references to
10089 NOTE_INSN_CALL_ARG_LOCATION.
10090 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10091 before dumping final insns.
10092 * except.c (emit_note_eh_region_end): Remove all references to
10093 NOTE_INSN_CALL_ARG_LOCATION.
10094 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10095 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10096 * config/arc/arc.c (hwloop_optimize): Likewise.
10097 * config/arm/arm.c (create_fix_barrier): Likewise.
10098 * config/s390/s390.c (s390_chunkify_start): Likewise.
10099 * config/sh/sh.c (find_barrier): Likewise.
10100 * config/i386/i386.c (rest_of_insert_endbranch,
10101 ix86_seh_fixup_eh_fallthru): Likewise.
10102 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10103 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10104 * config/frv/frv.c (frv_function_prologue): Likewise.
10105 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10107 (note_outside_basic_block_p): Remove all references to
10108 NOTE_INSN_CALL_ARG_LOCATION.
10109 * gengtype.c (adjust_field_rtx_def): Likewise.
10110 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10112 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10113 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10116 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10119 PR tree-optimization/84452
10120 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10121 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10124 2018-02-19 Martin Liska <mliska@suse.cz>
10127 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10129 2018-02-19 Martin Liska <mliska@suse.cz>
10130 Richard Sandiford <richard.sandiford@linaro.org>
10132 PR tree-optimization/82491
10133 * gimple-fold.c (get_base_constructor): Make earlier bail out
10136 2018-02-19 Carl Love <cel@us.ibm.com>
10138 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10139 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10141 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10142 P8V_BUILTIN_VEC_NEG.
10144 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10146 * config/rl78/rl78.md (movdf): New define expand.
10148 2018-02-19 Martin Liska <mliska@suse.cz>
10151 * doc/invoke.texi: Fix typo.
10152 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10154 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10156 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10157 handle rs6000_single_float and rs6000_double_float specially for
10160 2018-02-16 Jeff Law <law@redhat.com>
10162 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10163 (pop_regs): Use it.
10165 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10168 * ipa-inline.c (inline_small_functions): Fix a typo.
10170 2018-02-16 Nathan Sidwell <nathan@acm.org>
10172 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10174 2018-02-16 Carl Love <cel@us.ibm.com>
10176 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10177 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10178 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10179 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10180 expansion to P8V_BUILTIN_VEC_FLOAT2.
10182 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10184 PR rtl-optimization/70023
10185 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10186 src_regno into account.
10188 2018-02-16 Carl Love <cel@us.ibm.com>
10190 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10191 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10192 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10193 * config/rs6000/rs6000.c: Remove case statements for
10194 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10195 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10196 and P9V_BUILTIN_VEC_VINSERT4B.
10197 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10198 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10199 * config/rs6000/vsx.md:
10200 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10203 2018-02-16 Carl Love <cel@us.ibm.com>
10205 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10207 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10209 * config/rs6000/rs6000-c.c: Add the definitions for
10210 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10211 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10212 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10213 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10214 definition for insert4b and define insn *insert3b_internal.
10215 * doc/extend.texi: Add documentation for vec_extract4b.
10217 2018-02-16 Nathan Sidwell <nathan@acm.org>
10219 * doc/extend.texi (Backwards Compatibility): Mention friend
10220 injection. Note for-scope is deprecated.
10221 * doc/invoke.texi (-ffriend-injection): Deprecate.
10223 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10225 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10226 that moved to I2, also allow destinations that are a paradoxical
10227 subreg (instead of a normal reg).
10229 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10232 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10235 2018-02-16 Richard Biener <rguenther@suse.de>
10237 PR tree-optimization/84037
10238 PR tree-optimization/84016
10240 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10241 Adjust vec_construct for the fact we need additional higher latency
10242 128bit inserts for AVX256 and AVX512 vector builds.
10243 (ix86_add_stmt_cost): Scale vector construction cost for
10246 2018-02-16 Richard Biener <rguenther@suse.de>
10248 PR tree-optimization/84417
10249 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10250 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10251 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10253 2018-02-16 Martin Liska <mliska@suse.cz>
10256 * internal-fn.def (ASAN_CHECK): Set proper flags.
10257 (ASAN_MARK): Likewise.
10259 2018-02-16 Julia Koval <julia.koval@intel.com>
10261 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10262 from PTA_CANNONLAKE.
10264 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10267 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10268 Use ++iter rather than iter++ for std::list iterators.
10269 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10270 defer deleting them until all nodes in the forest are processed. Do
10271 free even leaf nodes. Change to_process into auto_vec.
10274 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10275 * vec.h (vec_default_construct): Use memset instead of placement new
10276 if BROKEN_VALUE_INITIALIZATION is defined.
10277 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10278 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10281 PR rtl-optimization/83723
10282 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10283 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10284 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10286 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10288 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10290 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10292 PR rtl-optimization/81443
10293 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10294 from inner REGs to paradoxical SUBREGs.
10296 2018-02-16 Richard Biener <rguenther@suse.de>
10298 PR tree-optimization/84399
10299 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10300 For operands we can analyze at their definition make sure we can
10301 analyze them at each use as well.
10303 2018-02-16 Richard Biener <rguenther@suse.de>
10305 PR tree-optimization/84190
10306 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10307 volatile accesses if the decl isn't volatile.
10309 2018-02-15 Jason Merrill <jason@redhat.com>
10311 PR c++/84314 - ICE with templates and fastcall attribute.
10312 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10313 TYPE_CANONICAL on an existing type.
10315 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10317 PR tree-optimization/84383
10318 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10319 dstoff nor call operand_equal_p if dstbase is NULL.
10321 PR tree-optimization/84334
10322 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10323 also a CONSTANT_CLASS_P, punt.
10325 2018-02-14 Jim Wilson <jimw@sifive.com>
10327 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10328 first SMALL_OPERAND check. New local min_second_step. Move assert
10329 to where locals are set. Add TARGET_RVC support.
10330 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10332 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10334 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10336 2018-02-14 Martin Sebor <msebor@redhat.com>
10338 PR tree-optimization/83698
10339 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10340 arrays constrain the offset range to their bounds.
10341 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10342 (builtin_access::overlap): Avoid setting the size of overlap if it's
10344 (maybe_diag_overlap): Also consider arrays when deciding what values
10345 of offsets to include in diagnostics.
10347 2018-02-14 Martin Sebor <msebor@redhat.com>
10350 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10351 that correspond to the kind of a declaration.
10353 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10356 * config/pa/pa.md: Load address of PIC label using the linkage table
10357 if the label is nonlocal.
10359 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10361 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10362 warning message if user requests -maltivec=be.
10363 * doc/invoke.texi: Document deprecation of -maltivec=be.
10365 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10368 * config/rs6000/rs6000-c.c: Update definitions for
10369 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10370 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10372 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10375 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10376 add _get_ssp intrinsics. Remove argument from
10377 __builtin_ia32_rdssp[d|q].
10378 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10379 * config/i386/i386-builtin.def: Remove argument from
10380 __builtin_ia32_rdssp[d|q].
10381 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10382 ix86_expand_special_args_builtin for _rdssp[d|q].
10383 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10384 Clear register before usage.
10385 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10386 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10388 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10390 PR tree-optimization/84357
10391 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10392 operand 1 of an ARRAY_REF too.
10394 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10397 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10398 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10400 (set_of_reg): New struct.
10401 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10402 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10403 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10405 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10406 Split into bitclr, bitset, bitinvert patterns if appropriate.
10407 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10408 use rx_fuse_in_memory_bitop.
10409 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10410 to named insn, correct maximum insn length.
10412 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10415 * machmode.def: Define a complex mode for PARTIAL_INT.
10416 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10418 * doc/rtl.texi: Document CSPImode.
10419 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10421 (msp430_hard_regno_nregs_with_padding): Likewise.
10423 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10426 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10428 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10430 PR rtl-optimization/84169
10431 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10432 we generated a parallel as new i3 and we split that to new i2 and i3
10433 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10434 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10435 those to i2, not i1. Partially rewrite this scan code.
10437 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10440 * stor-layout.c (place_field): For variable length fields, adjust
10441 offset_align afterwards not just based on the field's alignment,
10442 but also on the size.
10444 PR middle-end/84309
10445 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10446 of exps and logs in the use_exp2 case.
10448 2018-02-13 Jeff Law <law@redhat.com>
10450 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10451 entry for "vector".
10453 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10456 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10460 * common.opt (gas-loc-support, gas-locview-support): New.
10461 (ginline-points, ginternal-reset-location-views): New.
10462 * doc/invoke.texi: Document them. Use @itemx where intended.
10463 (gvariable-location-views): Adjust.
10464 * target.def (reset_location_view): New.
10465 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10466 (TARGET_RESET_LOCATION_VIEW): New.
10467 * doc/tm.texi: Rebuilt.
10468 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10469 (dwarf2out_default_as_locview_support): New.
10470 (output_asm_line_debug_info): Use option variables.
10471 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10472 (output_loc_list): Likewise.
10473 (add_high_low_attributes): Check option variables.
10474 Don't output entry view attribute in strict mode.
10475 (gen_inlined_subroutine_die): Check option variables.
10476 (dwarf2out_inline_entry): Likewise.
10477 (init_sections_and_labels): Likewise.
10478 (dwarf2out_early_finish): Likewise.
10479 (maybe_reset_location_view): New, from...
10480 (dwarf2out_var_location): ... here. Call it.
10481 * debug.h (dwarf2out_default_as_loc_support): Declare.
10482 (dwarf2out_default_as_locview_support): Declare.
10483 * hooks.c (hook_int_rtx_insn_0): New.
10484 * hooks.h (hook_int_rtx_insn_0): Declare.
10485 * toplev.c (process_options): Take -gas-loc-support and
10486 -gas-locview-support from dwarf2out. Enable
10487 -gvariable-location-views by default only with locview
10488 assembler support. Enable -ginternal-reset-location-views by
10489 default only if the target defines the corresponding hook.
10490 Enable -ginline-points by default if location views are
10491 enabled; force it disabled if statement frontiers are
10493 * tree-inline.c (expand_call_inline): Check option variables.
10494 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10496 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10498 PR tree-optimization/84321
10499 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10500 handling. Also check whether the anti-range contains any values
10501 that satisfy the mask; switch to a VR_RANGE if not.
10503 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10506 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10508 2018-02-13 Martin Jambor <mjambor@suse.cz>
10511 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10512 of call statements, also set location of a load to a temporary.
10514 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10516 * config/rl78/rl78.c (add_vector_labels): New function.
10517 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10518 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10519 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10520 which checks that no arguments are passed.
10521 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10522 * doc/extend.texi: Documentation for the new attribute.
10524 2018-02-13 Andreas Schwab <schwab@suse.de>
10526 * config/riscv/linux.h (CPP_SPEC): Define.
10528 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10531 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10532 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10533 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10534 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10535 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10536 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10537 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10538 temporarily for AES and PCLMUL builtins.
10540 PR tree-optimization/84339
10541 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10542 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10545 PR middle-end/84309
10546 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10547 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10548 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10550 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10552 * omp-simd-clone.h: New file.
10553 * omp-simd-clone.c: Include omp-simd-clone.h.
10554 (expand_simd_clones): No longer static.
10555 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10556 cgraph.h and omp-simd-clone.h.
10557 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10558 (vect_recog_widen_shift_pattern): Formatting fix.
10559 (vect_pattern_recog_1): Don't check optab for calls.
10562 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10563 operands[2] into a REG before using gen_lowpart on it.
10565 2018-02-12 Jeff Law <law@redhat.com>
10568 * config/sh/sh.c (find_barrier): Consider a sibling call
10571 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10572 successfully back substituting a reg.
10574 2018-02-12 Richard Biener <rguenther@suse.de>
10576 PR tree-optimization/84037
10577 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10578 parameter, move visited init to caller.
10579 (vect_slp_analyze_operations): Separate cost from validity
10580 check, initialize visited once for all instances.
10581 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10583 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10585 (vect_model_promotion_demotion_cost): Likewise.
10586 (vectorizable_bswap): Guard cost modeling with !slp_node
10587 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10589 (vectorizable_call): Likewise.
10590 (vectorizable_conversion): Likewise.
10591 (vectorizable_assignment): Likewise.
10592 (vectorizable_shift): Likewise.
10593 (vectorizable_operation): Likewise.
10594 (vectorizable_store): Likewise.
10595 (vectorizable_load): Likewise.
10596 (vectorizable_condition): Likewise.
10597 (vectorizable_comparison): Likewise.
10599 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10602 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10603 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10604 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10606 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10608 PR middle-end/83665
10609 * params.def (inline-min-speedup): Increase from 8 to 15.
10610 (max-inline-insns-auto): Decrease from 40 to 30.
10611 * ipa-split.c (consider_split): Add some buffer for function to
10612 be considered inlining candidate.
10613 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10616 2018-02-12 Richard Biener <rguenther@suse.de>
10618 PR tree-optimization/84037
10619 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10620 matched stmts if we cannot swap the non-matched ones.
10622 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10624 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10625 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10626 _mm_maskz_scalef_round_ss): New intrinsics.
10627 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10628 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10629 __builtin_ia32_scalefss_round): Remove.
10630 (__builtin_ia32_scalefsd_mask_round,
10631 __builtin_ia32_scalefss_mask_round): New intrinsics.
10632 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10633 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10634 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10635 "<round_constraint>")): Changed to ...
10636 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10637 "<round_scalar_constraint>")): ... this.
10638 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10639 %0, %1, %2<round_op3>}"): Changed to ...
10640 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10641 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10642 %2<round_scalar_mask_op3>}"): ... this.
10643 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10645 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10647 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10648 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10649 (_mm_maskz_sqrt_round_ss): New intrinsics.
10650 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10651 (__builtin_ia32_sqrtsd_mask_round)
10652 (__builtin_ia32_sqrtss_mask_round): New builtins.
10653 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10654 (__builtin_ia32_sqrtss_round): Remove.
10655 (__builtin_ia32_sqrtsd_mask_round)
10656 (__builtin_ia32_sqrtss_mask_round): New builtins.
10657 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10658 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10659 ((match_operand:VF_128 1 "vector_operand"
10660 "xBm,<round_constraint>")): Changed to ...
10661 ((match_operand:VF_128 1 "vector_operand"
10662 "xBm,<round_scalar_constraint>")): ... this.
10663 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10664 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10665 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10666 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10667 %<iptr>1<round_scalar_mask_op3>}): ... this.
10668 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10669 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10671 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10674 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10675 Cast vec_cmpeq result to correct type.
10676 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10677 Cast vec_cmpgt result to correct type.
10679 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10681 * final.c (final_scan_insn_1): Renamed from...
10682 (final_scan_insn): ... this. New wrapper, to recover
10683 seen from the outermost call in recursive ones.
10684 * config/sparc/sparc.c (output_return): Drop seen from call.
10685 (output_sibcall): Likewise.
10686 * config/visium/visium.c (output_branch): Likewise.
10688 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10690 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10693 2018-02-10 Alan Modra <amodra@gmail.com>
10696 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10697 Specify LR as an input.
10699 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10702 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10703 remove_member_access_dummy_vars): New functions.
10704 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10705 lower_omp_1, execute_lower_omp): Use them.
10707 PR rtl-optimization/84308
10708 * shrink-wrap.c (spread_components): Release todo vector.
10710 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10712 PR rtl-optimization/57193
10713 * ira-color.c (struct allocno_color_data): Add member
10714 conflict_allocno_hard_prefs.
10715 (update_conflict_allocno_hard_prefs): New.
10716 (bucket_allocno_compare_func): Add a preference based on
10717 conflict_allocno_hard_prefs.
10718 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10719 (color_allocnos): Remove a dead code. Initiate
10720 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10722 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10725 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10726 constraint from =wa to wa. Avoid a subreg on the output operand,
10727 instead use a pseudo and subreg it in a move.
10728 (p9_xxbrd_<mode>): Changed to ...
10729 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10730 (p9_xxbrd_v2df): New expander.
10731 (p9_xxbrw_<mode>): Changed to ...
10732 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10733 (p9_xxbrw_v4sf): New expander.
10735 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10737 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10739 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10742 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10743 multiply in 32-bit mode.
10744 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10745 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10748 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10750 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10751 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10752 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10753 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10755 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10758 * dwarf2out.c (is_trivial_indirect_ref): New function.
10759 (dwarf2out_late_global_decl): Do not generate a location
10760 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10761 and that are not defined in the current unit.
10763 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10765 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10766 instead of a libcall for UNORDERED.
10768 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10771 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10772 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10774 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10777 * config/s390/s390.c (s390_set_current_function): Invoke
10778 s390_indirect_branch_settings also if fndecl didn't change.
10780 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10782 * config/rs6000/rs6000.md (blockage): Set length to zero.
10784 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10786 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10788 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10791 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10792 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10796 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10797 PARALLEL incoming that failed vt_get_decl_and_offset check.
10799 PR middle-end/84237
10800 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10801 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10803 (get_variable_section): For decls in named .bss* sections pass true as
10804 second argument to bss_initializer_p.
10806 2018-02-09 Marek Polacek <polacek@redhat.com>
10807 Jakub Jelinek <jakub@redhat.com>
10810 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10811 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10812 Sync some changes from cxx_fold_indirect_ref.
10814 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10816 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10818 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10819 (BLOCK_INLINE_ENTRY_LABEL): New.
10820 (dwarf2out_var_location): Disregard inline entry markers.
10821 (inline_entry_data): New struct.
10822 (inline_entry_data_hasher): New hashtable type.
10823 (inline_entry_data_hasher::hash): New.
10824 (inline_entry_data_hasher::equal): New.
10825 (inline_entry_data_table): New variable.
10826 (add_high_low_attributes): Add DW_AT_entry_pc and
10827 DW_AT_GNU_entry_view attributes if a pending entry is found
10828 in inline_entry_data_table. Add old entry_pc attribute only
10829 if debug nonbinding markers are disabled.
10830 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10831 markers are enabled.
10832 (block_within_block_p, dwarf2out_inline_entry): New.
10833 (dwarf2out_finish): Check that no entries remained in
10834 inline_entry_data_table.
10835 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10836 (final_scan_insn, notice_source_line): Likewise.
10837 (rest_of_clean_state): Skip inline entry markers.
10838 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10840 * gimple.c (gimple_build_debug_inline_entry): New.
10841 * gimple.h (enum gimple_debug_subcode): Add
10842 GIMPLE_DEBUG_INLINE_ENTRY.
10843 (gimple_build_debug_inline_entry): Declare.
10844 (gimple_debug_inline_entry_p): New.
10845 (gimple_debug_nonbind_marker_p): Adjust.
10846 * insn-notes.def (INLINE_ENTRY): New.
10847 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10848 inline entry marker notes.
10849 (print_insn): Likewise.
10850 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10851 (INSN_DEBUG_MARKER_KIND): Likewise.
10852 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10853 * tree-inline.c (expand_call_inline): Build and insert
10854 debug_inline_entry stmt.
10855 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10856 inline entry blocks early, if nonbind markers are enabled.
10857 (dump_scope_block): Dump fragment info.
10858 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10859 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10860 (gimple_build_debug_inline_entry): New.
10861 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10862 Enable/disable inline entry points too.
10863 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10864 (DEBUG_INSN): Describe inline entry markers.
10866 * common.opt (gvariable-location-views): New.
10867 (gvariable-location-views=incompat5): New.
10868 * config.in: Rebuilt.
10869 * configure: Rebuilt.
10870 * configure.ac: Test assembler for view support.
10871 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10872 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10873 * dwarf2out.c (var_loc_view): New typedef.
10874 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10875 (dwarf2out_locviews_in_attribute): New.
10876 (dwarf2out_locviews_in_loclist): New.
10877 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10878 (enum dw_line_info_opcode): Add LI_adv_address.
10879 (struct dw_line_info_table): Add view.
10880 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10881 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10882 (zero_view_p): New variable.
10883 (ZERO_VIEW_P): New macro.
10884 (output_asm_line_debug_info): New.
10885 (struct var_loc_node): Add view.
10886 (add_AT_view_list, AT_loc_list): New.
10887 (add_var_loc_to_decl): Add view param. Test it against last.
10888 (new_loc_list): Add view params. Record them.
10889 (AT_loc_list_ptr): Handle loc and view lists.
10890 (view_list_to_loc_list_val_node): New.
10891 (print_dw_val): Handle dw_val_class_view_list.
10892 (size_of_die): Likewise.
10893 (value_format): Likewise.
10894 (loc_list_has_views): New.
10895 (gen_llsym): Set vl_symbol too.
10896 (maybe_gen_llsym, skip_loc_list_entry): New.
10897 (dwarf2out_maybe_output_loclist_view_pair): New.
10898 (output_loc_list): Output view list or entries too.
10899 (output_view_list_offset): New.
10900 (output_die): Handle dw_val_class_view_list.
10901 (output_dwarf_version): New.
10902 (output_compilation_unit_header): Use it.
10903 (output_skeleton_debug_sections): Likewise.
10904 (output_rnglists, output_line_info): Likewise.
10905 (output_pubnames, output_aranges): Update version comments.
10906 (output_one_line_info_table): Output view numbers in asm comments.
10907 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10908 Call maybe_gen_llsym.
10909 (loc_list_from_tree_1): Adjust.
10910 (add_AT_location_description): Create view list attribute if
10911 needed, check it's absent otherwise.
10912 (convert_cfa_to_fb_loc_list): Adjust.
10913 (maybe_emit_file): Call output_asm_line_debug_info for test.
10914 (dwarf2out_var_location): Reset views as needed. Precompute
10915 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10916 attribute. Set view.
10917 (new_line_info_table): Reset next view.
10918 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10919 (dwarf2out_source_line): Likewise. Output view resets and labels to
10920 the assembler, or select appropriate line info opcodes.
10921 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10922 (optimize_string_length): Catch it. Adjust.
10923 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10924 dw_val_class_view_list, and remove it if no longer needed.
10925 (hash_loc_list): Hash view numbers.
10926 (loc_list_hasher::equal): Compare them.
10927 (optimize_location_lists): Check whether a view list symbol is
10928 needed, and whether the locview attribute is present, and
10929 whether they match. Remove the locview attribute if no longer
10931 (index_location_lists): Call skip_loc_list_entry for test.
10932 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10933 Use output_dwarf_version.
10934 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10935 (struct dw_val_node): Add val_view_list.
10936 * final.c (SEEN_NEXT_VIEW): New.
10937 (set_next_view_needed): New.
10938 (clear_next_view_needed): New.
10939 (maybe_output_next_view): New.
10940 (final_start_function): Rename to...
10941 (final_start_function_1): ... this. Take pointer to FIRST,
10942 add SEEN parameter. Emit param bindings in the initial view.
10943 (final_start_function): Reintroduce SEEN-less interface.
10944 (final): Rename to...
10945 (final_1): ... this. Take SEEN parameter. Output final pending
10946 next view at the end.
10947 (final): Reintroduce seen-less interface.
10948 (final_scan_insn): Output pending next view before switching
10949 sections or ending a block. Mark the next view as needed when
10950 outputting variable locations. Notify debug backend of section
10951 changes, and of location view changes.
10952 (rest_of_handle_final): Adjust.
10953 * toplev.c (process_options): Autodetect value for debug variable
10954 location views option. Warn on incompat5 without -gdwarf-5.
10955 * doc/invoke.texi (gvariable-location-views): New.
10956 (gvariable-location-views=incompat5): New.
10957 (gno-variable-location-views): New.
10959 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10961 PR tree-optimization/84136
10962 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10963 that the result of find_edge is non-NULL.
10965 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10968 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10969 storing integer register in SImode. Fix cost of 256 and 512
10970 byte aligned SSE register store.
10972 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10974 * config/i386/i386.c (ix86_multiplication_cost): Fix
10975 multiplication cost for TARGET_AVX512DQ.
10977 2018-02-08 Marek Polacek <polacek@redhat.com>
10979 PR tree-optimization/84238
10980 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10983 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10985 PR tree-optimization/84265
10986 * tree-vect-stmts.c (vectorizable_store): Don't treat
10987 VMAT_CONTIGUOUS accesses as grouped.
10988 (vectorizable_load): Likewise.
10990 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10992 PR tree-optimization/81635
10993 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10994 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10995 (test_round_for_mask): New functions.
10996 (wide_int_cc_tests): Call test_round_for_mask.
10997 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10998 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10999 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11000 range returned by get_range_info.
11002 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11005 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11006 * symtab.c: Include builtins.h
11007 (symtab_node::output_to_lto_symbol_table_p): Move here
11008 from lto-streamer-out.c:output_symbol_p.
11009 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11010 (output_symbol_p): Move all logic to symtab.c
11011 (produce_symtab): Update.
11013 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11015 * config/s390/s390-opts.h (enum indirect_branch): Define.
11016 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11017 (s390_indirect_branch_via_thunk)
11018 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11019 (enum s390_indirect_branch_type): Define.
11020 * config/s390/s390.c (struct s390_frame_layout, struct
11021 machine_function): Remove.
11022 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11023 (indirect_branch_table_label_no, indirect_branch_table_name):
11025 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11026 (enum s390_indirect_branch_option): Define.
11027 (s390_return_addr_from_memory): New function.
11028 (s390_handle_string_attribute): New function.
11029 (s390_attribute_table): Add new attribute handler.
11030 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11031 (s390_indirect_branch_via_thunk): New function.
11032 (s390_indirect_branch_via_inline_thunk): New function.
11033 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11034 sibling call optimization for non z10 compiles.
11035 (s390_emit_call): Force indirect branch target to be a single
11036 register. Add r1 clobber for non-z10 compiles.
11037 (s390_emit_epilogue): Emit return jump via return_use expander.
11038 (s390_reorg): Handle JUMP_INSNs as execute targets.
11039 (s390_option_override_internal): Perform validity checks for the
11040 new command line options.
11041 (s390_indirect_branch_attrvalue): New function.
11042 (s390_indirect_branch_settings): New function.
11043 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11044 (s390_output_indirect_thunk_function): New function.
11045 (s390_code_end): Implement target hook.
11046 (s390_case_values_threshold): Implement target hook.
11047 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11049 * config/s390/s390.h (struct s390_frame_layout)
11050 (struct machine_function): Move here from s390.c.
11051 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11052 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11053 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11054 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11055 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11056 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11057 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11058 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11059 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11060 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11061 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11062 (mnemonic attribute): Add values which aren't recognized
11064 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11065 pattern for branch conversion. Fix mnemonic attribute.
11066 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11067 indirect branch via thunk if requested.
11068 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11069 ("*indirect_jump"): Disable for branch conversion using out of
11071 ("indirect_jump_via_thunk<mode>_z10")
11072 ("indirect_jump_via_thunk<mode>")
11073 ("indirect_jump_via_inlinethunk<mode>_z10")
11074 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11075 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11076 ("casesi_jump_via_inlinethunk<mode>_z10")
11077 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11078 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11079 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11080 ("*indirect2_jump"): Disable for branch conversion.
11081 ("casesi_jump"): Turn into expander and expand patterns for branch
11083 ("return_use"): New expander.
11084 ("*return"): Emit return via thunk and rename it to ...
11085 ("*return<mode>"): ... this one.
11086 * config/s390/s390.opt: Add new options and and enum for the
11089 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11091 * lra-constraints.c (match_reload): Unconditionally use
11092 gen_lowpart_SUBREG, rather than selecting between that
11093 and equivalent gen_rtx_SUBREG code.
11095 2018-02-08 Richard Biener <rguenther@suse.de>
11097 PR tree-optimization/84233
11098 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11099 changed flag instead of boguously re-using phi_inserted.
11101 2018-02-08 Martin Jambor <mjambor@suse.cz>
11103 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11104 static local variables.
11106 2018-02-08 Richard Biener <rguenther@suse.de>
11108 PR tree-optimization/84278
11109 * tree-vect-stmts.c (vectorizable_store): When looking for
11110 smaller vector types to perform grouped strided loads/stores
11111 make sure the mode is supported by the target.
11112 (vectorizable_load): Likewise.
11114 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11116 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11117 Increase LDP/STP opportunities by adding adjacent callee-saves.
11119 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11121 PR rtl-optimization/84068
11122 PR rtl-optimization/83459
11123 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11125 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11127 PR tree-optimization/84224
11128 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11129 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11130 non-zero arguments.
11132 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11135 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11136 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11137 count, remove one USE.
11139 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11141 * doc/install.texi (Configuration): Document the
11142 --with-long-double-format={ibm,ieee} PowerPC configuration
11146 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11147 Convert from define_expand to be define_insn_and_split. Rework
11148 float/double/_Float128 conversions to QI/HI/SImode to work with
11149 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11150 conversions to QI/HImode types did a store and then a load to
11151 truncate the value. For conversions to VSX registers, don't split
11152 the insn, instead emit the code directly. Use the code iterator
11153 any_fix to combine signed and unsigned conversions.
11154 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11155 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11156 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11157 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11158 (fix_<mode>di2_hw): Likewise.
11159 (fixuns_<mode>di2_hw): Likewise.
11160 (fix_<mode>si2_hw): Likewise.
11161 (fixuns_<mode>si2_hw): Likewise.
11162 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11163 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11164 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11165 fix<uns>_trunc<SFDF:mode>si2_p8.
11166 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11167 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11168 (fix<uns>_<mode>_mem): Likewise.
11169 (fctiw<u>z_<mode>_mem): Likewise.
11170 (fix<uns>_<mode>_mem): Likewise.
11171 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11172 the register allocator from doing a direct move to the GPRs to do
11173 a store, and instead use the ISA 3.0 store byte/half-word from
11174 vector register instruction. For IEEE 128-bit floating point,
11175 also optimize stores of 32-bit ints.
11176 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11178 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11180 * genextract.c (push_pathstr_operand): New function to support
11182 (walk_rtx): Call push_pathstr_operand.
11183 (print_path): Support [a-zA-Z].
11185 2018-02-07 Richard Biener <rguenther@suse.de>
11187 PR tree-optimization/84037
11188 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11189 (cse_and_gimplify_to_preheader): Declare.
11190 (vect_get_place_in_interleaving_chain): Likewise.
11191 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11193 (_loop_vec_info::~_loop_vec_info): Delete it.
11194 (cse_and_gimplify_to_preheader): New function.
11195 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11196 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11197 (vectorizable_load): Likewise. For grouped stores always base
11198 the IV on the first element.
11199 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11200 condition before gimplifying.
11202 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11204 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11205 *DIV_EXPR and *MOD_EXPR.
11207 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11210 * config/i386/i386.c (ix86_option_override_internal): Mask out
11211 the CF_SET bit when checking -fcf-protection.
11213 2018-02-07 Tom de Vries <tom@codesourcery.com>
11216 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11219 2018-02-07 Richard Biener <rguenther@suse.de>
11221 PR tree-optimization/84204
11222 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11225 PR tree-optimization/84205
11226 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11227 special-case isl_ast_op_zdiv_r.
11229 PR tree-optimization/84223
11230 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11231 Only add conditions from within the region.
11232 (gather_bbs::after_dom_children): Adjust.
11234 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11237 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11238 * config/avr/avr.md: Only post-reload split REG-REG moves if
11239 either register is GENERAL_REG_P.
11241 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11243 PR tree-optimization/84235
11244 * tree-ssa-scopedtables.c
11245 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11246 if the subtraction is performed in floating point type where NaNs are
11247 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11248 build 1. Formatting fix.
11250 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11253 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11254 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11255 and skip it regardless of bb boundaries. Use CALL_P macro,
11256 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11257 unnecessarily, formatting fix.
11259 2018-02-06 Michael Collison <michael.collison@arm.com>
11261 * config/arm/thumb2.md:
11262 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11263 (*thumb_mov_notscc): Ditto.
11265 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11268 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11269 unsigned_fix, not "s".
11271 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11273 * configure.ac (gcc_fn_eh_frame_ro): New function.
11274 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11275 correct .eh_frame permissions.
11276 * configure: Regenerate.
11278 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11280 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11281 irrelevant options.
11283 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11285 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11286 Display warning message for -mno-speculate-indirect-jumps.
11288 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11290 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11292 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11294 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11296 PR tree-optimization/84225
11297 * tree-eh.c (find_trapping_overflow): Only call
11298 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11300 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11303 * config/i386/i386.c: Reimplement the check of possible options
11304 -mibt/-mshstk conbination. Change error messages.
11305 * doc/invoke.texi: Fix a typo: remove extra '='.
11307 2018-02-06 Marek Polacek <polacek@redhat.com>
11309 PR tree-optimization/84228
11310 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11312 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11315 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11316 emitted arch directives.
11317 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11318 __ARM_FEATURE_COPROC before changing architectures.
11320 2018-02-06 Richard Biener <rguenther@suse.de>
11322 * config/i386/i386.c (print_reg): Fix typo.
11323 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11325 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11327 * configure: Regenerate.
11329 2018-02-05 Martin Sebor <msebor@redhat.com>
11331 PR tree-optimization/83369
11332 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11335 2018-02-05 Martin Liska <mliska@suse.cz>
11337 * doc/invoke.texi: Cherry-pick upstream r323995.
11339 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11341 * ira.c (ira_init_register_move_cost): Adjust comment.
11343 2018-02-05 Martin Liska <mliska@suse.cz>
11345 PR gcov-profile/84137
11346 * doc/gcov.texi: Fix typo in documentation.
11348 2018-02-05 Martin Liska <mliska@suse.cz>
11350 PR gcov-profile/83879
11351 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11352 using dlopen functionality.
11354 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11356 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11357 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11358 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11359 _mm_maskz_range_round_ss): New intrinsics.
11360 (__builtin_ia32_rangesd128_round)
11361 (__builtin_ia32_rangess128_round): Remove.
11362 (__builtin_ia32_rangesd128_mask_round,
11363 __builtin_ia32_rangess128_mask_round): New builtins.
11364 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11365 __builtin_ia32_rangess128_round): Remove.
11366 (__builtin_ia32_rangesd128_mask_round,
11367 __builtin_ia32_rangess128_mask_round): New builtins.
11368 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11369 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11370 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11371 "<round_saeonly_constraint>")): Changed to ...
11372 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11373 "<round_saeonly_scalar_constraint>")): ... this.
11374 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11375 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11376 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11377 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11378 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11380 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11382 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11384 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11385 Remove all values except native, 8540 and 8548.
11387 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11389 * config/i386/i386.c (ix86_output_function_return): Pass
11390 INVALID_REGNUM, instead of -1, as invalid register number to
11391 indirect_thunk_name and output_indirect_thunk.
11393 2018-02-02 Julia Koval <julia.koval@intel.com>
11395 * config.gcc: Add -march=icelake.
11396 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11397 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11398 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11399 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11400 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11401 (processor_target_table): Add icelake.
11402 (ix86_option_override_internal): Handle new PTAs.
11403 (get_builtin_code_for_version): Handle icelake.
11404 (M_INTEL_COREI7_ICELAKE): New.
11405 (fold_builtin_cpu): Handle icelake.
11406 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11407 * doc/invoke.texi: Add -march=icelake.
11409 2018-02-02 Julia Koval <julia.koval@intel.com>
11411 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11412 to wide_int_bitmask.
11413 * wide-int-bitmask.h: New.
11415 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11418 * config/i386/i386.md: Replace Pmode with word_mode in
11419 builtin_setjmp_setup and builtin_longjmp to support x32.
11421 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11425 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11427 (rs6000_supported_cpu_names): New static variable.
11428 (linux_cpu_translation_table): Likewise.
11429 (elf_platform) <cpu>: Define new static variable and use it.
11430 Translate kernel AT_PLATFORM name to canonical name if needed.
11431 Error if platform name is unknown.
11433 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11436 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11438 2018-02-01 Jeff Law <law@redhat.com>
11441 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11442 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11443 the scratch if RELEASE_VIA_POP is false.
11444 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11445 If we have to save a temporary register, decrement SIZE appropriately.
11446 Pass new arguments to release_scratch_register_on_entry.
11447 (ix86_adjust_stack_and_probe): Likewise.
11448 (ix86_emit_probe_stack_range): Pass new arguments to
11449 release_scratch_register_on_entry.
11451 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11453 PR rtl-optimization/84157
11454 * combine.c (change_zero_ext): Use REG_P predicate in
11455 front of HARD_REGISTER_P predicate.
11457 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11459 * config/avr/avr.c (avr_option_override): Move disabling of
11460 -fdelete-null-pointer-checks to...
11461 * common/config/avr/avr-common.c (avr_option_optimization_table):
11464 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11466 PR tree-optimization/81635
11467 * tree-data-ref.c (split_constant_offset_1): For types that
11468 wrap on overflow, try to use range info to prove that wrapping
11471 2018-02-01 Renlin Li <renlin.li@arm.com>
11474 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11475 TAILCALL_ADDR_REGS.
11476 (aarch64_register_move_cost): Likewise.
11477 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11478 TAILCALL_ADDR_REGS.
11479 (REG_CLASS_NAMES): Likewise.
11480 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11481 TAILCALL_ADDR_REGS. Remove IP registers.
11482 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11484 2018-02-01 Richard Biener <rguenther@suse.de>
11486 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11487 for specifying RPO order and allow NULL for that.
11488 * domwalk.c (dom_walker::dom_walker): Likewise.
11489 (dom_walker::walk): Handle NULL RPO order.
11490 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11492 (rewrite_update_dom_walker): Likewise.
11493 (mark_def_dom_walker): Likewise.
11495 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11497 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11498 (aarch64_maybe_expand_sve_subreg_move): Declare.
11499 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11500 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11502 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11503 that are semantically a reverse operation.
11504 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11505 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11506 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11508 (aarch64_can_change_mode_class): For big-endian, forbid changes
11509 between two SVE modes if they have different element sizes.
11511 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11513 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11514 the TImode handling for big-endian targets.
11516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11518 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11519 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11521 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11522 Remove BSWAP handing for big-endian targets and use the form of
11523 LD1RQ appropariate for the mode.
11525 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11527 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11528 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11529 duplicated element.
11531 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11534 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11535 check for operands that need to go through aarch64_sve_reload_be.
11537 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11539 PR tree-optimization/81661
11540 PR tree-optimization/84117
11541 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11542 * tree-eh.c: Include gimplify.h.
11543 (find_trapping_overflow, replace_trapping_overflow,
11544 rewrite_to_non_trapping_overflow): New functions.
11545 * tree-vect-loop.c: Include tree-eh.h.
11546 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11547 * tree-data-ref.c: Include tree-eh.h.
11548 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11550 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11552 PR rtl-optimization/84123
11553 * combine.c (change_zero_ext): Check if hard register satisfies
11554 can_change_dest_mode before calling gen_lowpart_SUBREG.
11556 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11559 * ira.c (ira_init_register_move_cost): Remove assert.
11561 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11563 PR rtl-optimization/84071
11564 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11565 * doc/tm.texi: Regenerate.
11567 2018-01-31 Richard Biener <rguenther@suse.de>
11569 PR tree-optimization/84132
11570 * tree-data-ref.c (analyze_miv_subscript): Properly
11571 check whether evolution_function_is_affine_multivariate_p
11572 before calling gcd_of_steps_may_divide_p.
11574 2018-01-31 Julia Koval <julia.koval@intel.com>
11577 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11578 * config/i386/i386.md (rdpid_rex64) New.
11579 (rdpid): Make 32bit only.
11581 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11584 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11585 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11587 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11590 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11592 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11594 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11596 PR rtl-optimization/84071
11597 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11598 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11600 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11602 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11603 (arc_attribute_table): Add 'aux' attribute.
11604 (arc_in_small_data_p): Consider aux like variables.
11605 (arc_is_aux_reg_p): New function.
11606 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11607 (arc_get_aux_arg): New function.
11608 (prepare_move_operands): Handle aux-register access.
11609 (arc_handle_aux_attribute): New function.
11610 * doc/extend.texi (ARC Variable attributes): Add subsection.
11612 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11614 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11615 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11616 (arc_attribute_table): Add 'uncached' attribute.
11617 (arc_print_operand): Print '.di' flag for uncached memory
11619 (arc_in_small_data_p): Do not consider for small data the uncached
11621 (arc_is_uncached_mem_p): New function.
11622 * config/arc/predicates.md (compact_store_memory_operand): Check
11623 for uncached memory accesses.
11624 (nonvol_nonimm_operand): Likewise.
11625 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11627 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11630 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11631 falign-loops=): Add Optimization flag.
11633 2018-01-30 Jeff Law <law@redhat.com>
11636 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11637 INT_REGISTERS_SAVED. Check it prior to calling
11638 get_scratch_register_on_entry.
11639 (ix86_adjust_stack_and_probe): Similarly.
11640 (ix86_emit_probe_stack_range): Similarly.
11641 (ix86_expand_prologue): Corresponding changes.
11643 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11646 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11647 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11649 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11652 * lra-constraints.c (curr_insn_transform): Process AND in the
11655 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11657 PR rtl-optimization/83986
11658 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11659 dependence against last_pending_memory_flush in addition to
11660 pending_jump_insns.
11662 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11664 PR tree-optimization/81611
11665 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11668 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11671 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11674 2018-01-30 Richard Biener <rguenther@suse.de>
11675 Jakub Jelinek <jakub@redhat.com>
11677 PR tree-optimization/84111
11678 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11679 inner loops added during recursion, as they don't have up-to-date
11682 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11685 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11686 (can_inline_edge_by_limits_p): ... here.
11687 (can_early_inline_edge_p, check_callers,
11688 update_caller_keys, update_callee_keys, recursive_inlining,
11689 add_new_edges_to_heap, speculation_useful_p,
11690 inline_small_functions,
11691 inline_small_functions, flatten_function,
11692 inline_to_all_callers_1): Update.
11694 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11696 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11699 2018-01-30 Richard Biener <rguenther@suse.de>
11701 PR tree-optimization/83008
11702 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11703 invariant and constant vector uses in stmts when they need
11704 more than one stmt.
11706 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11709 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11710 * configure: Regenerate.
11712 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11714 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11716 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11717 Use gen_rtx_REG rather than gen_lowpart.
11719 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11721 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11722 rather than 0 when creating partial subregs.
11724 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11726 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11729 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11732 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11733 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11734 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11735 flags. This restores the settings used before the 2017-07-24.
11736 Turning off pre increment/decrement/modify allows IVOPTS to
11737 optimize DF/SF loops where the index is an int.
11739 2018-01-29 Richard Biener <rguenther@suse.de>
11740 Kelvin Nilsen <kelvin@gcc.gnu.org>
11743 * tree-vect-stmts.c (vectorizable_call): Don't call
11744 targetm.vectorize_builtin_md_vectorized_function if callee is
11747 2018-01-22 Carl Love <cel@us.ibm.com>
11749 * doc/extend.tex: Fix typo in second arg in
11750 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11752 2018-01-29 Richard Biener <rguenther@suse.de>
11754 PR tree-optimization/84086
11755 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11756 (flush_ssaname_freelist): When SSA names were released reset
11757 the SCEV hash table.
11759 2018-01-29 Richard Biener <rguenther@suse.de>
11761 PR tree-optimization/84057
11762 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11763 removed paths when removing edges.
11765 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11767 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11768 -mfunction-return=@var{choice}.
11770 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11772 PR diagnostic/84034
11773 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11774 Handle CR like TAB.
11775 (layout::print_source_line): Likewise.
11776 (test_get_line_width_without_trailing_whitespace): Add test cases.
11778 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11780 PR middle-end/84040
11781 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11784 2018-01-26 Jim Wilson <jimw@sifive.com>
11786 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11788 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11791 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11793 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11794 and CMP + SUB-immediate -> SUBS.
11796 2018-01-26 Martin Sebor <msebor@redhat.com>
11798 PR tree-optimization/83896
11799 * tree-ssa-strlen.c (get_string_len): Rename...
11800 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11801 Avoid assuming length is constant.
11802 (handle_char_store): Use HOST_WIDE_INT for string length.
11804 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11807 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11808 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11810 2018-01-26 Richard Biener <rguenther@suse.de>
11812 PR rtl-optimization/84003
11813 * dse.c (record_store): Only record redundant stores when
11814 the earlier store aliases at least all accesses the later one does.
11816 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11818 PR rtl-optimization/83985
11819 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11820 REG_CFA_RESTORE insns.
11821 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11822 REG_CFA_RESTORE insns here.
11825 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11826 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11828 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11830 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11831 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11832 (arc_init): Likewise.
11833 (arc_override_options): Likewise.
11834 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11836 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11838 * config/arc/arc.h (TARGET_DBNZ): Define.
11839 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11840 properly set the tune attribute.
11841 (dbnz): Use TARGET_DBNZ guard.
11842 * config/arc/arc.opt (mtune): Add core3 option.
11844 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11846 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11847 recognize new pic like addresses.
11848 (arc_delegitimize_address): Clean up.
11850 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11852 * config/arc/arc-arches.def: Option mrf16 valid for all
11854 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11855 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11856 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11857 * config/arc/arc-tables.opt: Regenerate.
11858 * config/arc/arc.c (arc_conditional_register_usage): Handle
11859 reduced register file case.
11860 (arc_file_start): Set must have build attributes.
11861 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11862 mrf16 option value.
11863 * config/arc/arc.opt (mrf16): Add new option.
11864 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11865 * config/arc/genmultilib.awk: Handle new mrf16 option.
11866 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11867 * config/arc/t-multilib: Regenerate.
11868 * doc/invoke.texi (ARC Options): Document mrf16 option.
11870 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11872 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11873 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11874 (arc_attribute_table): Add 'secure_call' attribute.
11875 (arc_print_operand): Print secure call operand.
11876 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11878 (arc_is_secure_call_p): New function. * config/arc/arc.md
11879 (call_i): Add support for sjli instruction.
11880 (call_value_i): Likewise.
11881 * config/arc/constraints.md (Csc): New constraint.
11883 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11884 John Eric Martin <John.Martin@emmicro-us.com>
11886 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11887 * config/arc/arc.c (_arc_jli_section): New struct.
11888 (arc_jli_section): New type.
11889 (rc_jli_sections): New static variable.
11890 (arc_handle_jli_attribute): New function.
11891 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11892 (arc_file_end): New function.
11893 (TARGET_ASM_FILE_END): Define.
11894 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11895 (arc_add_jli_section): New function.
11896 (jli_call_scan): Likewise.
11897 (arc_reorg): Call jli_call_scan.
11898 (arc_output_addsi): Remove 'S' from printing asm operand.
11899 (arc_is_jli_call_p): New function.
11900 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11902 (movhi_insn): Likewise.
11903 (movsi_insn): Likewise.
11904 (movsi_set_cc_insn): Likewise.
11905 (loadqi_update): Likewise.
11906 (load_zeroextendqisi_update): Likewise.
11907 (load_signextendqisi_update): Likewise.
11908 (loadhi_update): Likewise.
11909 (load_zeroextendhisi_update): Likewise.
11910 (load_signextendhisi_update): Likewise.
11911 (loadsi_update): Likewise.
11912 (loadsf_update): Likewise.
11913 (movsicc_insn): Likewise.
11914 (bset_insn): Likewise.
11915 (bxor_insn): Likewise.
11916 (bclr_insn): Likewise.
11917 (bmsk_insn): Likewise.
11918 (bicsi3_insn): Likewise.
11919 (cmpsi_cc_c_insn): Likewise.
11920 (movsi_ne): Likewise.
11921 (movsi_cond_exec): Likewise.
11922 (clrsbsi2): Likewise.
11923 (norm_f): Likewise.
11933 (call_i): Remove 'S' asm letter, add jli instruction.
11934 (call_value_i): Likewise.
11935 * config/arc/arc.op (mjli-always): New option.
11936 * config/arc/constraints.md (Cji): New constraint.
11937 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11939 (subsf3_fpx): Likewise.
11940 (mulsf3_fpx): Likewise.
11941 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11943 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11944 function attrbutes.
11945 * doc/invoke.texi (ARC): Document mjli-always option.
11947 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11949 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11950 avoid addition with 0 and use incw and decw where possible.
11952 2018-01-26 Richard Biener <rguenther@suse.de>
11954 PR tree-optimization/81082
11955 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11956 association if it requires casting to unsigned.
11957 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11958 from fold_plusminus_mult_expr to catch important cases late when
11959 range info is available.
11961 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11963 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11964 * configure.ac (hidden_linkonce): New test.
11965 * configure: Regenerate.
11966 * config.in: Regenerate.
11968 2018-01-26 Julia Koval <julia.koval@intel.com>
11970 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11971 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11972 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11973 _mm_mask_bitshuffle_epi64_mask): Fix type.
11974 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11975 USI_FTYPE_V4DI_V4DI_USI): Remove.
11976 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11977 __builtin_ia32_vpshufbitqmb256_mask,
11978 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11979 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11980 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11982 2018-01-26 Alan Modra <amodra@gmail.com>
11985 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11986 UNSPEC_VBPERMQ. Sort other unspecs.
11988 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11990 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11992 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11994 PR middle-end/83055
11995 * predict.c (drop_profile): Do not push/pop cfun; update also
11997 (handle_missing_profiles): Fix logic looking for zero profiles.
11999 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12001 PR middle-end/83977
12002 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12003 on functions with #pragma omp declare simd or functions with simd
12005 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12006 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12007 Remove trailing \n from warning_at calls.
12009 2018-01-25 Tom de Vries <tom@codesourcery.com>
12012 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12013 for neutered workers.
12015 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12018 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12019 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12021 2018-01-24 Jeff Law <law@redhat.com>
12024 * i386.c (get_probe_interval): Move to earlier point.
12025 (ix86_compute_frame_layout): If -fstack-clash-protection and
12026 the frame is larger than the probe interval, then use pushes
12027 to save registers rather than reg->mem moves.
12028 (ix86_expand_prologue): Remove conditional for int_registers_saved
12031 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12034 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12035 min/max for never referenced object.
12037 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12039 PR middle-end/83977
12040 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12042 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12043 attributes from DECL_ATTRIBUTES (decl) without affecting
12044 DECL_ATTRIBUTES (current_function_decl).
12045 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12046 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12048 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12050 PR tree-optimization/83979
12051 * fold-const.c (fold_comparison): Use constant_boolean_node
12052 instead of boolean_{true,false}_node.
12054 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12056 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12059 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12061 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12062 Simplify the clause that sets the length attribute.
12063 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12064 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12065 clause that sets the length attribute.
12066 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12068 2018-01-24 Tom de Vries <tom@codesourcery.com>
12071 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12072 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12073 Add strict parameter.
12074 (prevent_branch_around_nothing): Insert dummy insn between branch to
12075 label and label with no ptx insn inbetween.
12076 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12078 2018-01-24 Tom de Vries <tom@codesourcery.com>
12081 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12082 for neutered threads in warp.
12083 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12085 2018-01-24 Richard Biener <rguenther@suse.de>
12087 PR tree-optimization/83176
12088 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12091 2018-01-24 Richard Biener <rguenther@suse.de>
12093 PR tree-optimization/82819
12094 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12095 code generating pluses that are no-ops in the target precision.
12097 2018-01-24 Richard Biener <rguenther@suse.de>
12099 PR middle-end/84000
12100 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12102 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12104 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12105 to merge probabilities.
12106 * predict.c (probably_never_executed): Also mark as cold functions
12107 with global 0 profile and guessed local profile.
12108 * profile-count.c (profile_probability::combine_with_count): New
12110 * profile-count.h (profile_probability::operator*,
12111 profile_probability::operator*=, profile_probability::operator/,
12112 profile_probability::operator/=): Reduce precision to adjusted
12113 and set value to guessed on contradictory divisions.
12114 (profile_probability::combine_with_freq): Remove.
12115 (profile_probability::combine_wiht_count): Declare.
12116 (profile_count::force_nonzero):: Set to adjusted.
12117 (profile_count::probability_in):: Set quality to adjusted.
12118 * tree-ssa-tail-merge.c (replace_block_by): Use
12119 combine_with_count.
12121 2018-01-23 Andrew Waterman <andrew@sifive.com>
12122 Jim Wilson <jimw@sifive.com>
12124 * config/riscv/riscv.c (riscv_stack_boundary): New.
12125 (riscv_option_override): Set riscv_stack_boundary. Handle
12126 riscv_preferred_stack_boundary_arg.
12127 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12128 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12129 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12130 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12131 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12132 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12134 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12137 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12138 of struct ix86_frame.
12139 (ix86_expand_epilogue): Likewise. Add a local variable for
12140 the reg_save_offset field in struct ix86_frame.
12142 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12144 PR tree-optimization/82604
12145 * tree-loop-distribution.c (enum partition_kind): New enum item
12146 PKIND_PARTIAL_MEMSET.
12147 (partition_builtin_p): Support above new enum item.
12148 (generate_code_for_partition): Ditto.
12149 (compute_access_range): Differentiate cases that equality can be
12150 proven at all loops, the innermost loops or no loops.
12151 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12152 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12153 (finalize_partitions, distribute_loop): Don't fuse partition of
12154 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12155 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12156 parloop is enabled.
12158 2018-01-23 Martin Liska <mliska@suse.cz>
12160 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12161 order to ignore the predictor.
12162 (PRED_POLYMORPHIC_CALL): Likewise.
12163 (PRED_RECURSIVE_CALL): Likewise.
12165 2018-01-23 Martin Liska <mliska@suse.cz>
12167 * tree-profile.c (tree_profiling): Print function header to
12168 aware reader which function we are working on.
12169 * value-prof.c (gimple_find_values_to_profile): Do not print
12170 not interesting value histograms.
12172 2018-01-23 Martin Liska <mliska@suse.cz>
12174 * profile-count.h (enum profile_quality): Add
12175 profile_uninitialized as the first value. Do not number values
12176 as they are zero based.
12177 (profile_count::verify): Update sanity check.
12178 (profile_probability::verify): Likewise.
12180 2018-01-23 Nathan Sidwell <nathan@acm.org>
12182 * doc/invoke.texi (ffor-scope): Deprecate.
12184 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12186 PR tree-optimization/83510
12187 * domwalk.c (set_all_edges_as_executable): New function.
12188 (dom_walker::dom_walker): Convert bool param
12189 "skip_unreachable_blocks" to enum reachability. Move setup of
12190 edge flags to set_all_edges_as_executable and only do it when
12191 reachability is REACHABLE_BLOCKS.
12192 * domwalk.h (enum dom_walker::reachability): New enum.
12193 (dom_walker::dom_walker): Convert bool param
12194 "skip_unreachable_blocks" to enum reachability.
12195 (set_all_edges_as_executable): New decl.
12196 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12197 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12199 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12200 but converting true to REACHABLE_BLOCKS.
12201 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12203 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12204 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12205 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12207 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12208 if check_all_array_refs will be called.
12210 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12212 * tree.c (selftest::test_location_wrappers): Add more test
12215 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12217 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12218 (selftest::test_bit_in_range): Likewise.
12220 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12223 * doc/sourcebuild.texi (vect_float): Say that the selector
12224 only describes the situation when -funsafe-math-optimizations is on.
12225 (vect_float_strict): Document.
12227 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12229 PR tree-optimization/83965
12230 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12231 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12232 instead of checking only for a reduction.
12233 (vect_recog_widen_sum_pattern): Likewise.
12235 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12237 * predict.c (probably_never_executed): Only use precise profile info.
12238 (compute_function_frequency): Skip after inlining hack since we now
12239 have quality checking.
12241 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12243 * profile-count.h (profile_probability::very_unlikely,
12244 profile_probability::unlikely, profile_probability::even): Set
12245 precision to guessed.
12247 2018-01-23 Richard Biener <rguenther@suse.de>
12249 PR tree-optimization/83963
12250 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12251 Properly terminate dominator walk when crossing the exit edge not
12252 when visiting its source block.
12254 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12257 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12258 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12260 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12262 PR tree-optimization/83957
12263 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12264 semicolon after for body surrounded by braces.
12266 PR tree-optimization/83081
12267 * profile-count.h (profile_probability::split): New method.
12268 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12269 Use profile_probability::split.
12270 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12271 when splitting a single conditional jump into 2.
12273 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12275 PR tree-optimization/69452
12276 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12279 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12281 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12282 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12283 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12285 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12287 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12289 * config/rl78/rl78.md (movdi): New define_expand.
12290 * config/rl78/rl78.c (rl78_split_movdi): New function.
12292 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12295 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12297 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12298 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12299 128-bit to produce an UNSPEC move to get the double word with the
12300 signbit and then a shift directly to do signbit.
12301 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12302 implementation with a new version that just does either a direct
12303 move or a regular move. Move memory interface to separate insns.
12304 Move insns so they are next to the expander.
12305 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12306 with signbit move. Split big and little endian case.
12307 (signbit<mode>2_dm_mem_le): Likewise.
12308 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12309 (signbit<mode>2_dm2): Likewise.
12311 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12313 * config/rl78/rl78.md (anddi3): New define_expand.
12315 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12317 * config/rl78/rl78.md (umindi3): New define_expand.
12319 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12321 * config/rl78/rl78.md (smindi3): New define_expand.
12323 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12325 * config/rl78/rl78.md (smaxdi3): New define_expand.
12327 2018-01-22 Carl Love <cel@us.ibm.com>
12329 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12330 LVX_V1TI): Add macro expansion.
12331 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12332 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12333 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12334 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12335 Change check to determine if the instruction is a byte reversing
12336 entry. Fix typo in comment.
12337 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12338 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12339 Add def_builtin calls for new builtins.
12340 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12341 Add define_insn expansion.
12343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12345 * config/rl78/rl78.md (umaxdi3): New define_expand.
12347 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12349 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12350 for non-QImode registers.
12352 2018-01-22 Richard Biener <rguenther@suse.de>
12354 PR tree-optimization/83963
12355 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12356 including the loop exit block.
12357 (scop_detection::merge_sese): Likewise.
12358 (scop_detection::add_scop): Do it here instead.
12360 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12362 * doc/sourcebuild.texi (arm_softfloat): Document.
12364 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12367 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12368 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12369 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12371 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12372 David Edelsohn <dje.gcc@gmail.com>
12375 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12376 Change "crset eq" to "crset 2".
12377 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12378 (*call_indirect_aix<mode>_nospec): Likewise.
12379 (*call_value_indirect_aix<mode>_nospec): Likewise.
12380 (*call_indirect_elfv2<mode>_nospec): Likewise.
12381 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12382 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12383 change assembly output from . to $.
12384 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12385 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12386 (*tablejump<mode>_internal1_nospec): Likewise.
12388 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12391 * config/sh/sh_optimize_sett_clrt.cc:
12392 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12394 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12396 PR tree-optimization/83940
12397 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12398 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12399 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12400 use it to pass back the definition type.
12401 (vect_check_store_rhs): Likewise rhs_dt_out.
12402 (vect_build_gather_load_calls): Add a mask_dt argument and use
12403 it instead of a call to vect_is_simple_use.
12404 (vectorizable_store): Update calls to vect_check_load_store_mask
12405 and vect_check_store_rhs. Use the dt returned by the latter instead
12406 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12407 instead of calls to vect_is_simple_use. Pass the scalar rather
12408 than the vector operand to vect_is_simple_use when handling
12409 second and subsequent copies of an rhs value.
12410 (vectorizable_load): Update calls to vect_check_load_store_mask
12411 and vect_build_gather_load_calls. Use the cached mask_dt and
12412 gs_info.offset_dt instead of calls to vect_is_simple_use.
12414 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12416 PR middle-end/83945
12417 * tree-emutls.c: Include gimplify.h.
12418 (lower_emutls_2): New function.
12419 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12420 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12421 it before further processing.
12424 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12425 UINTVAL (trueop1) instead of INTVAL (op1).
12427 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12431 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12432 INCOMING_FRAME_SP_OFFSET if not defined.
12433 (scan_trace): Add ENTRY argument. If true and
12434 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12435 emit a note to adjust the CFA offset.
12436 (create_cfi_notes): Adjust scan_trace callers.
12437 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12438 INCOMING_FRAME_SP_OFFSET in the CIE.
12439 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12440 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12442 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12443 * doc/tm.texi: Regenerated.
12445 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12447 PR rtl-optimization/83147
12448 * lra-constraints.c (remove_inheritance_pseudos): Use
12449 lra_substitute_pseudo_within_insn.
12451 2018-01-19 Tom de Vries <tom@codesourcery.com>
12452 Cesar Philippidis <cesar@codesourcery.com>
12455 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12457 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12460 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12461 spaces for function labels.
12463 2018-01-19 Martin Liska <mliska@suse.cz>
12465 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12466 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12467 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12468 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12469 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12470 (PRED_CONST_RETURN): Change from 69 to 65.
12471 (PRED_NULL_RETURN): Change from 91 to 71.
12472 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12473 (PRED_LOOP_GUARD): Change from 66 to 73.
12475 2018-01-19 Martin Liska <mliska@suse.cz>
12477 * predict.c (predict_insn_def): Add new assert.
12478 (struct branch_predictor): Change type to signed integer.
12479 (test_prediction_value_range): Amend test to cover
12480 PROB_UNINITIALIZED.
12481 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12482 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12483 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12484 (PRED_LOOP_IV_COMPARE): Likewise.
12485 * predict.h (PROB_UNINITIALIZED): Define new constant.
12487 2018-01-19 Martin Liska <mliska@suse.cz>
12489 * predict.c (dump_prediction): Add new format for
12490 analyze_brprob.py script which is enabled with -details
12492 * profile-count.h (precise_p): New function.
12494 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12496 PR tree-optimization/83922
12497 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12498 there are no statements that need masking.
12499 (vect_active_double_reduction_p): New function.
12500 (vect_analyze_loop_operations): Use it when handling phis that
12501 are not in the loop header.
12503 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12505 PR tree-optimization/83914
12506 * tree-vect-loop.c (vectorizable_induction): Don't convert
12507 init_expr or apply the peeling adjustment for inductions
12508 that are nested within the vectorized loop.
12510 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12512 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12515 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12519 * function.h (gimplify_parameters): Add gimple_seq * argument.
12520 * function.c: Include gimple.h and options.h.
12521 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12522 for the added local temporaries if needed.
12523 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12524 if there are any parameter cleanups, wrap whole body into a
12525 try/finally with the cleanups.
12527 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12530 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12531 Use GET_MODE_CLASS for scalar floating point.
12533 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12537 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12538 Fix call of call_cgraph_insertion_hooks.
12540 2018-01-18 Martin Sebor <msebor@redhat.com>
12542 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12544 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12547 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12550 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12553 * common.opt: (-ffile-prefix-map): New option.
12554 * opts.c (common_handle_option): Defer it.
12555 * opts-global.c (handle_common_deferred_options): Handle it.
12556 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12557 * file-prefix-map.h: New file.
12558 (remap_debug_filename, add_debug_prefix_map): ...here.
12559 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12560 * final.c (debug_prefix_map, add_debug_prefix_map
12561 remap_debug_filename): Move to...
12562 * file-prefix-map.c: New file.
12563 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12564 generalize, get rid of alloca(), use strrchr() instead of strchr().
12565 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12566 Implement in terms of add_prefix_map().
12567 (remap_macro_filename, remap_debug_filename): Implement in term of
12569 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12570 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12571 * dbxout.c: Include file-prefix-map.h.
12572 * varasm.c: Likewise.
12573 * vmsdbgout.c: Likewise.
12574 * xcoffout.c: Likewise.
12575 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12576 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12577 * doc/invoke.texi (-ffile-prefix-map): Document.
12578 (-fdebug-prefix-map): Update description.
12580 2018-01-18 Martin Liska <mliska@suse.cz>
12582 * config/i386/i386.c (indirect_thunk_name): Document that also
12584 (output_indirect_thunk): Document why both instructions
12585 (pause and lfence) are generated.
12587 2018-01-18 Richard Biener <rguenther@suse.de>
12589 PR tree-optimization/83887
12590 * graphite-scop-detection.c
12591 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12592 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12593 (scop_detection::merge_sese): Re-implement with a flood-fill
12594 algorithm that properly finds a SESE region if it exists.
12596 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12599 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12600 pointer_diff optimizations use view_convert instead of convert.
12602 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12604 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12605 Generate different code for -mno-speculate-indirect-jumps.
12606 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12607 (*call_indirect_aix<mode>): Disable for
12608 -mno-speculate-indirect-jumps.
12609 (*call_indirect_aix<mode>_nospec): New define_insn.
12610 (*call_value_indirect_aix<mode>): Disable for
12611 -mno-speculate-indirect-jumps.
12612 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12613 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12614 -mno-speculate-indirect-jumps.
12615 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12617 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12619 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12620 long double type, set the flags for noting the default long double
12621 type, even if we don't pass or return a long double type.
12623 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12626 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12629 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12631 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12632 support for merge[hl].
12633 (fold_mergehl_helper): New helper function.
12634 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12635 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12636 (altivec_vmrglw_direct): Add xxmrglw insn.
12638 2018-01-17 Andrew Waterman <andrew@sifive.com>
12640 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12641 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12643 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12646 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12647 call the lto_location_cache before reading the
12648 DECL_SOURCE_LOCATION of the types.
12650 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12651 Richard Sandiford <richard.sandiford@linaro.org>
12653 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12654 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12655 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12656 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12657 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12659 * config/aarch64/constraints.md (aarch64_movti_operand):
12661 * config/aarch64/predicates.md (Uti): Add new constraint.
12663 2018-01-17 Carl Love <cel@us.ibm.com>
12665 * config/rs6000/vsx.md (define_expand xl_len_r,
12666 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12667 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12669 (define_expand, define_insn): Move the shift left from the
12670 define_insn to the define_expand for lxvl and stxvl instructions.
12671 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12672 and XL_LEN_R definitions to PURE.
12674 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12676 * config/i386/i386.c (indirect_thunk_name): Declare regno
12677 as unsigned int. Compare regno with INVALID_REGNUM.
12678 (output_indirect_thunk): Ditto.
12679 (output_indirect_thunk_function): Ditto.
12680 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12681 in the call to output_indirect_thunk_function.
12683 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12685 PR middle-end/83884
12686 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12687 rather than the size of inner_type to determine the stack slot size
12688 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12690 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12693 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12696 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12698 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12699 endian Linux systems to optionally enable multilibs for selecting
12700 the long double type if the user configured an explicit type.
12701 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12702 have no long double multilibs if not defined.
12703 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12704 warn if the user used -mabi={ieee,ibm}longdouble and we built
12705 multilibs for long double.
12706 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12707 appropriate multilib option.
12708 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12710 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12711 for building long double multilibs.
12712 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12714 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12716 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12719 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12721 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12724 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12727 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12730 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12732 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12733 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12735 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12737 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12738 different rtl trees depending on TARGET_64BIT.
12739 (rs6000_gen_lvx): Likewise.
12741 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12743 * config/visium/visium.md (nop): Tweak comment.
12744 (hazard_nop): Likewise.
12746 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12748 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12749 -mspeculate-indirect-jumps.
12750 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12751 for -mno-speculate-indirect-jumps.
12752 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12753 (*call_value_indirect_elfv2<mode>): Disable for
12754 -mno-speculate-indirect-jumps.
12755 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12756 (indirect_jump): Emit different RTL for
12757 -mno-speculate-indirect-jumps.
12758 (*indirect_jump<mode>): Disable for
12759 -mno-speculate-indirect-jumps.
12760 (*indirect_jump<mode>_nospec): New define_insn.
12761 (tablejump): Emit different RTL for
12762 -mno-speculate-indirect-jumps.
12763 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12764 (tablejumpsi_nospec): New define_expand.
12765 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12766 (tablejumpdi_nospec): New define_expand.
12767 (*tablejump<mode>_internal1): Disable for
12768 -mno-speculate-indirect-jumps.
12769 (*tablejump<mode>_internal1_nospec): New define_insn.
12770 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12773 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12775 * caller-save.c (insert_save): Drop unnecessary parameter. All
12778 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12779 Richard Biener <rguenth@suse.de>
12782 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12783 return early, inline manually is_gimple_sizepos. Make sure if we
12784 call gimplify_expr we don't end up with a gimple constant.
12785 * tree.c (variably_modified_type_p): Don't return true for
12786 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12787 * gimplify.h (is_gimple_sizepos): Remove.
12789 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12791 PR tree-optimization/83857
12792 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12793 vectorizable_live_operation for pure SLP statements.
12794 (vectorizable_live_operation): Handle PHIs.
12796 2018-01-16 Richard Biener <rguenther@suse.de>
12798 PR tree-optimization/83867
12799 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12800 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12802 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12805 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12806 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12807 If off is not INTEGER_CST, issue a may not be aligned warning
12808 rather than isn't aligned. Use isn%'t rather than isn't.
12809 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12811 <case MULT_EXPR>: Improve the case when bottom and one of the
12812 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12813 operand, in that case check if the other operand is multiple of
12814 bottom divided by the INTEGER_CST operand.
12816 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12819 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12820 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12821 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12822 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12823 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12824 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12825 (pa_function_arg_size): New function.
12827 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12829 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12830 in a separate statement.
12832 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12834 PR tree-optimization/83847
12835 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12836 group gathers and scatters.
12838 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12840 PR rtl-optimization/86620
12841 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12843 PR rtl-optimization/83213
12844 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12845 to last if both are JUMP_INSNs.
12847 PR tree-optimization/83843
12848 * gimple-ssa-store-merging.c
12849 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12850 store_immediate_info for bswap/nop orig_stores.
12852 2018-01-15 Andrew Waterman <andrew@sifive.com>
12854 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12856 <UDIV>: Increase cost if !TARGET_DIV.
12858 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12860 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12861 (define_attr "cr_logical_3op"): New.
12862 (cceq_ior_compare): Adjust.
12863 (cceq_ior_compare_complement): Adjust.
12864 (*cceq_rev_compare): Adjust.
12865 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12866 (is_cracked_insn): Adjust.
12867 (insn_must_be_first_in_group): Adjust.
12868 * config/rs6000/40x.md: Adjust.
12869 * config/rs6000/440.md: Adjust.
12870 * config/rs6000/476.md: Adjust.
12871 * config/rs6000/601.md: Adjust.
12872 * config/rs6000/603.md: Adjust.
12873 * config/rs6000/6xx.md: Adjust.
12874 * config/rs6000/7450.md: Adjust.
12875 * config/rs6000/7xx.md: Adjust.
12876 * config/rs6000/8540.md: Adjust.
12877 * config/rs6000/cell.md: Adjust.
12878 * config/rs6000/e300c2c3.md: Adjust.
12879 * config/rs6000/e500mc.md: Adjust.
12880 * config/rs6000/e500mc64.md: Adjust.
12881 * config/rs6000/e5500.md: Adjust.
12882 * config/rs6000/e6500.md: Adjust.
12883 * config/rs6000/mpc.md: Adjust.
12884 * config/rs6000/power4.md: Adjust.
12885 * config/rs6000/power5.md: Adjust.
12886 * config/rs6000/power6.md: Adjust.
12887 * config/rs6000/power7.md: Adjust.
12888 * config/rs6000/power8.md: Adjust.
12889 * config/rs6000/power9.md: Adjust.
12890 * config/rs6000/rs64.md: Adjust.
12891 * config/rs6000/titan.md: Adjust.
12893 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12895 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12896 ix86_indirect_branch_register logic.
12898 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12900 * config/i386/constraints.md (Bs): Update
12901 ix86_indirect_branch_register check. Don't check
12902 ix86_indirect_branch_register with GOT_memory_operand.
12904 * config/i386/predicates.md (GOT_memory_operand): Don't check
12905 ix86_indirect_branch_register here.
12906 (GOT32_symbol_operand): Likewise.
12908 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12910 * config/i386/predicates.md (constant_call_address_operand):
12911 Rewrite ix86_indirect_branch_register logic.
12912 (sibcall_insn_operand): Likewise.
12914 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12916 * config/i386/constraints.md (Bs): Replace
12917 ix86_indirect_branch_thunk_register with
12918 ix86_indirect_branch_register.
12920 * config/i386/i386.md (indirect_jump): Likewise.
12921 (tablejump): Likewise.
12922 (*sibcall_memory): Likewise.
12923 (*sibcall_value_memory): Likewise.
12924 Peepholes of indirect call and jump via memory: Likewise.
12925 * config/i386/i386.opt: Likewise.
12926 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12927 (GOT_memory_operand): Likewise.
12928 (call_insn_operand): Likewise.
12929 (sibcall_insn_operand): Likewise.
12930 (GOT32_symbol_operand): Likewise.
12932 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12934 PR middle-end/83837
12935 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12936 type rather than type addr's type points to.
12937 (expand_omp_atomic_mutex): Likewise.
12938 (expand_omp_atomic): Likewise.
12940 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12943 * config/i386/i386.c (output_indirect_thunk_function): Use
12944 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12945 for __x86_return_thunk.
12947 2018-01-15 Richard Biener <rguenther@suse.de>
12949 PR middle-end/83850
12950 * expmed.c (extract_bit_field_1): Fix typo.
12952 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12955 * config/arm/iterators.md (VF): New mode iterator.
12956 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12957 Remove integer-related logic from pattern.
12958 (neon_vabd<mode>_3): Likewise.
12960 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12962 PR middle-end/82694
12963 * common.opt (fstrict-overflow): No longer an alias.
12964 (fwrapv-pointer): New option.
12965 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12966 also for pointer types based on flag_wrapv_pointer.
12967 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12968 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12969 opts->x_flag_wrapv got set.
12970 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12971 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12972 POINTER_TYPE_OVERFLOW_UNDEFINED.
12973 * match.pd: Likewise in address comparison pattern.
12974 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12976 2018-01-15 Richard Biener <rguenther@suse.de>
12979 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12980 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12981 Reset type names to their identifier if their TYPE_DECL doesn't
12982 have linkage (and thus is used for ODR and devirt).
12983 (save_debug_info_for_decl): Remove.
12984 (save_debug_info_for_type): Likewise.
12985 (add_tree_to_fld_list): Adjust.
12986 * tree-pretty-print.c (dump_generic_node): Make dumping of
12987 type names more robust.
12989 2018-01-15 Richard Biener <rguenther@suse.de>
12991 * BASE-VER: Bump to 8.0.1.
12993 2018-01-14 Martin Sebor <msebor@redhat.com>
12996 * builtins.c (check_access): Avoid warning when the no-warning bit
12999 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13001 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13002 * ira-color (allocno_hard_regs_compare): Likewise.
13004 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13007 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13008 Use .pushsection/.popsection.
13010 2018-01-14 Martin Sebor <msebor@redhat.com>
13013 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13015 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13017 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13018 entry from extra_headers.
13019 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13020 extra_headers, make the list bitwise identical to the i?86-*-* one.
13022 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13024 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13025 -mcmodel=large with -mindirect-branch=thunk,
13026 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13027 -mfunction-return=thunk-extern.
13028 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13029 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13030 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13032 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13034 * config/i386/i386.c (print_reg): Print the name of the full
13035 integer register without '%'.
13036 (ix86_print_operand): Handle 'V'.
13037 * doc/extend.texi: Document 'V' modifier.
13039 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13041 * config/i386/constraints.md (Bs): Disallow memory operand for
13042 -mindirect-branch-register.
13044 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13045 (GOT_memory_operand): Likewise.
13046 (call_insn_operand): Likewise.
13047 (sibcall_insn_operand): Likewise.
13048 (GOT32_symbol_operand): Likewise.
13049 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13050 for -mindirect-branch-register.
13051 (tablejump): Likewise.
13052 (*sibcall_memory): Likewise.
13053 (*sibcall_value_memory): Likewise.
13054 Disallow peepholes of indirect call and jump via memory for
13055 -mindirect-branch-register.
13056 (*call_pop): Replace m with Bw.
13057 (*call_value_pop): Likewise.
13058 (*sibcall_pop_memory): Replace m with Bs.
13059 * config/i386/i386.opt (mindirect-branch-register): New option.
13060 * doc/invoke.texi: Document -mindirect-branch-register option.
13062 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13064 * config/i386/i386-protos.h (ix86_output_function_return): New.
13065 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13066 set function_return_type.
13067 (indirect_thunk_name): Add ret_p to indicate thunk for function
13069 (output_indirect_thunk_function): Pass false to
13070 indirect_thunk_name.
13071 (ix86_output_indirect_branch_via_reg): Likewise.
13072 (ix86_output_indirect_branch_via_push): Likewise.
13073 (output_indirect_thunk_function): Create alias for function
13074 return thunk if regno < 0.
13075 (ix86_output_function_return): New function.
13076 (ix86_handle_fndecl_attribute): Handle function_return.
13077 (ix86_attribute_table): Add function_return.
13078 * config/i386/i386.h (machine_function): Add
13079 function_return_type.
13080 * config/i386/i386.md (simple_return_internal): Use
13081 ix86_output_function_return.
13082 (simple_return_internal_long): Likewise.
13083 * config/i386/i386.opt (mfunction-return=): New option.
13084 (indirect_branch): Mention -mfunction-return=.
13085 * doc/extend.texi: Document function_return function attribute.
13086 * doc/invoke.texi: Document -mfunction-return= option.
13088 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13090 * config/i386/i386-opts.h (indirect_branch): New.
13091 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13092 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13093 with local indirect jump when converting indirect call and jump.
13094 (ix86_set_indirect_branch_type): New.
13095 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13096 (indirectlabelno): New.
13097 (indirect_thunk_needed): Likewise.
13098 (indirect_thunk_bnd_needed): Likewise.
13099 (indirect_thunks_used): Likewise.
13100 (indirect_thunks_bnd_used): Likewise.
13101 (INDIRECT_LABEL): Likewise.
13102 (indirect_thunk_name): Likewise.
13103 (output_indirect_thunk): Likewise.
13104 (output_indirect_thunk_function): Likewise.
13105 (ix86_output_indirect_branch_via_reg): Likewise.
13106 (ix86_output_indirect_branch_via_push): Likewise.
13107 (ix86_output_indirect_branch): Likewise.
13108 (ix86_output_indirect_jmp): Likewise.
13109 (ix86_code_end): Call output_indirect_thunk_function if needed.
13110 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13112 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13113 (ix86_attribute_table): Add indirect_branch.
13114 * config/i386/i386.h (machine_function): Add indirect_branch_type
13115 and has_local_indirect_jump.
13116 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13118 (tablejump): Likewise.
13119 (*indirect_jump): Use ix86_output_indirect_jmp.
13120 (*tablejump_1): Likewise.
13121 (simple_return_indirect_internal): Likewise.
13122 * config/i386/i386.opt (mindirect-branch=): New option.
13123 (indirect_branch): New.
13126 (thunk-inline): Likewise.
13127 (thunk-extern): Likewise.
13128 * doc/extend.texi: Document indirect_branch function attribute.
13129 * doc/invoke.texi: Document -mindirect-branch= option.
13131 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13134 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13136 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13138 * ipa-inline.c (want_inline_small_function_p): Return false if
13139 inlining has already failed with CIF_FINAL_ERROR.
13140 (update_caller_keys): Call want_inline_small_function_p before
13142 (update_callee_keys): Likewise.
13144 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13146 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13148 (rs6000_quadword_masked_address_p): Likewise.
13149 (quad_aligned_load_p): Likewise.
13150 (quad_aligned_store_p): Likewise.
13151 (const_load_sequence_p): Add comment to describe the outer-most loop.
13152 (mimic_memory_attributes_and_flags): New function.
13153 (rs6000_gen_stvx): Likewise.
13154 (replace_swapped_aligned_store): Likewise.
13155 (rs6000_gen_lvx): Likewise.
13156 (replace_swapped_aligned_load): Likewise.
13157 (replace_swapped_load_constant): Capitalize argument name in
13158 comment describing this function.
13159 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13160 and stores that access quad-word aligned addresses and replace
13161 with stvx or lvx instructions when appropriate.
13162 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13163 New function prototype.
13164 (rs6000_quadword_masked_address_p): Likewise.
13165 (rs6000_gen_lvx): Likewise.
13166 (rs6000_gen_stvx): Likewise.
13167 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13168 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13169 when memory address is aligned.
13170 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13171 this split to select lvx instruction when memory address is aligned.
13172 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13173 instruction when memory address is aligned.
13174 (*vsx_le_perm_load_v16qi): Likewise.
13175 (four unnamed splitters): Modify to select the stvx instruction
13176 when memory is aligned.
13178 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13180 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13181 which appears in the queue multiple times.
13183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13184 Alan Hayward <alan.hayward@arm.com>
13185 David Sherwood <david.sherwood@arm.com>
13187 * tree-vectorizer.h (vec_lower_bound): New structure.
13188 (_loop_vec_info): Add check_nonzero and lower_bounds.
13189 (LOOP_VINFO_CHECK_NONZERO): New macro.
13190 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13191 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13192 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13193 fields. Make seg_len the distance travelled, not including the
13195 (dr_direction_indicator): Declare.
13196 (dr_zero_step_indicator): Likewise.
13197 (dr_known_forward_stride_p): Likewise.
13198 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13200 (runtime_alias_check_p): Allow runtime alias checks with
13202 (operator ==): Compare access_size and align.
13203 (prune_runtime_alias_test_list): Rework for new distinction between
13204 the access_size and seg_len.
13205 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13207 (get_segment_min_max): New function.
13208 (create_intersect_range_checks): Use it.
13209 (dr_step_indicator): New function.
13210 (dr_direction_indicator): Likewise.
13211 (dr_zero_step_indicator): Likewise.
13212 (dr_known_forward_stride_p): Likewise.
13213 * tree-loop-distribution.c (data_ref_segment_size): Return
13214 DR_STEP * (niters - 1).
13215 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13217 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13218 (vect_preserves_scalar_order_p): New function, split out from...
13219 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13220 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13221 (vect_vfa_access_size): New function.
13222 (vect_vfa_align): Likewise.
13223 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13224 (dump_lower_bound): New function.
13225 (vect_check_lower_bound): Likewise.
13226 (vect_small_gap_p): Likewise.
13227 (vectorizable_with_step_bound_p): Likewise.
13228 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13229 depencies if the vectorization factor is 1. Convert the checks
13230 for nonzero steps into checks on the bounds of DR_STEP. Try using
13231 a bunds check for variable steps if the minimum required step is
13232 relatively small. Update calls to the dr_with_seg_len
13233 constructor and to vect_compile_time_alias.
13234 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13236 (vect_loop_versioning): Call it.
13237 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13239 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13242 Alan Hayward <alan.hayward@arm.com>
13243 David Sherwood <david.sherwood@arm.com>
13245 * doc/sourcebuild.texi (vect_scatter_store): Document.
13246 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13248 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13250 * genopinit.c (main): Add supports_vec_scatter_store and
13251 supports_vec_scatter_store_cached to target_optabs.
13252 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13253 IFN_MASK_SCATTER_STORE.
13254 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13256 * internal-fn.h (internal_store_fn_p): Declare.
13257 (internal_fn_stored_value_index): Likewise.
13258 * internal-fn.c (scatter_store_direct): New macro.
13259 (expand_scatter_store_optab_fn): New function.
13260 (direct_scatter_store_optab_supported_p): New macro.
13261 (internal_store_fn_p): New function.
13262 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13263 IFN_MASK_SCATTER_STORE.
13264 (internal_fn_mask_index): Likewise.
13265 (internal_fn_stored_value_index): New function.
13266 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13267 for scatter stores.
13268 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13269 * optabs-query.c (supports_vec_scatter_store_p): New function.
13270 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13271 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13272 true for scatter stores.
13273 (vect_gather_scatter_fn_p): Handle scatter stores too.
13274 (vect_check_gather_scatter): Consider using scatter stores if
13275 supports_vec_scatter_store_p.
13276 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13277 scatter stores too.
13278 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13279 internal_fn_stored_value_index.
13280 (check_load_store_masking): Handle scatter stores too.
13281 (vect_get_store_rhs): Make public.
13282 (vectorizable_call): Use internal_store_fn_p.
13283 (vectorizable_store): Handle scatter store internal functions.
13284 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13285 when deciding whether the end of the group has been reached.
13286 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13287 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13288 (mask_scatter_store<mode>): New insns.
13290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13291 Alan Hayward <alan.hayward@arm.com>
13292 David Sherwood <david.sherwood@arm.com>
13294 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13295 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13296 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13298 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13299 Use vect_truncate_gather_scatter_offset if we can't treat the
13300 operation as a normal gather load or scatter store.
13301 (get_group_load_store_type): Take the gather_scatter_info
13302 as argument. Try using a gather load or scatter store for
13303 single-element groups.
13304 (get_load_store_type): Update calls to get_group_load_store_type
13305 and vect_use_strided_gather_scatters_p.
13307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13308 Alan Hayward <alan.hayward@arm.com>
13309 David Sherwood <david.sherwood@arm.com>
13311 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13312 optional tree argument.
13313 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13315 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13316 but continue to use the current value as a fallback.
13317 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13318 to compare the updates.
13319 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13320 (get_load_store_type): Use it when handling a strided access.
13321 (vect_get_strided_load_store_ops): New function.
13322 (vect_get_data_ptr_increment): Likewise.
13323 (vectorizable_load): Handle strided gather loads. Always pass
13324 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13327 Alan Hayward <alan.hayward@arm.com>
13328 David Sherwood <david.sherwood@arm.com>
13330 * doc/md.texi (gather_load@var{m}): Document.
13331 (mask_gather_load@var{m}): Likewise.
13332 * genopinit.c (main): Add supports_vec_gather_load and
13333 supports_vec_gather_load_cached to target_optabs.
13334 * optabs-tree.c (init_tree_optimization_optabs): Use
13335 ggc_cleared_alloc to allocate target_optabs.
13336 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13337 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13339 * internal-fn.h (internal_load_fn_p): Declare.
13340 (internal_gather_scatter_fn_p): Likewise.
13341 (internal_fn_mask_index): Likewise.
13342 (internal_gather_scatter_fn_supported_p): Likewise.
13343 * internal-fn.c (gather_load_direct): New macro.
13344 (expand_gather_load_optab_fn): New function.
13345 (direct_gather_load_optab_supported_p): New macro.
13346 (direct_internal_fn_optab): New function.
13347 (internal_load_fn_p): Likewise.
13348 (internal_gather_scatter_fn_p): Likewise.
13349 (internal_fn_mask_index): Likewise.
13350 (internal_gather_scatter_fn_supported_p): Likewise.
13351 * optabs-query.c (supports_at_least_one_mode_p): New function.
13352 (supports_vec_gather_load_p): Likewise.
13353 * optabs-query.h (supports_vec_gather_load_p): Declare.
13354 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13355 and memory_type field.
13356 (NUM_PATTERNS): Bump to 15.
13357 * tree-vect-data-refs.c: Include internal-fn.h.
13358 (vect_gather_scatter_fn_p): New function.
13359 (vect_describe_gather_scatter_call): Likewise.
13360 (vect_check_gather_scatter): Try using internal functions for
13361 gather loads. Recognize existing calls to a gather load function.
13362 (vect_analyze_data_refs): Consider using gather loads if
13363 supports_vec_gather_load_p.
13364 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13365 (vect_get_gather_scatter_offset_type): Likewise.
13366 (vect_convert_mask_for_vectype): Likewise.
13367 (vect_add_conversion_to_patterm): Likewise.
13368 (vect_try_gather_scatter_pattern): Likewise.
13369 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13370 (vect_vect_recog_func_ptrs): Add it.
13371 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13372 internal_fn_mask_index and internal_gather_scatter_fn_p.
13373 (check_load_store_masking): Take the gather_scatter_info as an
13374 argument and handle gather loads.
13375 (vect_get_gather_scatter_ops): New function.
13376 (vectorizable_call): Check internal_load_fn_p.
13377 (vectorizable_load): Likewise. Handle gather load internal
13379 (vectorizable_store): Update call to check_load_store_masking.
13380 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13381 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13382 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13383 (aarch64_gather_scale_operand_d): New predicates.
13384 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13385 (mask_gather_load<mode>): New insns.
13387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13388 Alan Hayward <alan.hayward@arm.com>
13389 David Sherwood <david.sherwood@arm.com>
13391 * optabs.def (fold_left_plus_optab): New optab.
13392 * doc/md.texi (fold_left_plus_@var{m}): Document.
13393 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13394 * internal-fn.c (fold_left_direct): Define.
13395 (expand_fold_left_optab_fn): Likewise.
13396 (direct_fold_left_optab_supported_p): Likewise.
13397 * fold-const-call.c (fold_const_fold_left): New function.
13398 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13399 * tree-parloops.c (valid_reduction_p): New function.
13400 (gather_scalar_reductions): Use it.
13401 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13402 (vect_finish_replace_stmt): Declare.
13403 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13404 (needs_fold_left_reduction_p): New function, split out from...
13405 (vect_is_simple_reduction): ...here. Accept reductions that
13406 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13407 (vect_force_simple_reduction): Also store the reduction type in
13408 the assignment's STMT_VINFO_REDUC_TYPE.
13409 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13410 (merge_with_identity): New function.
13411 (vect_expand_fold_left): Likewise.
13412 (vectorize_fold_left_reduction): Likewise.
13413 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13414 scalar phi in place for it. Check for target support and reject
13415 cases that would reassociate the operation. Defer the transform
13416 phase to vectorize_fold_left_reduction.
13417 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13418 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13419 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13423 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13424 call to ifc_temp_var.
13426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13427 Alan Hayward <alan.hayward@arm.com>
13428 David Sherwood <david.sherwood@arm.com>
13430 * target.def (legitimize_address_displacement): Take the original
13431 offset as a poly_int.
13432 * targhooks.h (default_legitimize_address_displacement): Update
13434 * targhooks.c (default_legitimize_address_displacement): Likewise.
13435 * doc/tm.texi: Regenerate.
13436 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13437 as an argument, moving assert of ad->disp == ad->disp_term to...
13438 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13439 Try calling targetm.legitimize_address_displacement before expanding
13440 the address rather than afterwards, and adjust for the new interface.
13441 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13442 Match the new hook interface. Handle SVE addresses.
13443 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13444 new hook interface.
13446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13448 * Makefile.in (OBJS): Add early-remat.o.
13449 * target.def (select_early_remat_modes): New hook.
13450 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13451 * doc/tm.texi: Regenerate.
13452 * targhooks.h (default_select_early_remat_modes): Declare.
13453 * targhooks.c (default_select_early_remat_modes): New function.
13454 * timevar.def (TV_EARLY_REMAT): New timevar.
13455 * passes.def (pass_early_remat): New pass.
13456 * tree-pass.h (make_pass_early_remat): Declare.
13457 * early-remat.c: New file.
13458 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13460 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13462 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13463 Alan Hayward <alan.hayward@arm.com>
13464 David Sherwood <david.sherwood@arm.com>
13466 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13467 vfm1 with a bound_epilog parameter.
13468 (vect_do_peeling): Update calls accordingly, and move the prologue
13469 call earlier in the function. Treat the base bound_epilog as 0 for
13470 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13471 this base when peeling for gaps.
13472 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13473 with fully-masked loops.
13474 (vect_estimate_min_profitable_iters): Handle the single peeled
13475 iteration in that case.
13477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13478 Alan Hayward <alan.hayward@arm.com>
13479 David Sherwood <david.sherwood@arm.com>
13481 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13482 single-element interleaving even if the size is not a power of 2.
13483 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13484 accesses for single-element interleaving if the group size is
13487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13488 Alan Hayward <alan.hayward@arm.com>
13489 David Sherwood <david.sherwood@arm.com>
13491 * doc/md.texi (fold_extract_last_@var{m}): Document.
13492 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13493 * optabs.def (fold_extract_last_optab): New optab.
13494 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13495 * internal-fn.c (fold_extract_direct): New macro.
13496 (expand_fold_extract_optab_fn): Likewise.
13497 (direct_fold_extract_optab_supported_p): Likewise.
13498 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13499 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13500 EXTRACT_LAST_REDUCTION.
13501 (get_initial_def_for_reduction): Do not create an initial vector
13502 for EXTRACT_LAST_REDUCTION reductions.
13503 (vectorizable_reduction): Leave the scalar phi in place for
13504 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13505 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13506 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13507 transform phase to vectorizable_condition.
13508 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13510 (vect_finish_stmt_generation): ...here.
13511 (vect_finish_replace_stmt): New function.
13512 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13513 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13515 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13518 Alan Hayward <alan.hayward@arm.com>
13519 David Sherwood <david.sherwood@arm.com>
13521 * doc/md.texi (extract_last_@var{m}): Document.
13522 * optabs.def (extract_last_optab): New optab.
13523 * internal-fn.def (EXTRACT_LAST): New internal function.
13524 * internal-fn.c (cond_unary_direct): New macro.
13525 (expand_cond_unary_optab_fn): Likewise.
13526 (direct_cond_unary_optab_supported_p): Likewise.
13527 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13528 loops using EXTRACT_LAST.
13529 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13530 (extract_last_<mode>): ...this optab.
13531 (vec_extract<mode><Vel>): Update accordingly.
13533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13534 Alan Hayward <alan.hayward@arm.com>
13535 David Sherwood <david.sherwood@arm.com>
13537 * target.def (empty_mask_is_expensive): New hook.
13538 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13539 * doc/tm.texi: Regenerate.
13540 * targhooks.h (default_empty_mask_is_expensive): Declare.
13541 * targhooks.c (default_empty_mask_is_expensive): New function.
13542 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13543 if the target says that empty masks are expensive.
13544 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13546 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13549 Alan Hayward <alan.hayward@arm.com>
13550 David Sherwood <david.sherwood@arm.com>
13552 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13553 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13554 (vect_use_loop_mask_for_alignment_p): New function.
13555 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13556 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13557 niters_skip argument. Make sure that the first niters_skip elements
13558 of the first iteration are inactive.
13559 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13560 Update call to vect_set_loop_masks_directly.
13561 (get_misalign_in_elems): New function, split out from...
13562 (vect_gen_prolog_loop_niters): ...here.
13563 (vect_update_init_of_dr): Take a code argument that specifies whether
13564 the adjustment should be added or subtracted.
13565 (vect_update_init_of_drs): Likewise.
13566 (vect_prepare_for_masked_peels): New function.
13567 (vect_do_peeling): Skip prologue peeling if we're using a mask
13568 instead. Update call to vect_update_inits_of_drs.
13569 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13571 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13572 alignment. Do not include the number of peeled iterations in
13573 the minimum threshold in that case.
13574 (vectorizable_induction): Adjust the start value down by
13575 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13576 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13577 Take the number of skipped iterations into account when calculating
13579 * tree-vect-stmts.c (vect_gen_while_not): New function.
13581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13582 Alan Hayward <alan.hayward@arm.com>
13583 David Sherwood <david.sherwood@arm.com>
13585 * doc/sourcebuild.texi (vect_fully_masked): Document.
13586 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13587 default value to 0.
13588 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13590 (vect_analyze_loop_2): ...here. Don't check the vectorization
13591 factor against the number of loop iterations if the loop is
13594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13595 Alan Hayward <alan.hayward@arm.com>
13596 David Sherwood <david.sherwood@arm.com>
13598 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13599 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13600 (dump_groups): Update accordingly.
13601 (iv_use::mem_type): New member variable.
13602 (address_p): New function.
13603 (record_use): Add a mem_type argument and initialize the new
13605 (record_group_use): Add a mem_type argument. Use address_p.
13606 Remove obsolete null checks of base_object. Update call to record_use.
13607 (find_interesting_uses_op): Update call to record_group_use.
13608 (find_interesting_uses_cond): Likewise.
13609 (find_interesting_uses_address): Likewise.
13610 (get_mem_type_for_internal_fn): New function.
13611 (find_address_like_use): Likewise.
13612 (find_interesting_uses_stmt): Try find_address_like_use before
13613 calling find_interesting_uses_op.
13614 (addr_offset_valid_p): Use the iv mem_type field as the type
13615 of the addressed memory.
13616 (add_autoinc_candidates): Likewise.
13617 (get_address_cost): Likewise.
13618 (split_small_address_groups_p): Use address_p.
13619 (split_address_groups): Likewise.
13620 (add_iv_candidate_for_use): Likewise.
13621 (autoinc_possible_for_pair): Likewise.
13622 (rewrite_groups): Likewise.
13623 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13624 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13625 (get_alias_ptr_type_for_ptr_address): New function.
13626 (rewrite_use_address): Rewrite address uses in calls that were
13627 identified by find_address_like_use.
13629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13630 Alan Hayward <alan.hayward@arm.com>
13631 David Sherwood <david.sherwood@arm.com>
13633 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13635 * gimple-expr.h (is_gimple_addressable: Likewise.
13636 * gimple-expr.c (is_gimple_address): Likewise.
13637 * internal-fn.c (expand_call_mem_ref): New function.
13638 (expand_mask_load_optab_fn): Use it.
13639 (expand_mask_store_optab_fn): Likewise.
13641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13642 Alan Hayward <alan.hayward@arm.com>
13643 David Sherwood <david.sherwood@arm.com>
13645 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13646 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13647 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13648 (cond_umax@var{mode}): Document.
13649 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13650 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13651 (cond_umin_optab, cond_umax_optab): New optabs.
13652 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13653 (COND_IOR, COND_XOR): New internal functions.
13654 * internal-fn.h (get_conditional_internal_fn): Declare.
13655 * internal-fn.c (cond_binary_direct): New macro.
13656 (expand_cond_binary_optab_fn): Likewise.
13657 (direct_cond_binary_optab_supported_p): Likewise.
13658 (get_conditional_internal_fn): New function.
13659 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13660 Cope with reduction statements that are vectorized as calls rather
13662 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13663 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13664 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13665 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13666 (UNSPEC_COND_EOR): New unspecs.
13667 (optab): Add mappings for them.
13668 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13669 (sve_int_op, sve_fp_op): New int attributes.
13671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13672 Alan Hayward <alan.hayward@arm.com>
13673 David Sherwood <david.sherwood@arm.com>
13675 * optabs.def (while_ult_optab): New optab.
13676 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13677 * internal-fn.def (WHILE_ULT): New internal function.
13678 * internal-fn.h (direct_internal_fn_supported_p): New override
13679 that takes two types as argument.
13680 * internal-fn.c (while_direct): New macro.
13681 (expand_while_optab_fn): New function.
13682 (convert_optab_supported_p): Likewise.
13683 (direct_while_optab_supported_p): New macro.
13684 * wide-int.h (wi::udiv_ceil): New function.
13685 * tree-vectorizer.h (rgroup_masks): New structure.
13686 (vec_loop_masks): New typedef.
13687 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13688 and fully_masked_p.
13689 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13690 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13691 (vect_max_vf): New function.
13692 (slpeel_make_loop_iterate_ntimes): Delete.
13693 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13694 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13695 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13696 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13697 internal-fn.h, stor-layout.h and optabs-query.h.
13698 (vect_set_loop_mask): New function.
13699 (add_preheader_seq): Likewise.
13700 (add_header_seq): Likewise.
13701 (interleave_supported_p): Likewise.
13702 (vect_maybe_permute_loop_masks): Likewise.
13703 (vect_set_loop_masks_directly): Likewise.
13704 (vect_set_loop_condition_masked): Likewise.
13705 (vect_set_loop_condition_unmasked): New function, split out from
13706 slpeel_make_loop_iterate_ntimes.
13707 (slpeel_make_loop_iterate_ntimes): Rename to..
13708 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13709 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13710 (vect_do_peeling): Update call accordingly.
13711 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13713 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13714 mask_compare_type, can_fully_mask_p and fully_masked_p.
13715 (release_vec_loop_masks): New function.
13716 (_loop_vec_info): Use it to free the loop masks.
13717 (can_produce_all_loop_masks_p): New function.
13718 (vect_get_max_nscalars_per_iter): Likewise.
13719 (vect_verify_full_masking): Likewise.
13720 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13721 retries, and free the mask rgroups before retrying. Check loop-wide
13722 reasons for disallowing fully-masked loops. Make the final decision
13723 about whether use a fully-masked loop or not.
13724 (vect_estimate_min_profitable_iters): Do not assume that peeling
13725 for the number of iterations will be needed for fully-masked loops.
13726 (vectorizable_reduction): Disable fully-masked loops.
13727 (vectorizable_live_operation): Likewise.
13728 (vect_halve_mask_nunits): New function.
13729 (vect_double_mask_nunits): Likewise.
13730 (vect_record_loop_mask): Likewise.
13731 (vect_get_loop_mask): Likewise.
13732 (vect_transform_loop): Handle the case in which the final loop
13733 iteration might handle a partial vector. Call vect_set_loop_condition
13734 instead of slpeel_make_loop_iterate_ntimes.
13735 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13736 (check_load_store_masking): New function.
13737 (prepare_load_store_mask): Likewise.
13738 (vectorizable_store): Handle fully-masked loops.
13739 (vectorizable_load): Likewise.
13740 (supportable_widening_operation): Use vect_halve_mask_nunits for
13742 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13743 (vect_gen_while): New function.
13744 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13745 (aarch64_uqdec<mode>): New insn.
13747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13752 (reduc_xor_scal_optab): New optabs.
13753 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13754 (reduc_xor_scal_@var{m}): Document.
13755 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13756 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13757 internal functions.
13758 * fold-const-call.c (fold_const_call): Handle them.
13759 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13760 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13761 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13762 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13763 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13764 (UNSPEC_XORV): New unspecs.
13765 (optab): Add entries for them.
13766 (BITWISEV): New int iterator.
13767 (bit_reduc_op): New int attributes.
13769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13770 Alan Hayward <alan.hayward@arm.com>
13771 David Sherwood <david.sherwood@arm.com>
13773 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13774 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13775 * optabs.def (vec_shl_insert_optab): New optab.
13776 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13777 (duplicate_and_interleave): Likewise.
13778 * tree-vect-loop.c: Include internal-fn.h.
13779 (neutral_op_for_slp_reduction): New function, split out from
13780 get_initial_defs_for_reduction.
13781 (get_initial_def_for_reduction): Handle option 2 for variable-length
13782 vectors by loading the neutral value into a vector and then shifting
13783 the initial value into element 0.
13784 (get_initial_defs_for_reduction): Replace the code argument with
13785 the neutral value calculated by neutral_op_for_slp_reduction.
13786 Use gimple_build_vector for constant-length vectors.
13787 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13788 but the first group_size elements have a neutral value.
13789 Use duplicate_and_interleave otherwise.
13790 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13791 Update call to get_initial_defs_for_reduction. Handle SLP
13792 reductions for variable-length vectors by creating one vector
13793 result for each scalar result, with the elements associated
13794 with other scalar results stubbed out with the neutral value.
13795 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13796 Require IFN_VEC_SHL_INSERT for double reductions on
13797 variable-length vectors, or SLP reductions that have
13798 a neutral value. Require can_duplicate_and_interleave_p
13799 support for variable-length unchained SLP reductions if there
13800 is no neutral value, such as for MIN/MAX reductions. Also require
13801 the number of vector elements to be a multiple of the number of
13802 SLP statements when doing variable-length unchained SLP reductions.
13803 Update call to vect_create_epilog_for_reduction.
13804 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13805 and remove initial values.
13806 (duplicate_and_interleave): Make public.
13807 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13808 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13815 (can_duplicate_and_interleave_p): New function.
13816 (vect_get_and_check_slp_defs): Take the vector of statements
13817 rather than just the current one. Remove excess parentheses.
13818 Restriction rejectinon of vect_constant_def and vect_external_def
13819 for variable-length vectors to boolean types, or types for which
13820 can_duplicate_and_interleave_p is false.
13821 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13822 (duplicate_and_interleave): New function.
13823 (vect_get_constant_vectors): Use gimple_build_vector for
13824 constant-length vectors and suitable variable-length constant
13825 vectors. Use duplicate_and_interleave for other variable-length
13826 vectors. Don't defer the update when inserting new statements.
13828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13833 min_profitable_iters doesn't go negative.
13835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13836 Alan Hayward <alan.hayward@arm.com>
13837 David Sherwood <david.sherwood@arm.com>
13839 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13840 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13841 * optabs.def (vec_mask_load_lanes_optab): New optab.
13842 (vec_mask_store_lanes_optab): Likewise.
13843 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13844 (MASK_STORE_LANES): Likewise.
13845 * internal-fn.c (mask_load_lanes_direct): New macro.
13846 (mask_store_lanes_direct): Likewise.
13847 (expand_mask_load_optab_fn): Handle masked operations.
13848 (expand_mask_load_lanes_optab_fn): New macro.
13849 (expand_mask_store_optab_fn): Handle masked operations.
13850 (expand_mask_store_lanes_optab_fn): New macro.
13851 (direct_mask_load_lanes_optab_supported_p): Likewise.
13852 (direct_mask_store_lanes_optab_supported_p): Likewise.
13853 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13855 (vect_load_lanes_supported): Likewise.
13856 * tree-vect-data-refs.c (strip_conversion): New function.
13857 (can_group_stmts_p): Likewise.
13858 (vect_analyze_data_ref_accesses): Use it instead of checking
13859 for a pair of assignments.
13860 (vect_store_lanes_supported): Take a masked_p parameter.
13861 (vect_load_lanes_supported): Likewise.
13862 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13863 vect_store_lanes_supported and vect_load_lanes_supported.
13864 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13865 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13866 parameter. Don't allow gaps for masked accesses.
13867 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13868 and vect_load_lanes_supported.
13869 (get_load_store_type): Take a masked_p parameter and update
13870 call to get_group_load_store_type.
13871 (vectorizable_store): Update call to get_load_store_type.
13872 Handle IFN_MASK_STORE_LANES.
13873 (vectorizable_load): Update call to get_load_store_type.
13874 Handle IFN_MASK_LOAD_LANES.
13876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13877 Alan Hayward <alan.hayward@arm.com>
13878 David Sherwood <david.sherwood@arm.com>
13880 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13882 * config/aarch64/aarch64-protos.h
13883 (aarch64_sve_struct_memory_operand_p): Declare.
13884 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13885 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13886 (VPRED, vpred): Handle SVE structure modes.
13887 * config/aarch64/constraints.md (Utx): New constraint.
13888 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13889 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13890 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13891 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13892 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13893 structure modes. Split into pieces after RA.
13894 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13895 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13897 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13898 SVE structure modes.
13899 (aarch64_classify_address): Likewise.
13900 (sizetochar): Move earlier in file.
13901 (aarch64_print_operand): Handle SVE register lists.
13902 (aarch64_array_mode): New function.
13903 (aarch64_sve_struct_memory_operand_p): Likewise.
13904 (TARGET_ARRAY_MODE): Redefine.
13906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13907 Alan Hayward <alan.hayward@arm.com>
13908 David Sherwood <david.sherwood@arm.com>
13910 * target.def (array_mode): New target hook.
13911 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13912 * doc/tm.texi: Regenerate.
13913 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13914 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13915 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13916 targetm.array_mode.
13917 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13921 Alan Hayward <alan.hayward@arm.com>
13922 David Sherwood <david.sherwood@arm.com>
13924 * fold-const.c (fold_binary_loc): Check the argument types
13925 rather than the result type when testing for a vector operation.
13927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13929 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13930 * doc/tm.texi: Regenerate.
13932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13937 (sve): Document new AArch64 extension.
13938 * doc/md.texi (w): Extend the description of the AArch64
13939 constraint to include SVE vectors.
13940 (Upl, Upa): Document new AArch64 predicate constraints.
13941 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13943 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13944 (msve-vector-bits=): New option.
13945 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13946 SVE when these are disabled.
13947 (sve): New extension.
13948 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13949 modes. Adjust their number of units based on aarch64_sve_vg.
13950 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13951 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13952 aarch64_addr_query_type.
13953 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13954 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13955 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13956 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13957 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13958 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13959 (aarch64_simd_imm_zero_p): Delete.
13960 (aarch64_check_zero_based_sve_index_immediate): Declare.
13961 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13962 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13963 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13964 (aarch64_sve_float_mul_immediate_p): Likewise.
13965 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13966 rather than an rtx.
13967 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13968 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13969 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13970 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13971 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13972 (aarch64_regmode_natural_size): Likewise.
13973 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13974 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13976 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13977 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13978 for VG and the SVE predicate registers.
13979 (V_ALIASES): Add a "z"-prefixed alias.
13980 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13981 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13982 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13983 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13984 (REG_CLASS_NAMES): Add entries for them.
13985 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13986 and the predicate registers.
13987 (aarch64_sve_vg): Declare.
13988 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13989 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13990 (REGMODE_NATURAL_SIZE): Define.
13991 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13993 * config/aarch64/aarch64.c: Include cfgrtl.h.
13994 (simd_immediate_info): Add a constructor for series vectors,
13995 and an associated step field.
13996 (aarch64_sve_vg): New variable.
13997 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13998 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13999 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14000 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14001 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14002 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14003 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14004 (aarch64_get_mask_mode): New functions.
14005 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14006 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14007 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14008 predicate modes and predicate registers. Explicitly restrict
14009 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14010 to store a vector mode if it is recognized by
14011 aarch64_classify_vector_mode.
14012 (aarch64_regmode_natural_size): New function.
14013 (aarch64_hard_regno_caller_save_mode): Return the original mode
14015 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14016 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14017 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14018 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14020 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14021 does not overlap dest if the function is frame-related. Handle
14023 (aarch64_split_add_offset): New function.
14024 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14025 them aarch64_add_offset.
14026 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14027 and update call to aarch64_sub_sp.
14028 (aarch64_add_cfa_expression): New function.
14029 (aarch64_expand_prologue): Pass extra temporary registers to the
14030 functions above. Handle the case in which we need to emit new
14031 DW_CFA_expressions for registers that were originally saved
14032 relative to the stack pointer, but now have to be expressed
14033 relative to the frame pointer.
14034 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14036 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14037 IP0 and IP1 values for SVE frames.
14038 (aarch64_expand_vec_series): New function.
14039 (aarch64_expand_sve_widened_duplicate): Likewise.
14040 (aarch64_expand_sve_const_vector): Likewise.
14041 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14042 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14043 into the register, rather than emitting a SET directly.
14044 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14045 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14046 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14047 (offset_9bit_signed_scaled_p): New functions.
14048 (aarch64_replicate_bitmask_imm): New function.
14049 (aarch64_bitmask_imm): Use it.
14050 (aarch64_cannot_force_const_mem): Reject expressions involving
14051 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14052 (aarch64_classify_index): Handle SVE indices, by requiring
14053 a plain register index with a scale that matches the element size.
14054 (aarch64_classify_address): Handle SVE addresses. Assert that
14055 the mode of the address is VOIDmode or an integer mode.
14056 Update call to aarch64_classify_symbol.
14057 (aarch64_classify_symbolic_expression): Update call to
14058 aarch64_classify_symbol.
14059 (aarch64_const_vec_all_in_range_p): New function.
14060 (aarch64_print_vector_float_operand): Likewise.
14061 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14062 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14063 and the FP immediates 1.0 and 0.5.
14064 (aarch64_print_address_internal): Handle SVE addresses.
14065 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14066 (aarch64_regno_regclass): Handle predicate registers.
14067 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14069 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14070 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14071 (aarch64_convert_sve_vector_bits): New function.
14072 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14073 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14074 rather than an rtx.
14075 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14076 Handle SVE vector and predicate modes. Accept VL-based constants
14077 that need only one temporary register, and VL offsets that require
14078 no temporary registers.
14079 (aarch64_conditional_register_usage): Mark the predicate registers
14080 as fixed if SVE isn't available.
14081 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14082 Return true for SVE vector and predicate modes.
14083 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14084 rather than an unsigned int. Handle SVE modes.
14085 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14087 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14089 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14090 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14091 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14092 (aarch64_sve_float_mul_immediate_p): New functions.
14093 (aarch64_sve_valid_immediate): New function.
14094 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14095 Explicitly reject structure modes. Check for INDEX constants.
14096 Handle PTRUE and PFALSE constants.
14097 (aarch64_check_zero_based_sve_index_immediate): New function.
14098 (aarch64_simd_imm_zero_p): Delete.
14099 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14100 vector modes. Accept constants in the range of CNT[BHWD].
14101 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14102 ask for an Advanced SIMD mode.
14103 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14104 (aarch64_simd_vector_alignment): Handle SVE predicates.
14105 (aarch64_vectorize_preferred_vector_alignment): New function.
14106 (aarch64_simd_vector_alignment_reachable): Use it instead of
14108 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14109 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14111 (MAX_VECT_LEN): Delete.
14112 (expand_vec_perm_d): Add a vec_flags field.
14113 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14114 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14115 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14117 (aarch64_evpc_rev): Rename to...
14118 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14119 (aarch64_evpc_rev_global): New function.
14120 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14121 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14123 (aarch64_evpc_sve_tbl): New function.
14124 (aarch64_expand_vec_perm_const_1): Update after rename of
14125 aarch64_evpc_rev. Handle SVE permutes too, trying
14126 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14127 than aarch64_evpc_tbl.
14128 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14129 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14130 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14131 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14132 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14133 (aarch64_expand_sve_vcond): New functions.
14134 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14135 of aarch64_vector_mode_p.
14136 (aarch64_dwarf_poly_indeterminate_value): New function.
14137 (aarch64_compute_pressure_classes): Likewise.
14138 (aarch64_can_change_mode_class): Likewise.
14139 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14140 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14141 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14142 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14143 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14144 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14145 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14146 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14148 (Dn, Dl, Dr): Accept const as well as const_vector.
14149 (Dz): Likewise. Compare against CONST0_RTX.
14150 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14151 of "vector" where appropriate.
14152 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14153 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14154 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14155 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14156 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14157 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14158 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14159 (v_int_equiv): Extend to SVE modes.
14160 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14162 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14163 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14164 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14165 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14166 (SVE_COND_FP_CMP): New int iterators.
14167 (perm_hilo): Handle the new unpack unspecs.
14168 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14170 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14171 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14172 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14173 (aarch64_equality_operator, aarch64_constant_vector_operand)
14174 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14175 (aarch64_sve_nonimmediate_operand): Likewise.
14176 (aarch64_sve_general_operand): Likewise.
14177 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14178 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14179 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14180 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14181 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14182 (aarch64_sve_float_arith_immediate): Likewise.
14183 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14184 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14185 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14186 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14187 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14188 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14189 (aarch64_sve_float_arith_operand): Likewise.
14190 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14191 (aarch64_sve_float_mul_operand): Likewise.
14192 (aarch64_sve_vec_perm_operand): Likewise.
14193 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14194 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14195 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14196 as well as const_vector.
14197 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14198 in file. Use CONST0_RTX and CONSTM1_RTX.
14199 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14200 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14201 Use aarch64_simd_imm_zero.
14202 * config/aarch64/aarch64-sve.md: New file.
14203 * config/aarch64/aarch64.md: Include it.
14204 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14205 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14206 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14207 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14208 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14209 (sve): New attribute.
14210 (enabled): Disable instructions with the sve attribute unless
14212 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14213 aarch64_expand_mov_immediate.
14214 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14215 CNT[BHSD] immediates.
14216 (movti): Split CONST_POLY_INT moves into two halves.
14217 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14218 Split additions that need a temporary here if the destination
14219 is the stack pointer.
14220 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14221 (*add<mode>3_poly_1): New instruction.
14222 (set_clobber_cc): New expander.
14224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14226 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14227 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14228 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14229 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14230 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14231 Change innermode from fixed_mode_size to machine_mode.
14232 (simplify_subreg): Update call accordingly. Handle a constant-sized
14233 subreg of a variable-length CONST_VECTOR.
14235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14236 Alan Hayward <alan.hayward@arm.com>
14237 David Sherwood <david.sherwood@arm.com>
14239 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14240 (add_offset_to_base): New function, split out from...
14241 (create_mem_ref): ...here. When handling a scale other than 1,
14242 check first whether the address is valid without the offset.
14243 Add it into the base if so, leaving the index and scale as-is.
14245 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14248 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14249 fold_for_warn before checking if arg2 is INTEGER_CST.
14251 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14253 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14254 (store_multiple_operation): Delete.
14255 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14256 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14257 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14258 guarded by TARGET_STRING.
14259 (rs6000_output_load_multiple): Delete.
14260 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14261 OPTION_MASK_STRING / TARGET_STRING handling.
14262 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14263 (const rs6000_opt_masks) <"string">: Change mask to 0.
14264 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14265 (MASK_STRING): Delete.
14266 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14268 (load_multiple): Delete.
14275 (store_multiple): Delete.
14282 (movmemsi_8reg): Delete.
14283 (corresponding unnamed define_insn): Delete.
14284 (movmemsi_6reg): Delete.
14285 (corresponding unnamed define_insn): Delete.
14286 (movmemsi_4reg): Delete.
14287 (corresponding unnamed define_insn): Delete.
14288 (movmemsi_2reg): Delete.
14289 (corresponding unnamed define_insn): Delete.
14290 (movmemsi_1reg): Delete.
14291 (corresponding unnamed define_insn): Delete.
14292 * config/rs6000/rs6000.opt (mno-string): New.
14293 (mstring): Replace by deprecation warning stub.
14294 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14296 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14298 * regrename.c (regrename_do_replace): If replacing the same
14299 reg multiple times, try to reuse last created gen_raw_REG.
14302 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14303 main to workaround a bug in GDB.
14305 2018-01-12 Tom de Vries <tom@codesourcery.com>
14308 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14310 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14312 PR rtl-optimization/80481
14313 * ira-color.c (get_cap_member): New function.
14314 (allocnos_conflict_by_live_ranges_p): Use it.
14315 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14316 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14318 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14321 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14322 (*saddl_se_1): Ditto.
14323 (*ssubsi_1): Ditto.
14324 (*ssubl_se_1): Ditto.
14326 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14328 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14329 rather than wi::to_widest for DR_INITs.
14330 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14331 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14332 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14334 (vect_analyze_group_access_1): Note that here.
14336 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14338 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14339 polynomial type sizes.
14341 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14343 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14344 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14345 (gimple_add_tmp_var): Likewise.
14347 2018-01-12 Martin Liska <mliska@suse.cz>
14349 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14350 (gimple_alloc_sizes): Likewise.
14351 (dump_gimple_statistics): Use PRIu64 in printf format.
14352 * gimple.h: Change uint64_t to int.
14354 2018-01-12 Martin Liska <mliska@suse.cz>
14356 * tree-core.h: Use uint64_t instead of int.
14357 * tree.c (tree_node_counts): Likewise.
14358 (tree_node_sizes): Likewise.
14359 (dump_tree_statistics): Use PRIu64 in printf format.
14361 2018-01-12 Martin Liska <mliska@suse.cz>
14363 * Makefile.in: As qsort_chk is implemented in vec.c, add
14364 vec.o to linkage of gencfn-macros.
14365 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14366 passing the info to record_node_allocation_statistics.
14367 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14369 * ggc-common.c (struct ggc_usage): Add operator== and use
14370 it in operator< and compare function.
14371 * mem-stats.h (struct mem_usage): Likewise.
14372 * vec.c (struct vec_usage): Remove operator< and compare
14373 function. Can be simply inherited.
14375 2018-01-12 Martin Jambor <mjambor@suse.cz>
14378 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14379 * tree-ssa-math-opts.c: Include domwalk.h.
14380 (convert_mult_to_fma_1): New function.
14381 (fma_transformation_info): New type.
14382 (fma_deferring_state): Likewise.
14383 (cancel_fma_deferring): New function.
14384 (result_of_phi): Likewise.
14385 (last_fma_candidate_feeds_initial_phi): Likewise.
14386 (convert_mult_to_fma): Added deferring logic, split actual
14387 transformation to convert_mult_to_fma_1.
14388 (math_opts_dom_walker): New type.
14389 (math_opts_dom_walker::after_dom_children): New method, body moved
14390 here from pass_optimize_widening_mul::execute, added deferring logic
14392 (pass_optimize_widening_mul::execute): Moved most of code to
14393 math_opts_dom_walker::after_dom_children.
14394 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14395 * config/i386/i386.c (ix86_option_override_internal): Added
14396 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14398 2018-01-12 Richard Biener <rguenther@suse.de>
14401 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14402 inline instance vars.
14404 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14407 * config/rx/rx.c (rx_is_restricted_memory_address):
14408 Handle SUBREG case.
14410 2018-01-12 Richard Biener <rguenther@suse.de>
14412 PR tree-optimization/80846
14413 * target.def (split_reduction): New target hook.
14414 * targhooks.c (default_split_reduction): New function.
14415 * targhooks.h (default_split_reduction): Declare.
14416 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14417 target requests first reduce vectors by combining low and high
14419 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14420 (get_vectype_for_scalar_type_and_size): Export.
14421 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14422 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14423 * doc/tm.texi: Regenerate.
14424 * config/i386/i386.c (ix86_split_reduction): Implement
14425 TARGET_VECTORIZE_SPLIT_REDUCTION.
14427 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14430 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14431 in PIC mode except for TARGET_VXWORKS_RTP.
14432 * config/sparc/sparc.c: Include cfgrtl.h.
14433 (TARGET_INIT_PIC_REG): Define.
14434 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14435 (sparc_pic_register_p): New predicate.
14436 (sparc_legitimate_address_p): Use it.
14437 (sparc_legitimize_pic_address): Likewise.
14438 (sparc_delegitimize_address): Likewise.
14439 (sparc_mode_dependent_address_p): Likewise.
14440 (gen_load_pcrel_sym): Remove 4th parameter.
14441 (load_got_register): Adjust call to above. Remove obsolete stuff.
14442 (sparc_expand_prologue): Do not call load_got_register here.
14443 (sparc_flat_expand_prologue): Likewise.
14444 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14445 (sparc_use_pseudo_pic_reg): New function.
14446 (sparc_init_pic_reg): Likewise.
14447 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14448 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14450 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14452 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14453 Add item for branch_cost.
14455 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14457 PR rtl-optimization/83565
14458 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14459 not extend the result to a larger mode for rotate operations.
14460 (num_sign_bit_copies1): Likewise.
14462 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14465 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14467 Use values-Xc.o for -pedantic.
14468 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14470 2018-01-12 Martin Liska <mliska@suse.cz>
14473 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14475 (possible_polymorphic_call_targets): Use it.
14476 (ipa_devirt): Likewise.
14478 2018-01-12 Martin Liska <mliska@suse.cz>
14480 * profile-count.h (enum profile_quality): Use 0 as invalid
14481 enum value of profile_quality.
14483 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14485 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14486 -mext-string options.
14488 2018-01-12 Richard Biener <rguenther@suse.de>
14490 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14491 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14492 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14494 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14496 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14498 * configure.ac (--with-long-double-format): Add support for the
14499 configuration option to change the default long double format on
14501 * config.gcc (powerpc*-linux*-*): Likewise.
14502 * configure: Regenerate.
14503 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14504 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14505 used without modification.
14507 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14509 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14510 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14511 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14512 MISC_BUILTIN_SPEC_BARRIER.
14513 (rs6000_init_builtins): Likewise.
14514 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14516 (speculation_barrier): New define_insn.
14517 * doc/extend.texi: Document __builtin_speculation_barrier.
14519 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14522 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14523 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14524 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14526 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14527 integral modes instead of "ss" and "sd".
14528 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14529 vectors with 32-bit and 64-bit elements.
14530 (vecdupssescalarmodesuffix): New mode attribute.
14531 (vec_dup<mode>): Use it.
14533 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14536 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14537 frame if argument is passed on stack.
14539 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14542 * ree.c (combine_reaching_defs): Optimize also
14543 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14544 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14546 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14548 PR middle-end/83189
14549 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14551 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14553 PR middle-end/83718
14554 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14555 after they are computed.
14557 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14559 PR tree-optimization/83695
14560 * gimple-loop-linterchange.cc
14561 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14562 reset cached scev information after interchange.
14563 (pass_linterchange::execute): Remove call to scev_reset_htab.
14565 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14567 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14568 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14569 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14570 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14571 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14572 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14573 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14574 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14575 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14576 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14577 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14578 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14579 (V_lane_reg): Likewise.
14580 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14582 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14583 (vfmal_lane_low<mode>_intrinsic,
14584 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14585 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14586 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14587 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14588 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14589 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14591 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14593 * config/arm/arm-cpus.in (fp16fml): New feature.
14594 (ALL_SIMD): Add fp16fml.
14595 (armv8.2-a): Add fp16fml as an option.
14596 (armv8.3-a): Likewise.
14597 (armv8.4-a): Add fp16fml as part of fp16.
14598 * config/arm/arm.h (TARGET_FP16FML): Define.
14599 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14601 * config/arm/arm-modes.def (V2HF): Define.
14602 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14603 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14604 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14605 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14606 vfmsl_low, vfmsl_high): New set of builtins.
14607 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14608 (vfml_op): New code attribute.
14609 (VFMLHALVES): New int iterator.
14610 (VFML, VFMLSEL): New mode attributes.
14611 (V_reg): Define mapping for V2HF.
14612 (V_hi, V_lo): New mode attributes.
14613 (VF_constraint): Likewise.
14614 (vfml_half, vfml_half_selector): New int attributes.
14615 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14617 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14618 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14620 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14621 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14622 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14623 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14625 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14626 Document new effective target and option set.
14628 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14630 * config/arm/arm-cpus.in (armv8_4): New feature.
14631 (ARMv8_4a): New fgroup.
14632 (armv8.4-a): New arch.
14633 * config/arm/arm-tables.opt: Regenerate.
14634 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14635 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14636 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14637 Add matching rules for -march=armv8.4-a and extensions.
14638 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14640 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14643 * config/rx/rx.md (BW): New mode attribute.
14644 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14646 2018-01-11 Richard Biener <rguenther@suse.de>
14648 PR tree-optimization/83435
14649 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14650 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14651 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14653 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14654 Alan Hayward <alan.hayward@arm.com>
14655 David Sherwood <david.sherwood@arm.com>
14657 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14659 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14660 (aarch64_print_address_internal): Use it to check for a zero offset.
14662 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14663 Alan Hayward <alan.hayward@arm.com>
14664 David Sherwood <david.sherwood@arm.com>
14666 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14667 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14668 Return a poly_int64 rather than a HOST_WIDE_INT.
14669 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14670 rather than a HOST_WIDE_INT.
14671 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14672 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14673 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14674 final_offset from HOST_WIDE_INT to poly_int64.
14675 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14676 to_constant when getting the number of units in an Advanced SIMD
14678 (aarch64_builtin_vectorized_function): Check for a constant number
14680 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14682 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14683 attribute instead of GET_MODE_NUNITS.
14684 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14685 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14686 GET_MODE_SIZE for fixed-size registers.
14687 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14688 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14689 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14690 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14691 (aarch64_print_operand, aarch64_print_address_internal)
14692 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14693 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14694 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14695 Handle polynomial GET_MODE_SIZE.
14696 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14697 wider than SImode without modification.
14698 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14699 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14700 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14701 passing and returning SVE modes.
14702 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14703 rather than GEN_INT.
14704 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14705 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14706 (aarch64_allocate_and_probe_stack_space): Likewise.
14707 (aarch64_layout_frame): Cope with polynomial offsets.
14708 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14709 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14710 polynomial offsets.
14711 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14712 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14713 poly_int64 rather than a HOST_WIDE_INT.
14714 (aarch64_get_separate_components, aarch64_process_components)
14715 (aarch64_expand_prologue, aarch64_expand_epilogue)
14716 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14717 (aarch64_anchor_offset): New function, split out from...
14718 (aarch64_legitimize_address): ...here.
14719 (aarch64_builtin_vectorization_cost): Handle polynomial
14720 TYPE_VECTOR_SUBPARTS.
14721 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14723 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14724 number of elements from the PARALLEL rather than the mode.
14725 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14726 rather than GET_MODE_BITSIZE.
14727 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14728 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14729 (aarch64_expand_vec_perm_const_1): Handle polynomial
14730 d->perm.length () and d->perm elements.
14731 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14732 Apply to_constant to d->perm elements.
14733 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14734 polynomial CONST_VECTOR_NUNITS.
14735 (aarch64_move_pointer): Take amount as a poly_int64 rather
14737 (aarch64_progress_pointer): Avoid temporary variable.
14738 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14739 the mode attribute instead of GET_MODE.
14741 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14742 Alan Hayward <alan.hayward@arm.com>
14743 David Sherwood <david.sherwood@arm.com>
14745 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14746 x exists before using it.
14747 (aarch64_add_constant_internal): Rename to...
14748 (aarch64_add_offset_1): ...this. Replace regnum with separate
14749 src and dest rtxes. Handle the case in which they're different,
14750 including when the offset is zero. Replace scratchreg with an rtx.
14751 Use 2 additions if there is no spare register into which we can
14752 move a 16-bit constant.
14753 (aarch64_add_constant): Delete.
14754 (aarch64_add_offset): Replace reg with separate src and dest
14755 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14756 Use aarch64_add_offset_1.
14757 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14758 an rtx rather than an int. Take the delta as a poly_int64
14759 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14760 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14761 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14762 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14763 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14764 and aarch64_add_sp.
14765 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14766 aarch64_add_constant.
14768 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14770 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14771 Use scalar_float_mode.
14773 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14775 * config/aarch64/aarch64-simd.md
14776 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14777 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14778 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14779 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14780 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14781 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14782 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14783 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14784 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14785 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14787 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14790 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14791 targ_options->x_arm_arch_string is non NULL.
14793 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14795 * config/aarch64/aarch64.h
14796 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14798 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14801 * expmed.c (emit_store_flag_force): Swap if const op0
14802 and change VOIDmode to mode of op0.
14804 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14806 PR rtl-optimization/83761
14807 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14808 than bytes to mode_for_size.
14810 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14812 PR middle-end/83189
14813 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14814 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14817 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14819 PR middle-end/83575
14820 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14821 when in layout mode.
14822 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14823 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14826 2018-01-10 Michael Collison <michael.collison@arm.com>
14828 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14829 * config/aarch64/aarch64-option-extension.def: Add
14830 AARCH64_OPT_EXTENSION of 'fp16fml'.
14831 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14832 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14833 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14834 * config/aarch64/constraints.md (Ui7): New constraint.
14835 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14836 (VFMLA_SEL_W): Ditto.
14839 (VFMLA16_LOW): New int iterator.
14840 (VFMLA16_HIGH): Ditto.
14841 (UNSPEC_FMLAL): New unspec.
14842 (UNSPEC_FMLSL): Ditto.
14843 (UNSPEC_FMLAL2): Ditto.
14844 (UNSPEC_FMLSL2): Ditto.
14845 (f16mac): New code attribute.
14846 * config/aarch64/aarch64-simd-builtins.def
14847 (aarch64_fmlal_lowv2sf): Ditto.
14848 (aarch64_fmlsl_lowv2sf): Ditto.
14849 (aarch64_fmlalq_lowv4sf): Ditto.
14850 (aarch64_fmlslq_lowv4sf): Ditto.
14851 (aarch64_fmlal_highv2sf): Ditto.
14852 (aarch64_fmlsl_highv2sf): Ditto.
14853 (aarch64_fmlalq_highv4sf): Ditto.
14854 (aarch64_fmlslq_highv4sf): Ditto.
14855 (aarch64_fmlal_lane_lowv2sf): Ditto.
14856 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14857 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14858 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14859 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14860 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14861 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14862 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14863 (aarch64_fmlal_lane_highv2sf): Ditto.
14864 (aarch64_fmlsl_lane_highv2sf): Ditto.
14865 (aarch64_fmlal_laneq_highv2sf): Ditto.
14866 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14867 (aarch64_fmlalq_lane_highv4sf): Ditto.
14868 (aarch64_fmlsl_lane_highv4sf): Ditto.
14869 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14870 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14871 * config/aarch64/aarch64-simd.md:
14872 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14873 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14874 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14875 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14876 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14877 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14878 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14879 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14880 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14881 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14882 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14883 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14884 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14885 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14886 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14887 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14888 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14889 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14890 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14891 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14892 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14893 (vfmlsl_low_u32): Ditto.
14894 (vfmlalq_low_u32): Ditto.
14895 (vfmlslq_low_u32): Ditto.
14896 (vfmlal_high_u32): Ditto.
14897 (vfmlsl_high_u32): Ditto.
14898 (vfmlalq_high_u32): Ditto.
14899 (vfmlslq_high_u32): Ditto.
14900 (vfmlal_lane_low_u32): Ditto.
14901 (vfmlsl_lane_low_u32): Ditto.
14902 (vfmlal_laneq_low_u32): Ditto.
14903 (vfmlsl_laneq_low_u32): Ditto.
14904 (vfmlalq_lane_low_u32): Ditto.
14905 (vfmlslq_lane_low_u32): Ditto.
14906 (vfmlalq_laneq_low_u32): Ditto.
14907 (vfmlslq_laneq_low_u32): Ditto.
14908 (vfmlal_lane_high_u32): Ditto.
14909 (vfmlsl_lane_high_u32): Ditto.
14910 (vfmlal_laneq_high_u32): Ditto.
14911 (vfmlsl_laneq_high_u32): Ditto.
14912 (vfmlalq_lane_high_u32): Ditto.
14913 (vfmlslq_lane_high_u32): Ditto.
14914 (vfmlalq_laneq_high_u32): Ditto.
14915 (vfmlslq_laneq_high_u32): Ditto.
14916 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14917 (AARCH64_FL_FOR_ARCH8_4): New.
14918 (AARCH64_ISA_F16FML): New ISA flag.
14919 (TARGET_F16FML): New feature flag for fp16fml.
14920 (doc/invoke.texi): Document new fp16fml option.
14922 2018-01-10 Michael Collison <michael.collison@arm.com>
14924 * config/aarch64/aarch64-builtins.c:
14925 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14926 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14927 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14928 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14929 (AARCH64_ISA_SHA3): New ISA flag.
14930 (TARGET_SHA3): New feature flag for sha3.
14931 * config/aarch64/iterators.md (sha512_op): New int attribute.
14932 (CRYPTO_SHA512): New int iterator.
14933 (UNSPEC_SHA512H): New unspec.
14934 (UNSPEC_SHA512H2): Ditto.
14935 (UNSPEC_SHA512SU0): Ditto.
14936 (UNSPEC_SHA512SU1): Ditto.
14937 * config/aarch64/aarch64-simd-builtins.def
14938 (aarch64_crypto_sha512hqv2di): New builtin.
14939 (aarch64_crypto_sha512h2qv2di): Ditto.
14940 (aarch64_crypto_sha512su0qv2di): Ditto.
14941 (aarch64_crypto_sha512su1qv2di): Ditto.
14942 (aarch64_eor3qv8hi): Ditto.
14943 (aarch64_rax1qv2di): Ditto.
14944 (aarch64_xarqv2di): Ditto.
14945 (aarch64_bcaxqv8hi): Ditto.
14946 * config/aarch64/aarch64-simd.md:
14947 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14948 (aarch64_crypto_sha512su0qv2di): Ditto.
14949 (aarch64_crypto_sha512su1qv2di): Ditto.
14950 (aarch64_eor3qv8hi): Ditto.
14951 (aarch64_rax1qv2di): Ditto.
14952 (aarch64_xarqv2di): Ditto.
14953 (aarch64_bcaxqv8hi): Ditto.
14954 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14955 (vsha512h2q_u64): Ditto.
14956 (vsha512su0q_u64): Ditto.
14957 (vsha512su1q_u64): Ditto.
14958 (veor3q_u16): Ditto.
14959 (vrax1q_u64): Ditto.
14960 (vxarq_u64): Ditto.
14961 (vbcaxq_u16): Ditto.
14962 * config/arm/types.md (crypto_sha512): New type attribute.
14963 (crypto_sha3): Ditto.
14964 (doc/invoke.texi): Document new sha3 option.
14966 2018-01-10 Michael Collison <michael.collison@arm.com>
14968 * config/aarch64/aarch64-builtins.c:
14969 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14970 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14971 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14972 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14973 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14974 (AARCH64_ISA_SM4): New ISA flag.
14975 (TARGET_SM4): New feature flag for sm4.
14976 * config/aarch64/aarch64-simd-builtins.def
14977 (aarch64_sm3ss1qv4si): Ditto.
14978 (aarch64_sm3tt1aq4si): Ditto.
14979 (aarch64_sm3tt1bq4si): Ditto.
14980 (aarch64_sm3tt2aq4si): Ditto.
14981 (aarch64_sm3tt2bq4si): Ditto.
14982 (aarch64_sm3partw1qv4si): Ditto.
14983 (aarch64_sm3partw2qv4si): Ditto.
14984 (aarch64_sm4eqv4si): Ditto.
14985 (aarch64_sm4ekeyqv4si): Ditto.
14986 * config/aarch64/aarch64-simd.md:
14987 (aarch64_sm3ss1qv4si): Ditto.
14988 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14989 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14990 (aarch64_sm4eqv4si): Ditto.
14991 (aarch64_sm4ekeyqv4si): Ditto.
14992 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14993 (sm3part_op): Ditto.
14994 (CRYPTO_SM3TT): Ditto.
14995 (CRYPTO_SM3PART): Ditto.
14996 (UNSPEC_SM3SS1): New unspec.
14997 (UNSPEC_SM3TT1A): Ditto.
14998 (UNSPEC_SM3TT1B): Ditto.
14999 (UNSPEC_SM3TT2A): Ditto.
15000 (UNSPEC_SM3TT2B): Ditto.
15001 (UNSPEC_SM3PARTW1): Ditto.
15002 (UNSPEC_SM3PARTW2): Ditto.
15003 (UNSPEC_SM4E): Ditto.
15004 (UNSPEC_SM4EKEY): Ditto.
15005 * config/aarch64/constraints.md (Ui2): New constraint.
15006 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15007 * config/arm/types.md (crypto_sm3): New type attribute.
15008 (crypto_sm4): Ditto.
15009 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15010 (vsm3tt1aq_u32): Ditto.
15011 (vsm3tt1bq_u32): Ditto.
15012 (vsm3tt2aq_u32): Ditto.
15013 (vsm3tt2bq_u32): Ditto.
15014 (vsm3partw1q_u32): Ditto.
15015 (vsm3partw2q_u32): Ditto.
15016 (vsm4eq_u32): Ditto.
15017 (vsm4ekeyq_u32): Ditto.
15018 (doc/invoke.texi): Document new sm4 option.
15020 2018-01-10 Michael Collison <michael.collison@arm.com>
15022 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15023 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15024 (AARCH64_FL_FOR_ARCH8_4): New.
15025 (AARCH64_FL_V8_4): New flag.
15026 (doc/invoke.texi): Document new armv8.4-a option.
15028 2018-01-10 Michael Collison <michael.collison@arm.com>
15030 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15031 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15032 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15033 * config/aarch64/aarch64-option-extension.def: Add
15034 AARCH64_OPT_EXTENSION of 'sha2'.
15035 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15036 (crypto): Disable sha2 and aes if crypto disabled.
15037 (crypto): Enable aes and sha2 if enabled.
15038 (simd): Disable sha2 and aes if simd disabled.
15039 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15041 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15042 (TARGET_SHA2): New feature flag for sha2.
15043 (TARGET_AES): New feature flag for aes.
15044 * config/aarch64/aarch64-simd.md:
15045 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15046 conditional on TARGET_AES.
15047 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15048 (aarch64_crypto_sha1hsi): Make pattern conditional
15050 (aarch64_crypto_sha1hv4si): Ditto.
15051 (aarch64_be_crypto_sha1hv4si): Ditto.
15052 (aarch64_crypto_sha1su1v4si): Ditto.
15053 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15054 (aarch64_crypto_sha1su0v4si): Ditto.
15055 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15056 (aarch64_crypto_sha256su0v4si): Ditto.
15057 (aarch64_crypto_sha256su1v4si): Ditto.
15058 (doc/invoke.texi): Document new aes and sha2 options.
15060 2018-01-10 Martin Sebor <msebor@redhat.com>
15062 PR tree-optimization/83781
15063 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15066 2018-01-11 Martin Sebor <msebor@gmail.com>
15067 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15069 PR tree-optimization/83501
15070 PR tree-optimization/81703
15072 * tree-ssa-strlen.c (get_string_cst): Rename...
15073 (get_string_len): ...to this. Handle global constants.
15074 (handle_char_store): Adjust.
15076 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15077 Jim Wilson <jimw@sifive.com>
15079 * config/riscv/riscv-protos.h (riscv_output_return): New.
15080 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15081 (riscv_attribute_table, riscv_output_return),
15082 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15083 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15084 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15086 (riscv_expand_prologue): Add early return for naked function.
15087 (riscv_expand_epilogue): Likewise.
15088 (riscv_function_ok_for_sibcall): Return false for naked function.
15089 (riscv_set_current_function): New.
15090 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15091 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15092 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15093 * doc/extend.texi (RISC-V Function Attributes): New.
15095 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15097 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15098 check for 128-bit long double before checking TCmode.
15099 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15100 128-bit long doubles before checking TFmode or TCmode.
15101 (FLOAT128_IBM_P): Likewise.
15103 2018-01-10 Martin Sebor <msebor@redhat.com>
15105 PR tree-optimization/83671
15106 * builtins.c (c_strlen): Unconditionally return zero for the empty
15108 Use -Warray-bounds for warnings.
15109 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15110 for non-constant array indices with COMPONENT_REF, arrays of
15111 arrays, and pointers to arrays.
15112 (gimple_fold_builtin_strlen): Determine and set length range for
15113 non-constant character arrays.
15115 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15117 PR middle-end/81897
15118 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15121 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15123 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15125 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15128 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15129 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15130 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15131 indexed_or_indirect_operand predicate.
15132 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15133 (*vsx_le_perm_load_v8hi): Likewise.
15134 (*vsx_le_perm_load_v16qi): Likewise.
15135 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15136 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15137 (*vsx_le_perm_store_v8hi): Likewise.
15138 (*vsx_le_perm_store_v16qi): Likewise.
15139 (eight unnamed splitters): Likewise.
15141 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15143 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15144 * config/rs6000/emmintrin.h: Likewise.
15145 * config/rs6000/mmintrin.h: Likewise.
15146 * config/rs6000/xmmintrin.h: Likewise.
15148 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15151 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15153 * tree.c (tree_nop_conversion): Return true for location wrapper
15155 (maybe_wrap_with_location): New function.
15156 (selftest::check_strip_nops): New function.
15157 (selftest::test_location_wrappers): New function.
15158 (selftest::tree_c_tests): Call it.
15159 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15160 (maybe_wrap_with_location): New decl.
15161 (EXPR_LOCATION_WRAPPER_P): New macro.
15162 (location_wrapper_p): New inline function.
15163 (tree_strip_any_location_wrapper): New inline function.
15165 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15168 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15169 stack_realign_offset for the largest alignment of stack slot
15171 (ix86_find_max_used_stack_alignment): New function.
15172 (ix86_finalize_stack_frame_flags): Use it. Set
15173 max_used_stack_alignment if we don't realign stack.
15174 * config/i386/i386.h (machine_function): Add
15175 max_used_stack_alignment.
15177 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15179 * config/arm/arm.opt (-mbranch-cost): New option.
15180 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15183 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15186 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15187 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15189 2018-01-10 Richard Biener <rguenther@suse.de>
15192 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15193 early out so it also covers the case where we have a non-NULL
15196 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15198 PR tree-optimization/83753
15199 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15200 for non-strided grouped accesses if the number of elements is 1.
15202 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15205 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15206 * i386.h (TARGET_USE_GATHER): Define.
15207 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15209 2018-01-10 Martin Liska <mliska@suse.cz>
15212 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15213 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15215 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15216 CLEANUP_NO_PARTITIONING is not set.
15218 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15220 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15221 for vectors, as a partial revert of r254296.
15222 * rtl.h (const_vec_p): Delete.
15223 (const_vec_duplicate_p): Don't test for vector CONSTs.
15224 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15225 * expmed.c (make_tree): Likewise.
15228 * common.md (E, F): Use CONSTANT_P instead of checking for
15230 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15231 checking for CONST_VECTOR.
15233 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15235 PR middle-end/83575
15236 * predict.c (force_edge_cold): Handle in more sane way edges
15237 with no prediction.
15239 2018-01-09 Carl Love <cel@us.ibm.com>
15241 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15243 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15244 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15245 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15246 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15247 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15248 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15249 * config/rs6000/rs6000-protos.h: Add extern defition for
15250 rs6000_generate_float2_double_code.
15251 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15253 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15254 (float2_v2df): Add define_expand.
15256 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15259 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15260 op_mode in the force_to_mode call.
15262 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15264 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15265 instead of checking each element individually.
15266 (aarch64_evpc_uzp): Likewise.
15267 (aarch64_evpc_zip): Likewise.
15268 (aarch64_evpc_ext): Likewise.
15269 (aarch64_evpc_rev): Likewise.
15270 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15271 instead of checking each element individually. Return true without
15273 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15274 whether all selected elements come from the same input, instead of
15275 checking each element individually. Remove calls to gen_rtx_REG,
15276 start_sequence and end_sequence and instead assert that no rtl is
15279 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15281 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15282 order of HIGH and CONST checks.
15284 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15286 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15287 if the destination isn't an SSA_NAME.
15289 2018-01-09 Richard Biener <rguenther@suse.de>
15291 PR tree-optimization/83668
15292 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15294 (canonicalize_loop_form): ... here, renamed from ...
15295 (canonicalize_loop_closed_ssa_form): ... this and amended to
15296 swap successor edges for loop exit blocks to make us use
15297 the RPO order we need for initial schedule generation.
15299 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15301 PR tree-optimization/64811
15302 * match.pd: When optimizing comparisons with Inf, avoid
15303 introducing or losing exceptions from comparisons with NaN.
15305 2018-01-09 Martin Liska <mliska@suse.cz>
15308 * asan.c (shadow_mem_size): Add gcc_assert.
15310 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15312 Don't save registers in main().
15315 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15316 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15317 * config/avr/avr.c (avr_set_current_function): Don't error if
15318 naked, OS_task or OS_main are specified at the same time.
15319 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15321 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15323 * common/config/avr/avr-common.c (avr_option_optimization_table):
15324 Switch on -mmain-is-OS_task for optimizing compilations.
15326 2018-01-09 Richard Biener <rguenther@suse.de>
15328 PR tree-optimization/83572
15329 * graphite.c: Include cfganal.h.
15330 (graphite_transform_loops): Connect infinite loops to exit
15331 and remove fake edges at the end.
15333 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15335 * ipa-inline.c (edge_badness): Revert accidental checkin.
15337 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15340 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15341 symbols; not inline clones.
15343 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15346 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15347 hard registers. Formatting fixes.
15349 PR preprocessor/83722
15350 * gcc.c (try_generate_repro): Pass
15351 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15352 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15355 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15356 Kito Cheng <kito.cheng@gmail.com>
15358 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15359 (riscv_leaf_function_p): Delete.
15360 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15362 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15364 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15366 (do_ifelse): New function.
15367 (do_isel): New function.
15368 (do_sub3): New function.
15369 (do_add3): New function.
15370 (do_load_mask_compare): New function.
15371 (do_overlap_load_compare): New function.
15372 (expand_compare_loop): New function.
15373 (expand_block_compare): Call expand_compare_loop() when appropriate.
15374 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15375 option description.
15376 (-mblock-compare-inline-loop-limit): New option.
15378 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15381 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15382 Reverse order of second and third operands in first alternative.
15383 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15384 of first and second elements in UNSPEC_VPERMR vector.
15385 (altivec_expand_vec_perm_le): Likewise.
15387 2018-01-08 Jeff Law <law@redhat.com>
15389 PR rtl-optimizatin/81308
15390 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15391 (process_switch): If group_case_labels makes a change, then set
15393 (pass_convert_switch::execute): If a switch is converted, then
15394 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15396 PR rtl-optimization/81308
15397 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15400 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15402 PR target/83663 - Revert r255946
15403 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15404 generation for cases where splatting a value is not useful.
15405 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15406 across a vec_duplicate and a paradoxical subreg forming a vector
15407 mode to a vec_concat.
15409 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15411 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15412 -march=armv8.3-a variants.
15413 * config/arm/t-multilib: Likewise.
15414 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15416 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15418 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15420 (cceq_ior_compare_complement): Give it a name so I can use it, and
15421 change boolean_or_operator predicate to boolean_operator so it can
15422 be used to generate a crand.
15423 (eqne): New code iterator.
15424 (bd/bd_neg): New code_attrs.
15425 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15426 a single define_insn.
15427 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15428 decrement (bdnzt/bdnzf/bdzt/bdzf).
15429 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15430 with the new names of the branch decrement patterns, and added the
15431 names of the branch decrement conditional patterns.
15433 2018-01-08 Richard Biener <rguenther@suse.de>
15435 PR tree-optimization/83563
15436 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15439 2018-01-08 Richard Biener <rguenther@suse.de>
15441 PR middle-end/83713
15442 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15444 2018-01-08 Richard Biener <rguenther@suse.de>
15446 PR tree-optimization/83685
15447 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15448 references to abnormals.
15450 2018-01-08 Richard Biener <rguenther@suse.de>
15453 * dwarf2out.c (output_indirect_strings): Handle empty
15454 skeleton_debug_str_hash.
15455 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15457 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15459 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15460 (emit_store_direct): Likewise.
15461 (arc_trampoline_adjust_address): Likewise.
15462 (arc_asm_trampoline_template): New function.
15463 (arc_initialize_trampoline): Use asm_trampoline_template.
15464 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15465 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15466 * config/arc/arc.md (flush_icache): Delete pattern.
15468 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15470 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15471 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15474 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15477 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15478 by not USED_FOR_TARGET.
15479 (make_pass_resolve_sw_modes): Likewise.
15481 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15483 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15486 2018-01-08 Richard Biener <rguenther@suse.de>
15488 PR middle-end/83580
15489 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15491 2018-01-08 Richard Biener <rguenther@suse.de>
15493 PR middle-end/83517
15494 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15496 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15498 PR middle-end/81897
15499 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15500 basic blocks with a small number of successors.
15501 (convert_control_dep_chain_into_preds): Improve handling of
15503 (dump_predicates): Split apart into...
15504 (dump_pred_chain): ...here...
15505 (dump_pred_info): ...and here.
15506 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15507 (can_chain_union_be_invalidated_p): Improve check for invalidation
15509 (uninit_uses_cannot_happen): Avoid unnecessary if
15510 convert_control_dep_chain_into_preds yielded nothing.
15512 2018-01-06 Martin Sebor <msebor@redhat.com>
15514 PR tree-optimization/83640
15515 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15516 subtracting negative offset from size.
15517 (builtin_access::overlap): Adjust offset bounds of the access to fall
15518 within the size of the object if possible.
15520 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15522 PR rtl-optimization/83699
15523 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15524 extract_bit_field_as_subreg to cases in which the extracted
15525 value is also a vector.
15527 * lra-constraints.c (process_alt_operands): Test for the equivalence
15528 substitutions when detecting a possible reload cycle.
15530 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15533 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15534 by default if flag_selective_schedling{,2}. Formatting fixes.
15536 PR rtl-optimization/83682
15537 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15538 if it has non-VECTOR_MODE element mode.
15539 (vec_duplicate_p): Likewise.
15541 PR middle-end/83694
15542 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15543 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15545 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15548 * config/i386/i386-builtin.def
15549 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15550 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15551 Require also OPTION_MASK_ISA_AVX512F in addition to
15552 OPTION_MASK_ISA_GFNI.
15553 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15554 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15555 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15556 to OPTION_MASK_ISA_GFNI.
15557 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15558 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15559 OPTION_MASK_ISA_AVX512BW.
15560 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15561 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15562 addition to OPTION_MASK_ISA_GFNI.
15563 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15564 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15565 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15566 to OPTION_MASK_ISA_GFNI.
15567 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15568 a requirement for all ISAs rather than any of them with a few
15570 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15572 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15573 bitmasks to be enabled with 3 exceptions, instead of requiring any
15574 enabled ISA with lots of exceptions.
15575 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15576 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15577 Change avx512bw in isa attribute to avx512f.
15578 * config/i386/sgxintrin.h: Add license boilerplate.
15579 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15580 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15581 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15582 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15584 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15585 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15586 temporarily sse2 rather than sse if not enabled already.
15589 * config/i386/sse.md (VI248_VLBW): Rename to ...
15590 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15591 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15592 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15593 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15594 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15595 mode iterator instead of VI248_VLBW.
15597 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15599 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15600 (record_modified): Skip clobbers; add debug output.
15601 (param_change_prob): Use sreal frequencies.
15603 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15605 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15606 punt for user-aligned variables.
15608 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15610 * tree-chrec.c (chrec_contains_symbols): Return true for
15613 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15616 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15617 of (x|y) == x for BICS pattern.
15619 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15621 PR tree-optimization/83605
15622 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15623 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15626 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15628 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15629 * config/epiphany/rtems.h: New file.
15631 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15632 Uros Bizjak <ubizjak@gmail.com>
15635 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15636 QIreg_operand instead of register_operand predicate.
15637 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15638 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15639 comments instead of -fmitigate[-_]rop.
15641 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15644 * cgraphunit.c (symbol_table::compile): Switch to text_section
15645 before calling assembly_start debug hook.
15646 * run-rtl-passes.c (run_rtl_passes): Likewise.
15649 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15651 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15652 range_int_cst_p rather than !symbolic_range_p before calling
15653 extract_range_from_multiplicative_op_1.
15655 2018-01-04 Jeff Law <law@redhat.com>
15657 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15658 redundant test in assertion.
15660 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15662 * doc/rtl.texi: Document machine_mode wrapper classes.
15664 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15666 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15667 using tree_to_uhwi.
15669 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15671 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15672 the VEC_PERM_EXPR fold to fail.
15674 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15677 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15678 to switched_sections.
15680 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15683 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15684 test for d.testing.
15686 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15689 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15690 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15692 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15695 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15696 is BLKmode and bitpos not zero or mode change is needed.
15698 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15701 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15704 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15707 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15708 instead of MULT rtx. Update all corresponding splitters.
15709 (*saddl_se): Ditto.
15710 (*ssub<modesuffix>): Ditto.
15711 (*ssubl_se): Ditto.
15712 (*cmp_sadd_di): Update split patterns.
15713 (*cmp_sadd_si): Ditto.
15714 (*cmp_sadd_sidi): Ditto.
15715 (*cmp_ssub_di): Ditto.
15716 (*cmp_ssub_si): Ditto.
15717 (*cmp_ssub_sidi): Ditto.
15718 * config/alpha/predicates.md (const23_operand): New predicate.
15719 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15720 Look for ASHIFT, not MULT inner operand.
15721 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15723 2018-01-04 Martin Liska <mliska@suse.cz>
15725 PR gcov-profile/83669
15726 * gcov.c (output_intermediate_file): Add version to intermediate
15728 * doc/gcov.texi: Document new field 'version' in intermediate
15729 file format. Fix location of '-k' option of gcov command.
15731 2018-01-04 Martin Liska <mliska@suse.cz>
15734 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15736 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15738 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15740 2018-01-03 Martin Sebor <msebor@redhat.com>
15742 PR tree-optimization/83655
15743 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15744 checking calls with invalid arguments.
15746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15748 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15749 (vectorizable_mask_load_store): Delete.
15750 (vectorizable_call): Return false for masked loads and stores.
15751 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15752 instead of gimple_assign_rhs1.
15753 (vectorizable_load): Handle IFN_MASK_LOAD.
15754 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15758 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15760 (vectorizable_mask_load_store): ...here.
15761 (vectorizable_load): ...and here.
15763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15765 * tree-vect-stmts.c (vect_build_all_ones_mask)
15766 (vect_build_zero_merge_argument): New functions, split out from...
15767 (vectorizable_load): ...here.
15769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15771 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15773 (vectorizable_mask_load_store): ...here.
15774 (vectorizable_store): ...and here.
15776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15778 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15780 (vectorizable_mask_load_store): ...here.
15782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15784 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15785 (vect_model_store_cost): Take a vec_load_store_type instead of a
15787 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15788 (vect_model_store_cost): Take a vec_load_store_type instead of a
15790 (vectorizable_mask_load_store): Update accordingly.
15791 (vectorizable_store): Likewise.
15792 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15796 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15797 IFN_MASK_LOAD calls here rather than...
15798 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15801 Alan Hayward <alan.hayward@arm.com>
15802 David Sherwood <david.sherwood@arm.com>
15804 * expmed.c (extract_bit_field_1): For vector extracts,
15805 fall back to extract_bit_field_as_subreg if vec_extract
15808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15809 Alan Hayward <alan.hayward@arm.com>
15810 David Sherwood <david.sherwood@arm.com>
15812 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15813 they are variable or constant sized.
15814 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15815 slots for constant-sized data.
15817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15818 Alan Hayward <alan.hayward@arm.com>
15819 David Sherwood <david.sherwood@arm.com>
15821 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15822 handling COND_EXPRs with boolean comparisons, try to find a better
15823 basis for the mask type than the boolean itself.
15825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15827 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15828 is calculated and how it can be overridden.
15829 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15830 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15832 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15836 Alan Hayward <alan.hayward@arm.com>
15837 David Sherwood <david.sherwood@arm.com>
15839 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15840 Remove the mode argument.
15841 (aarch64_simd_valid_immediate): Remove the mode and inverse
15843 * config/aarch64/iterators.md (bitsize): New iterator.
15844 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15845 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15846 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15847 aarch64_simd_valid_immediate.
15848 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15849 (aarch64_reg_or_bic_imm): Likewise.
15850 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15851 with an insn_type enum and msl with a modifier_type enum.
15852 Replace element_width with a scalar_mode. Change the shift
15853 to unsigned int. Add constructors for scalar_float_mode and
15854 scalar_int_mode elements.
15855 (aarch64_vect_float_const_representable_p): Delete.
15856 (aarch64_can_const_movi_rtx_p)
15857 (aarch64_simd_scalar_immediate_valid_for_move)
15858 (aarch64_simd_make_constant): Update call to
15859 aarch64_simd_valid_immediate.
15860 (aarch64_advsimd_valid_immediate_hs): New function.
15861 (aarch64_advsimd_valid_immediate): Likewise.
15862 (aarch64_simd_valid_immediate): Remove mode and inverse
15863 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15864 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15865 and aarch64_float_const_representable_p on the result.
15866 (aarch64_output_simd_mov_immediate): Remove mode argument.
15867 Update call to aarch64_simd_valid_immediate and use of
15868 simd_immediate_info.
15869 (aarch64_output_scalar_simd_mov_immediate): Update call
15872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15873 Alan Hayward <alan.hayward@arm.com>
15874 David Sherwood <david.sherwood@arm.com>
15876 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15877 (mode_nunits): Likewise CONST_MODE_NUNITS.
15878 * machmode.def (ADJUST_NUNITS): Document.
15879 * genmodes.c (mode_data::need_nunits_adj): New field.
15880 (blank_mode): Update accordingly.
15881 (adj_nunits): New variable.
15882 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15884 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15885 listed in adj_nunits.
15886 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15887 listed in adj_nunits. Don't emit case statements for such modes.
15888 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15889 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15890 nothing if adj_nunits is nonnull.
15891 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15892 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15893 (emit_mode_fbit): Update use of print_maybe_const_decl.
15894 (emit_move_size): Likewise. Treat the array as non-const
15896 (emit_mode_adjustments): Handle adj_nunits.
15898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15900 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15901 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15902 (VECTOR_MODES): Use it.
15903 (make_vector_modes): Take the prefix as an argument.
15905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15906 Alan Hayward <alan.hayward@arm.com>
15907 David Sherwood <david.sherwood@arm.com>
15909 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15910 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15911 for MODE_VECTOR_BOOL.
15912 * machmode.def (VECTOR_BOOL_MODE): Document.
15913 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15914 (make_vector_bool_mode): New function.
15915 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15917 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15918 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15920 * stor-layout.c (int_mode_for_mode): Likewise.
15921 * tree.c (build_vector_type_for_mode): Likewise.
15922 * varasm.c (output_constant_pool_2): Likewise.
15923 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15924 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15925 for MODE_VECTOR_BOOL.
15926 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15927 of mode class checks.
15928 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15929 instead of a list of mode class checks.
15930 (expand_vector_scalar_condition): Likewise.
15931 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15934 Alan Hayward <alan.hayward@arm.com>
15935 David Sherwood <david.sherwood@arm.com>
15937 * machmode.h (mode_size): Change from unsigned short to
15939 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15940 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15941 or if measurement_type is not polynomial.
15942 (fixed_size_mode::includes_p): Check for constant-sized modes.
15943 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15944 return a poly_uint16 rather than an unsigned short.
15945 (emit_mode_size): Change the type of mode_size from unsigned short
15946 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15947 (emit_mode_adjustments): Cope with polynomial vector sizes.
15948 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15950 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15952 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15953 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15954 * caller-save.c (setup_save_areas): Likewise.
15955 (replace_reg_with_saved_mem): Likewise.
15956 * calls.c (emit_library_call_value_1): Likewise.
15957 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15958 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15959 (gen_lowpart_for_combine): Likewise.
15960 * convert.c (convert_to_integer_1): Likewise.
15961 * cse.c (equiv_constant, cse_insn): Likewise.
15962 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15963 (cselib_subst_to_values): Likewise.
15964 * dce.c (word_dce_process_block): Likewise.
15965 * df-problems.c (df_word_lr_mark_ref): Likewise.
15966 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15967 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15968 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15969 (rtl_for_decl_location): Likewise.
15970 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15971 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15972 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15973 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15974 (expand_expr_real_1): Likewise.
15975 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15976 (pad_below): Likewise.
15977 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15978 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15979 * ira.c (get_subreg_tracking_sizes): Likewise.
15980 * ira-build.c (ira_create_allocno_objects): Likewise.
15981 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15982 (ira_sort_regnos_for_alter_reg): Likewise.
15983 * ira-costs.c (record_operand_costs): Likewise.
15984 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15985 (resolve_simple_move): Likewise.
15986 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15987 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15988 (lra_constraints): Likewise.
15989 (CONST_POOL_OK_P): Reject variable-sized modes.
15990 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15991 (add_pseudo_to_slot, lra_spill): Likewise.
15992 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15993 * optabs-query.c (get_best_extraction_insn): Likewise.
15994 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15995 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15996 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15997 * recog.c (offsettable_address_addr_space_p): Likewise.
15998 * regcprop.c (maybe_mode_change): Likewise.
15999 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16000 * regrename.c (build_def_use): Likewise.
16001 * regstat.c (dump_reg_info): Likewise.
16002 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16003 (find_reloads, find_reloads_subreg_address): Likewise.
16004 * reload1.c (eliminate_regs_1): Likewise.
16005 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16006 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16007 (simplify_binary_operation_1, simplify_subreg): Likewise.
16008 * targhooks.c (default_function_arg_padding): Likewise.
16009 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16010 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16011 (verify_gimple_assign_ternary): Likewise.
16012 * tree-inline.c (estimate_move_cost): Likewise.
16013 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16014 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16015 (get_address_cost_ainc): Likewise.
16016 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16017 (vect_supportable_dr_alignment): Likewise.
16018 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16019 (vectorizable_reduction): Likewise.
16020 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16021 (vectorizable_operation, vectorizable_load): Likewise.
16022 * tree.c (build_same_sized_truth_vector_type): Likewise.
16023 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16024 * var-tracking.c (emit_note_insn_var_location): Likewise.
16025 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16026 (ADDR_VEC_ALIGN): Likewise.
16028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16029 Alan Hayward <alan.hayward@arm.com>
16030 David Sherwood <david.sherwood@arm.com>
16032 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16034 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16035 or if measurement_type is polynomial.
16036 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16037 * combine.c (make_extraction): Likewise.
16038 * dse.c (find_shift_sequence): Likewise.
16039 * dwarf2out.c (mem_loc_descriptor): Likewise.
16040 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16041 (extract_bit_field, extract_low_bits): Likewise.
16042 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16043 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16044 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16045 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16046 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16047 * reload.c (find_reloads): Likewise.
16048 * reload1.c (alter_reg): Likewise.
16049 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16050 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16051 * tree-if-conv.c (predicate_mem_writes): Likewise.
16052 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16053 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16054 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16055 * valtrack.c (dead_debug_insert_temp): Likewise.
16056 * varasm.c (mergeable_constant_section): Likewise.
16057 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16060 Alan Hayward <alan.hayward@arm.com>
16061 David Sherwood <david.sherwood@arm.com>
16063 * expr.c (expand_assignment): Cope with polynomial mode sizes
16064 when assigning to a CONCAT.
16066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16067 Alan Hayward <alan.hayward@arm.com>
16068 David Sherwood <david.sherwood@arm.com>
16070 * machmode.h (mode_precision): Change from unsigned short to
16072 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16074 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16075 or if measurement_type is not polynomial.
16076 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16077 in which the mode is already known to be a scalar_int_mode.
16078 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16079 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16081 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16082 for GET_MODE_PRECISION.
16083 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16084 for GET_MODE_PRECISION.
16085 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16087 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16088 (expand_field_assignment, make_extraction): Likewise.
16089 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16090 (get_last_value): Likewise.
16091 * convert.c (convert_to_integer_1): Likewise.
16092 * cse.c (cse_insn): Likewise.
16093 * expr.c (expand_expr_real_1): Likewise.
16094 * lra-constraints.c (simplify_operand_subreg): Likewise.
16095 * optabs-query.c (can_atomic_load_p): Likewise.
16096 * optabs.c (expand_atomic_load): Likewise.
16097 (expand_atomic_store): Likewise.
16098 * ree.c (combine_reaching_defs): Likewise.
16099 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16100 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16101 * tree.h (type_has_mode_precision_p): Likewise.
16102 * ubsan.c (instrument_si_overflow): Likewise.
16104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16105 Alan Hayward <alan.hayward@arm.com>
16106 David Sherwood <david.sherwood@arm.com>
16108 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16109 polynomial numbers of units.
16110 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16111 (valid_vector_subparts_p): New function.
16112 (build_vector_type): Remove temporary shim and take the number
16113 of units as a poly_uint64 rather than an int.
16114 (build_opaque_vector_type): Take the number of units as a
16115 poly_uint64 rather than an int.
16116 * tree.c (build_vector_from_ctor): Handle polynomial
16117 TYPE_VECTOR_SUBPARTS.
16118 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16119 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16120 (build_vector_from_val): If the number of units is variable,
16121 use build_vec_duplicate_cst for constant operands and
16122 VEC_DUPLICATE_EXPR otherwise.
16123 (make_vector_type): Remove temporary is_constant ().
16124 (build_vector_type, build_opaque_vector_type): Take the number of
16125 units as a poly_uint64 rather than an int.
16126 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16128 * cfgexpand.c (expand_debug_expr): Likewise.
16129 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16130 (store_constructor, expand_expr_real_1): Likewise.
16131 (const_scalar_mask_from_tree): Likewise.
16132 * fold-const-call.c (fold_const_reduction): Likewise.
16133 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16134 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16135 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16136 (fold_relational_const): Likewise.
16137 (native_interpret_vector): Likewise. Change the size from an
16138 int to an unsigned int.
16139 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16140 TYPE_VECTOR_SUBPARTS.
16141 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16142 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16143 duplicating a non-constant operand into a variable-length vector.
16144 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16145 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16146 * ipa-icf.c (sem_variable::equals): Likewise.
16147 * match.pd: Likewise.
16148 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16149 * print-tree.c (print_node): Likewise.
16150 * stor-layout.c (layout_type): Likewise.
16151 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16152 * tree-cfg.c (verify_gimple_comparison): Likewise.
16153 (verify_gimple_assign_binary): Likewise.
16154 (verify_gimple_assign_ternary): Likewise.
16155 (verify_gimple_assign_single): Likewise.
16156 * tree-pretty-print.c (dump_generic_node): Likewise.
16157 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16158 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16159 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16160 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16161 (vect_shift_permute_load_chain): Likewise.
16162 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16163 (expand_vector_condition, optimize_vector_constructor): Likewise.
16164 (lower_vec_perm, get_compute_type): Likewise.
16165 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16166 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16167 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16168 (vect_recog_mask_conversion_pattern): Likewise.
16169 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16170 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16171 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16172 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16173 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16174 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16175 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16176 (supportable_widening_operation): Likewise.
16177 (supportable_narrowing_operation): Likewise.
16178 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16180 * varasm.c (output_constant): Likewise.
16182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16183 Alan Hayward <alan.hayward@arm.com>
16184 David Sherwood <david.sherwood@arm.com>
16186 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16187 so that both the length == 3 and length != 3 cases set up their
16188 own permute vectors. Add comments explaining why we know the
16189 number of elements is constant.
16190 (vect_permute_load_chain): Likewise.
16192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16193 Alan Hayward <alan.hayward@arm.com>
16194 David Sherwood <david.sherwood@arm.com>
16196 * machmode.h (mode_nunits): Change from unsigned char to
16198 (ONLY_FIXED_SIZE_MODES): New macro.
16199 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16200 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16201 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16203 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16204 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16205 or if measurement_type is not polynomial.
16206 * genmodes.c (ZERO_COEFFS): New macro.
16207 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16209 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16210 Use ZERO_COEFFS when emitting initializers.
16211 * data-streamer.h (bp_pack_poly_value): New function.
16212 (bp_unpack_poly_value): Likewise.
16213 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16214 for GET_MODE_NUNITS.
16215 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16216 for GET_MODE_NUNITS.
16217 * tree.c (make_vector_type): Remove temporary shim and make
16218 the real function take the number of units as a poly_uint64
16219 rather than an int.
16220 (build_vector_type_for_mode): Handle polynomial nunits.
16221 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16222 * emit-rtl.c (const_vec_series_p_1): Likewise.
16223 (gen_rtx_CONST_VECTOR): Likewise.
16224 * fold-const.c (test_vec_duplicate_folding): Likewise.
16225 * genrecog.c (validate_pattern): Likewise.
16226 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16227 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16228 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16229 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16230 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16231 * rtlanal.c (subreg_get_info): Likewise.
16232 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16233 (vect_grouped_load_supported): Likewise.
16234 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16235 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16236 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16237 (simplify_const_unary_operation, simplify_binary_operation_1)
16238 (simplify_const_binary_operation, simplify_ternary_operation)
16239 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16240 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16241 instead of CONST_VECTOR_NUNITS.
16242 * varasm.c (output_constant_pool_2): Likewise.
16243 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16244 explicit-encoded elements in the XVEC for variable-length vectors.
16246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16248 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16251 Alan Hayward <alan.hayward@arm.com>
16252 David Sherwood <david.sherwood@arm.com>
16254 * coretypes.h (fixed_size_mode): Declare.
16255 (fixed_size_mode_pod): New typedef.
16256 * builtins.h (target_builtins::x_apply_args_mode)
16257 (target_builtins::x_apply_result_mode): Change type to
16258 fixed_size_mode_pod.
16259 * builtins.c (apply_args_size, apply_result_size, result_vector)
16260 (expand_builtin_apply_args_1, expand_builtin_apply)
16261 (expand_builtin_return): Update accordingly.
16263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16265 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16266 * cselib.c (cselib_hash_rtx): Likewise.
16267 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16268 CONST_VECTOR encoding.
16270 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16271 Jeff Law <law@redhat.com>
16274 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16275 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16276 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16277 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16280 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16281 explicitly probe *sp in a noreturn function if there were any callee
16282 register saves or frame pointer is needed.
16284 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16287 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16288 BLKmode for ternary, binary or unary expressions.
16291 * var-tracking.c (delete_vta_debug_insn): New inline function.
16292 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16293 insns from get_insns () to NULL instead of each bb separately.
16294 Use delete_vta_debug_insn. No longer static.
16295 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16296 delete_vta_debug_insns callers.
16297 * rtl.h (delete_vta_debug_insns): Declare.
16298 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16299 instead of variable_tracking_main.
16301 2018-01-03 Martin Sebor <msebor@redhat.com>
16303 PR tree-optimization/83603
16304 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16305 arguments past the endof the argument list in functions declared
16306 without a prototype.
16307 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16308 Avoid checking when arguments are null.
16310 2018-01-03 Martin Sebor <msebor@redhat.com>
16313 * doc/extend.texi (attribute const): Fix a typo.
16314 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16315 issuing -Wsuggest-attribute for void functions.
16317 2018-01-03 Martin Sebor <msebor@redhat.com>
16319 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16320 offset_int::from instead of wide_int::to_shwi.
16321 (maybe_diag_overlap): Remove assertion.
16322 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16323 * gimple-ssa-sprintf.c (format_directive): Same.
16324 (parse_directive): Same.
16325 (sprintf_dom_walker::compute_format_length): Same.
16326 (try_substitute_return_value): Same.
16328 2018-01-03 Jeff Law <law@redhat.com>
16330 PR middle-end/83654
16331 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16332 non-constant residual for zero at runtime and avoid probing in
16333 that case. Reorganize code for trailing problem to mirror handling
16336 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16338 PR tree-optimization/83501
16339 * tree-ssa-strlen.c (get_string_cst): New.
16340 (handle_char_store): Call get_string_cst.
16342 2018-01-03 Martin Liska <mliska@suse.cz>
16344 PR tree-optimization/83593
16345 * tree-ssa-strlen.c: Include tree-cfg.h.
16346 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16347 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16348 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16350 (strlen_dom_walker::before_dom_children): Call
16351 gimple_purge_dead_eh_edges. Dump tranformation with details
16353 (strlen_dom_walker::before_dom_children): Update call by adding
16354 new argument cleanup_eh.
16355 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16357 2018-01-03 Martin Liska <mliska@suse.cz>
16360 * cif-code.def (VARIADIC_THUNK): New enum value.
16361 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16364 2018-01-03 Jan Beulich <jbeulich@suse.com>
16366 * sse.md (mov<mode>_internal): Tighten condition for when to use
16367 vmovdqu<ssescalarsize> for TI and OI modes.
16369 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16371 Update copyright years.
16373 2018-01-03 Martin Liska <mliska@suse.cz>
16376 * ipa-visibility.c (function_and_variable_visibility): Skip
16377 functions with noipa attribure.
16379 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16381 * gcc.c (process_command): Update copyright notice dates.
16382 * gcov-dump.c (print_version): Ditto.
16383 * gcov.c (print_version): Ditto.
16384 * gcov-tool.c (print_version): Ditto.
16385 * gengtype.c (create_file): Ditto.
16386 * doc/cpp.texi: Bump @copying's copyright year.
16387 * doc/cppinternals.texi: Ditto.
16388 * doc/gcc.texi: Ditto.
16389 * doc/gccint.texi: Ditto.
16390 * doc/gcov.texi: Ditto.
16391 * doc/install.texi: Ditto.
16392 * doc/invoke.texi: Ditto.
16394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16396 * vector-builder.h (vector_builder::m_full_nelts): Change from
16397 unsigned int to poly_uint64.
16398 (vector_builder::full_nelts): Update prototype accordingly.
16399 (vector_builder::new_vector): Likewise.
16400 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16401 (vector_builder::operator ==): Likewise.
16402 (vector_builder::finalize): Likewise.
16403 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16404 Take the number of elements as a poly_uint64 rather than an
16406 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16407 from unsigned int to poly_uint64.
16408 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16409 (vec_perm_indices::new_vector): Likewise.
16410 (vec_perm_indices::length): Likewise.
16411 (vec_perm_indices::nelts_per_input): Likewise.
16412 (vec_perm_indices::input_nelts): Likewise.
16413 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16414 number of elements per input as a poly_uint64 rather than an
16415 unsigned int. Use the original encoding for variable-length
16416 vectors, rather than clamping each individual element.
16417 For the second and subsequent elements in each pattern,
16418 clamp the step and base before clamping their sum.
16419 (vec_perm_indices::series_p): Handle polynomial element counts.
16420 (vec_perm_indices::all_in_range_p): Likewise.
16421 (vec_perm_indices_to_tree): Likewise.
16422 (vec_perm_indices_to_rtx): Likewise.
16423 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16424 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16425 (tree_vector_builder::new_binary_operation): Handle polynomial
16426 element counts. Return false if we need to know the number
16427 of elements at compile time.
16428 * fold-const.c (fold_vec_perm): Punt if the number of elements
16429 isn't known at compile time.
16431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16433 * vec-perm-indices.h (vec_perm_builder): Change element type
16434 from HOST_WIDE_INT to poly_int64.
16435 (vec_perm_indices::element_type): Update accordingly.
16436 (vec_perm_indices::clamp): Handle polynomial element_types.
16437 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16438 (vec_perm_indices::all_in_range_p): Likewise.
16439 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16441 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16442 polynomial vec_perm_indices element types.
16443 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16444 * fold-const.c (fold_vec_perm): Likewise.
16445 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16446 * tree-vect-generic.c (lower_vec_perm): Likewise.
16447 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16448 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16449 element type to HOST_WIDE_INT.
16451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16452 Alan Hayward <alan.hayward@arm.com>
16453 David Sherwood <david.sherwood@arm.com>
16455 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16456 rather than an int. Use plus_constant.
16457 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16458 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16461 Alan Hayward <alan.hayward@arm.com>
16462 David Sherwood <david.sherwood@arm.com>
16464 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16465 a HOST_WIDE_INT to a poly_int64.
16467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16468 Alan Hayward <alan.hayward@arm.com>
16469 David Sherwood <david.sherwood@arm.com>
16471 * calls.c (load_register_parameters): Cope with polynomial
16472 mode sizes. Require a constant size for BLKmode parameters
16473 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16474 forces a parameter to be padded at the lsb end in order to
16475 fill a complete number of words, require the parameter size
16476 to be ordered wrt UNITS_PER_WORD.
16478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16479 Alan Hayward <alan.hayward@arm.com>
16480 David Sherwood <david.sherwood@arm.com>
16482 * reload1.c (spill_stack_slot_width): Change element type
16483 from unsigned int to poly_uint64_pod.
16484 (alter_reg): Treat mode sizes as polynomial.
16486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16487 Alan Hayward <alan.hayward@arm.com>
16488 David Sherwood <david.sherwood@arm.com>
16490 * reload.c (complex_word_subreg_p): New function.
16491 (reload_inner_reg_of_subreg, push_reload): Use it.
16493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16494 Alan Hayward <alan.hayward@arm.com>
16495 David Sherwood <david.sherwood@arm.com>
16497 * lra-constraints.c (process_alt_operands): Reject matched
16498 operands whose sizes aren't ordered.
16499 (match_reload): Refer to this check here.
16501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16502 Alan Hayward <alan.hayward@arm.com>
16503 David Sherwood <david.sherwood@arm.com>
16505 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16506 that the mode size is in the set {1, 2, 4, 8, 16}.
16508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16509 Alan Hayward <alan.hayward@arm.com>
16510 David Sherwood <david.sherwood@arm.com>
16512 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16513 Use plus_constant instead of gen_rtx_PLUS.
16515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16516 Alan Hayward <alan.hayward@arm.com>
16517 David Sherwood <david.sherwood@arm.com>
16519 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16520 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16521 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16522 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16523 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16524 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16525 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16526 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16527 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16528 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16530 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16531 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16532 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16533 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16534 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16535 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16536 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16537 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16538 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16539 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16541 * expr.c (emit_move_resolve_push): Treat the input and result
16542 of PUSH_ROUNDING as a poly_int64.
16543 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16544 (emit_push_insn): Likewise.
16545 * lra-eliminations.c (mark_not_eliminable): Likewise.
16546 * recog.c (push_operand): Likewise.
16547 * reload1.c (elimination_effects): Likewise.
16548 * rtlanal.c (nonzero_bits1): Likewise.
16549 * calls.c (store_one_arg): Likewise. Require the padding to be
16550 known at compile time.
16552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16553 Alan Hayward <alan.hayward@arm.com>
16554 David Sherwood <david.sherwood@arm.com>
16556 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16557 Use plus_constant instead of gen_rtx_PLUS.
16559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16560 Alan Hayward <alan.hayward@arm.com>
16561 David Sherwood <david.sherwood@arm.com>
16563 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16564 rather than an int.
16566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16567 Alan Hayward <alan.hayward@arm.com>
16568 David Sherwood <david.sherwood@arm.com>
16570 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16571 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16572 via stack temporaries. Treat the mode size as polynomial too.
16574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16575 Alan Hayward <alan.hayward@arm.com>
16576 David Sherwood <david.sherwood@arm.com>
16578 * expr.c (expand_expr_real_2): When handling conversions involving
16579 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16580 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16581 as a poly_uint64 too.
16583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16584 Alan Hayward <alan.hayward@arm.com>
16585 David Sherwood <david.sherwood@arm.com>
16587 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16590 Alan Hayward <alan.hayward@arm.com>
16591 David Sherwood <david.sherwood@arm.com>
16593 * combine.c (can_change_dest_mode): Handle polynomial
16594 REGMODE_NATURAL_SIZE.
16595 * expmed.c (store_bit_field_1): Likewise.
16596 * expr.c (store_constructor): Likewise.
16597 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16598 and polynomial REGMODE_NATURAL_SIZE.
16599 (gen_lowpart_common): Likewise.
16600 * reginfo.c (record_subregs_of_mode): Likewise.
16601 * rtlanal.c (read_modify_subreg_p): Likewise.
16603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16604 Alan Hayward <alan.hayward@arm.com>
16605 David Sherwood <david.sherwood@arm.com>
16607 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16608 numbers of elements.
16610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16611 Alan Hayward <alan.hayward@arm.com>
16612 David Sherwood <david.sherwood@arm.com>
16614 * match.pd: Cope with polynomial numbers of vector elements.
16616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16617 Alan Hayward <alan.hayward@arm.com>
16618 David Sherwood <david.sherwood@arm.com>
16620 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16621 in a POINTER_PLUS_EXPR.
16623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16624 Alan Hayward <alan.hayward@arm.com>
16625 David Sherwood <david.sherwood@arm.com>
16627 * omp-simd-clone.c (simd_clone_subparts): New function.
16628 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16629 (ipa_simd_modify_function_body): Likewise.
16631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16632 Alan Hayward <alan.hayward@arm.com>
16633 David Sherwood <david.sherwood@arm.com>
16635 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16636 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16637 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16638 (expand_vector_condition, vector_element): Likewise.
16639 (subparts_gt): New function.
16640 (get_compute_type): Use subparts_gt.
16641 (count_type_subparts): Delete.
16642 (expand_vector_operations_1): Use subparts_gt instead of
16643 count_type_subparts.
16645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16646 Alan Hayward <alan.hayward@arm.com>
16647 David Sherwood <david.sherwood@arm.com>
16649 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16650 (vect_compile_time_alias): ...this new function. Do the calculation
16651 on poly_ints rather than trees.
16652 (vect_prune_runtime_alias_test_list): Update call accordingly.
16654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16655 Alan Hayward <alan.hayward@arm.com>
16656 David Sherwood <david.sherwood@arm.com>
16658 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16660 (vect_schedule_slp_instance): Likewise.
16662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16663 Alan Hayward <alan.hayward@arm.com>
16664 David Sherwood <david.sherwood@arm.com>
16666 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16667 constant and extern definitions for variable-length vectors.
16668 (vect_get_constant_vectors): Note that the number of units
16669 is known to be constant.
16671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16672 Alan Hayward <alan.hayward@arm.com>
16673 David Sherwood <david.sherwood@arm.com>
16675 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16676 of units as polynomial. Choose between WIDE and NARROW based
16679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16680 Alan Hayward <alan.hayward@arm.com>
16681 David Sherwood <david.sherwood@arm.com>
16683 * tree-vect-stmts.c (simd_clone_subparts): New function.
16684 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16687 Alan Hayward <alan.hayward@arm.com>
16688 David Sherwood <david.sherwood@arm.com>
16690 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16691 vectors as polynomial. Use build_index_vector for
16692 IFN_GOMP_SIMD_LANE.
16694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16695 Alan Hayward <alan.hayward@arm.com>
16696 David Sherwood <david.sherwood@arm.com>
16698 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16699 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16700 for variable-length vectors.
16701 (vectorizable_mask_load_store): Treat the number of units as
16702 polynomial, asserting that it is constant if the condition has
16703 already been enforced.
16704 (vectorizable_store, vectorizable_load): Likewise.
16706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16707 Alan Hayward <alan.hayward@arm.com>
16708 David Sherwood <david.sherwood@arm.com>
16710 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16711 of units as polynomial. Punt if we can't tell at compile time
16712 which vector contains the final result.
16714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16715 Alan Hayward <alan.hayward@arm.com>
16716 David Sherwood <david.sherwood@arm.com>
16718 * tree-vect-loop.c (vectorizable_induction): Treat the number
16719 of units as polynomial. Punt on SLP inductions. Use an integer
16720 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16721 cast of such a series for variable-length floating-point
16724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16725 Alan Hayward <alan.hayward@arm.com>
16726 David Sherwood <david.sherwood@arm.com>
16728 * tree.h (build_index_vector): Declare.
16729 * tree.c (build_index_vector): New function.
16730 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16731 of units as polynomial, forcibly converting it to a constant if
16732 vectorizable_reduction has already enforced the condition.
16733 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16734 to create a {1,2,3,...} vector.
16735 (vectorizable_reduction): Treat the number of units as polynomial.
16736 Choose vectype_in based on the largest scalar element size rather
16737 than the smallest number of units. Enforce the restrictions
16740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16741 Alan Hayward <alan.hayward@arm.com>
16742 David Sherwood <david.sherwood@arm.com>
16744 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16745 number of units as polynomial.
16747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16748 Alan Hayward <alan.hayward@arm.com>
16749 David Sherwood <david.sherwood@arm.com>
16751 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16752 * target.def (autovectorize_vector_sizes): Return the vector sizes
16753 by pointer, using vector_sizes rather than a bitmask.
16754 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16755 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16756 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16758 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16759 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16760 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16761 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16762 * omp-general.c (omp_max_vf): Likewise.
16763 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16764 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16765 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16766 * tree-vect-slp.c (vect_slp_bb): Likewise.
16767 * doc/tm.texi: Regenerate.
16768 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16770 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16771 the vector size as a poly_uint64 rather than an unsigned int.
16772 (current_vector_size): Change from an unsigned int to a poly_uint64.
16773 (get_vectype_for_scalar_type): Update accordingly.
16774 * tree.h (build_truth_vector_type): Take the size and number of
16775 units as a poly_uint64 rather than an unsigned int.
16776 (build_vector_type): Add a temporary overload that takes
16777 the number of units as a poly_uint64 rather than an unsigned int.
16778 * tree.c (make_vector_type): Likewise.
16779 (build_truth_vector_type): Take the number of units as a poly_uint64
16780 rather than an unsigned int.
16782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16783 Alan Hayward <alan.hayward@arm.com>
16784 David Sherwood <david.sherwood@arm.com>
16786 * target.def (get_mask_mode): Take the number of units and length
16787 as poly_uint64s rather than unsigned ints.
16788 * targhooks.h (default_get_mask_mode): Update accordingly.
16789 * targhooks.c (default_get_mask_mode): Likewise.
16790 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16791 * doc/tm.texi: Regenerate.
16793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16794 Alan Hayward <alan.hayward@arm.com>
16795 David Sherwood <david.sherwood@arm.com>
16797 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16798 * omp-general.c (omp_max_vf): Likewise.
16799 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16800 (expand_omp_simd): Handle polynomial safelen.
16801 * omp-low.c (omplow_simd_context): Add a default constructor.
16802 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16803 (lower_rec_simd_input_clauses): Update accordingly.
16804 (lower_rec_input_clauses): Likewise.
16806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16807 Alan Hayward <alan.hayward@arm.com>
16808 David Sherwood <david.sherwood@arm.com>
16810 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16811 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16812 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16813 (vect_analyze_slp_cost): Likewise.
16814 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16815 (vect_model_load_cost): Likewise.
16817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16818 Alan Hayward <alan.hayward@arm.com>
16819 David Sherwood <david.sherwood@arm.com>
16821 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16822 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16823 from an unsigned int * to a poly_uint64_pod *.
16824 (calculate_unrolling_factor): New function.
16825 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16832 from an unsigned int to a poly_uint64.
16833 (_loop_vec_info::slp_unrolling_factor): Likewise.
16834 (_loop_vec_info::vectorization_factor): Change from an int
16836 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16837 (vect_get_num_vectors): New function.
16838 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16839 (vect_get_num_copies): Use vect_get_num_vectors.
16840 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16841 to an unsigned int *.
16842 (vect_analyze_data_refs): Change min_vf from an int * to a
16844 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16845 than an unsigned HOST_WIDE_INT.
16846 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16847 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16848 to an unsigned int *.
16849 (vect_analyze_data_ref_dependences): Likewise.
16850 (vect_compute_data_ref_alignment): Handle polynomial vf.
16851 (vect_enhance_data_refs_alignment): Likewise.
16852 (vect_prune_runtime_alias_test_list): Likewise.
16853 (vect_shift_permute_load_chain): Likewise.
16854 (vect_supportable_dr_alignment): Likewise.
16855 (dependence_distance_ge_vf): Take the vectorization factor as a
16856 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16857 (vect_analyze_data_refs): Change min_vf from an int * to a
16859 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16860 vfm1 as a poly_uint64 rather than an int. Make the same change
16861 for the returned bound_scalar.
16862 (vect_gen_vector_loop_niters): Handle polynomial vf.
16863 (vect_do_peeling): Likewise. Update call to
16864 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16865 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16867 * tree-vect-loop.c (vect_determine_vectorization_factor)
16868 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16869 (vect_get_known_peeling_cost): Likewise.
16870 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16871 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16872 (vect_transform_loop): Likewise. Use the lowest possible VF when
16873 updating the upper bounds of the loop.
16874 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16875 rather than an int.
16876 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16877 polynomial unroll factors.
16878 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16879 (vect_make_slp_decision): Likewise.
16880 (vect_supported_load_permutation_p): Likewise, and polynomial
16882 (vect_analyze_slp_cost): Handle polynomial vf.
16883 (vect_slp_analyze_node_operations): Likewise.
16884 (vect_slp_analyze_bb_1): Likewise.
16885 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16886 than an unsigned HOST_WIDE_INT.
16887 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16888 (vectorizable_load): Handle polynomial vf.
16889 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16891 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16894 Alan Hayward <alan.hayward@arm.com>
16895 David Sherwood <david.sherwood@arm.com>
16897 * match.pd: Handle bit operations involving three constants
16898 and try to fold one pair.
16900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16902 * tree-vect-loop-manip.c: Include gimple-fold.h.
16903 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16904 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16905 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16906 Add a path that uses a step of VF instead of 1, but disable it
16908 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16909 and niters_no_overflow parameters. Update calls to
16910 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16911 Create a new SSA name if the latter choses to use a ste other
16912 than zero, and return it via niters_vector_mult_vf_var.
16913 * tree-vect-loop.c (vect_transform_loop): Update calls to
16914 vect_do_peeling, vect_gen_vector_loop_niters and
16915 slpeel_make_loop_iterate_ntimes.
16916 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16917 (vect_gen_vector_loop_niters): Update declarations after above changes.
16919 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16921 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16922 128-bit round to integer instructions.
16923 (ceil<mode>2): Likewise.
16924 (btrunc<mode>2): Likewise.
16925 (round<mode>2): Likewise.
16927 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16929 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16930 unaligned VSX load/store on P8/P9.
16931 (expand_block_clear): Allow the use of unaligned VSX
16932 load/store on P8/P9.
16934 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16936 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16938 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16939 swap associated with both a load and a store.
16941 2018-01-02 Andrew Waterman <andrew@sifive.com>
16943 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16944 * config/riscv/riscv.md (clear_cache): Use it.
16946 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16948 * web.c: Remove out-of-date comment.
16950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16952 * expr.c (fixup_args_size_notes): Check that any existing
16953 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16954 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16955 (emit_single_push_insn): ...here.
16957 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16959 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16960 (const_vector_encoded_nelts): New function.
16961 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16962 (const_vector_int_elt, const_vector_elt): Declare.
16963 * emit-rtl.c (const_vector_int_elt_1): New function.
16964 (const_vector_elt): Likewise.
16965 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16966 of CONST_VECTOR_ELT.
16968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16970 * expr.c: Include rtx-vector-builder.h.
16971 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16972 directly on the tree encoding.
16973 (const_vector_from_tree): Likewise.
16974 * optabs.c: Include rtx-vector-builder.h.
16975 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16976 sequence of "u" values.
16977 * vec-perm-indices.c: Include rtx-vector-builder.h.
16978 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16979 directly on the vec_perm_indices encoding.
16981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16983 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16984 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16985 * rtx-vector-builder.h: New file.
16986 * rtx-vector-builder.c: Likewise.
16987 * rtl.h (rtx_def::u2): Add a const_vector field.
16988 (CONST_VECTOR_NPATTERNS): New macro.
16989 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16990 (CONST_VECTOR_DUPLICATE_P): Likewise.
16991 (CONST_VECTOR_STEPPED_P): Likewise.
16992 (CONST_VECTOR_ENCODED_ELT): Likewise.
16993 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16994 (unwrap_const_vec_duplicate): Likewise.
16995 (const_vec_series_p): Check for a non-duplicated vector encoding.
16996 Say that the function only returns true for integer vectors.
16997 * emit-rtl.c: Include rtx-vector-builder.h.
16998 (gen_const_vec_duplicate_1): Delete.
16999 (gen_const_vector): Call gen_const_vec_duplicate instead of
17000 gen_const_vec_duplicate_1.
17001 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17002 (gen_const_vec_duplicate): Use rtx_vector_builder.
17003 (gen_const_vec_series): Likewise.
17004 (gen_rtx_CONST_VECTOR): Likewise.
17005 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17006 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17007 Build a new vector rather than modifying a CONST_VECTOR in-place.
17008 (handle_special_swappables): Update call accordingly.
17009 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17010 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17011 Build a new vector rather than modifying a CONST_VECTOR in-place.
17012 (handle_special_swappables): Update call accordingly.
17014 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17016 * simplify-rtx.c (simplify_const_binary_operation): Use
17017 CONST_VECTOR_ELT instead of XVECEXP.
17019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17021 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17022 the selector elements to be different from the data elements
17023 if the selector is a VECTOR_CST.
17024 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17025 ssizetype for the selector.
17027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17029 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17030 before testing each element individually.
17031 * tree-vect-generic.c (lower_vec_perm): Likewise.
17033 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17035 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17036 * selftest-run-tests.c (selftest::run_tests): Call it.
17037 * vector-builder.h (vector_builder::operator ==): New function.
17038 (vector_builder::operator !=): Likewise.
17039 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17040 (vec_perm_indices::all_from_input_p): New function.
17041 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17042 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17043 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17044 instead of reading the VECTOR_CST directly. Detect whether both
17045 vector inputs are the same before constructing the vec_perm_indices,
17046 and update the number of inputs argument accordingly. Use the
17047 utility functions added above. Only construct sel2 if we need to.
17049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17051 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17052 the broadcast of the low byte.
17053 (expand_mult_highpart): Use an explicit encoding for the permutes.
17054 * optabs-query.c (can_mult_highpart_p): Likewise.
17055 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17056 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17057 (vectorizable_bswap): Likewise.
17058 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17059 explicit encoding for the power-of-2 permutes.
17060 (vect_permute_store_chain): Likewise.
17061 (vect_grouped_load_supported): Likewise.
17062 (vect_permute_load_chain): Likewise.
17064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17066 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17067 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17068 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17069 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17070 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17071 (vect_gen_perm_mask_any): Likewise.
17073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17075 * int-vector-builder.h: New file.
17076 * vec-perm-indices.h: Include int-vector-builder.h.
17077 (vec_perm_indices): Redefine as an int_vector_builder.
17078 (auto_vec_perm_indices): Delete.
17079 (vec_perm_builder): Redefine as a stand-alone class.
17080 (vec_perm_indices::vec_perm_indices): New function.
17081 (vec_perm_indices::clamp): Likewise.
17082 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17083 (vec_perm_indices::new_vector): New function.
17084 (vec_perm_indices::new_expanded_vector): Update for new
17085 vec_perm_indices class.
17086 (vec_perm_indices::rotate_inputs): New function.
17087 (vec_perm_indices::all_in_range_p): Operate directly on the
17088 encoded form, without computing elided elements.
17089 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17090 encoding. Update for new vec_perm_indices class.
17091 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17092 the given vec_perm_builder.
17093 (expand_vec_perm_var): Update vec_perm_builder constructor.
17094 (expand_mult_highpart): Use vec_perm_builder instead of
17095 auto_vec_perm_indices.
17096 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17097 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17098 or double series encoding as appropriate.
17099 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17100 vec_perm_indices instead of auto_vec_perm_indices.
17101 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17102 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17103 (vect_permute_store_chain): Likewise.
17104 (vect_grouped_load_supported): Likewise.
17105 (vect_permute_load_chain): Likewise.
17106 (vect_shift_permute_load_chain): Likewise.
17107 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17108 (vect_transform_slp_perm_load): Likewise.
17109 (vect_schedule_slp_instance): Likewise.
17110 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17111 (vectorizable_mask_load_store): Likewise.
17112 (vectorizable_bswap): Likewise.
17113 (vectorizable_store): Likewise.
17114 (vectorizable_load): Likewise.
17115 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17116 vec_perm_indices instead of auto_vec_perm_indices. Use
17117 tree_to_vec_perm_builder to read the vector from a tree.
17118 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17119 vec_perm_builder instead of a vec_perm_indices.
17120 (have_whole_vector_shift): Use vec_perm_builder and
17121 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17122 truncation to calc_vec_perm_mask_for_shift.
17123 (vect_create_epilog_for_reduction): Likewise.
17124 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17125 from auto_vec_perm_indices to vec_perm_indices.
17126 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17127 instead of changing individual elements.
17128 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17129 the vector in d.perm.
17130 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17131 from auto_vec_perm_indices to vec_perm_indices.
17132 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17133 instead of changing individual elements.
17134 (arm_vectorize_vec_perm_const): Use new_vector to install
17135 the vector in d.perm.
17136 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17137 Update vec_perm_builder constructor.
17138 (rs6000_expand_interleave): Likewise.
17139 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17140 (rs6000_expand_interleave): Likewise.
17142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17144 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17145 to qimode could truncate the indices.
17146 * optabs.c (expand_vec_perm_var): Likewise.
17148 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17150 * Makefile.in (OBJS): Add vec-perm-indices.o.
17151 * vec-perm-indices.h: New file.
17152 * vec-perm-indices.c: Likewise.
17153 * target.h (vec_perm_indices): Replace with a forward class
17155 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17156 * optabs.h: Include vec-perm-indices.h.
17157 (expand_vec_perm): Delete.
17158 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17159 (expand_vec_perm_const): Declare.
17160 * target.def (vec_perm_const_ok): Replace with...
17161 (vec_perm_const): ...this new hook.
17162 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17163 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17164 * doc/tm.texi: Regenerate.
17165 * optabs.def (vec_perm_const): Delete.
17166 * doc/md.texi (vec_perm_const): Likewise.
17167 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17168 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17169 expand_vec_perm for constant permutation vectors. Assert that
17170 the mode of variable permutation vectors is the integer equivalent
17171 of the mode that is being permuted.
17172 * optabs-query.h (selector_fits_mode_p): Declare.
17173 * optabs-query.c: Include vec-perm-indices.h.
17174 (selector_fits_mode_p): New function.
17175 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17176 is defined, instead of checking whether the vec_perm_const_optab
17177 exists. Use targetm.vectorize.vec_perm_const instead of
17178 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17179 fit in the vector mode before using a variable permute.
17180 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17181 vec_perm_indices instead of an rtx.
17182 (expand_vec_perm): Replace with...
17183 (expand_vec_perm_const): ...this new function. Take the selector
17184 as a vec_perm_indices rather than an rtx. Also take the mode of
17185 the selector. Update call to shift_amt_for_vec_perm_mask.
17186 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17187 Use vec_perm_indices::new_expanded_vector to expand the original
17188 selector into bytes. Check whether the indices fit in the vector
17189 mode before using a variable permute.
17190 (expand_vec_perm_var): Make global.
17191 (expand_mult_highpart): Use expand_vec_perm_const.
17192 * fold-const.c: Includes vec-perm-indices.h.
17193 * tree-ssa-forwprop.c: Likewise.
17194 * tree-vect-data-refs.c: Likewise.
17195 * tree-vect-generic.c: Likewise.
17196 * tree-vect-loop.c: Likewise.
17197 * tree-vect-slp.c: Likewise.
17198 * tree-vect-stmts.c: Likewise.
17199 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17201 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17202 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17203 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17204 (aarch64_vectorize_vec_perm_const): ...this new function.
17205 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17206 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17207 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17208 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17209 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17210 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17211 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17213 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17214 check for NEON modes.
17215 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17216 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17217 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17218 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17220 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17221 the old VEC_PERM_CONST conditions.
17222 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17223 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17224 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17225 (ia64_vectorize_vec_perm_const_ok): Merge into...
17226 (ia64_vectorize_vec_perm_const): ...this new function.
17227 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17228 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17229 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17230 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17231 * config/mips/mips.c (mips_expand_vec_perm_const)
17232 (mips_vectorize_vec_perm_const_ok): Merge into...
17233 (mips_vectorize_vec_perm_const): ...this new function.
17234 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17235 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17236 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17237 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17238 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17239 (rs6000_expand_vec_perm_const): Delete.
17240 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17242 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17243 (altivec_expand_vec_perm_const_le): Take each operand individually.
17244 Operate on constant selectors rather than rtxes.
17245 (altivec_expand_vec_perm_const): Likewise. Update call to
17246 altivec_expand_vec_perm_const_le.
17247 (rs6000_expand_vec_perm_const): Delete.
17248 (rs6000_vectorize_vec_perm_const_ok): Delete.
17249 (rs6000_vectorize_vec_perm_const): New function.
17250 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17251 an element count and rtx array.
17252 (rs6000_expand_extract_even): Update call accordingly.
17253 (rs6000_expand_interleave): Likewise.
17254 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17255 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17256 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17257 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17258 (rs6000_expand_vec_perm_const): Delete.
17259 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17260 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17261 (altivec_expand_vec_perm_const_le): Take each operand individually.
17262 Operate on constant selectors rather than rtxes.
17263 (altivec_expand_vec_perm_const): Likewise. Update call to
17264 altivec_expand_vec_perm_const_le.
17265 (rs6000_expand_vec_perm_const): Delete.
17266 (rs6000_vectorize_vec_perm_const_ok): Delete.
17267 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17268 reference to the SPE evmerge intructions.
17269 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17270 an element count and rtx array.
17271 (rs6000_expand_extract_even): Update call accordingly.
17272 (rs6000_expand_interleave): Likewise.
17273 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17274 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17276 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17280 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17281 vector mode and that that mode matches the mode of the data
17283 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17284 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17285 directly using expand_vec_perm_1 when forcing selectors into
17287 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17289 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17291 * optabs-query.h (can_vec_perm_p): Delete.
17292 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17293 * optabs-query.c (can_vec_perm_p): Split into...
17294 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17295 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17296 particular selector is valid.
17297 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17298 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17299 (vect_grouped_load_supported): Likewise.
17300 (vect_shift_permute_load_chain): Likewise.
17301 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17302 (vect_transform_slp_perm_load): Likewise.
17303 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17304 (vectorizable_bswap): Likewise.
17305 (vect_gen_perm_mask_checked): Likewise.
17306 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17307 implementations of variable permutation vectors into account
17308 when deciding which selector to use.
17309 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17310 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17311 with a false third argument.
17312 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17313 to test whether the constant selector is valid and can_vec_perm_var_p
17314 to test whether a variable selector is valid.
17316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17318 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17319 * optabs-query.c (can_vec_perm_p): Likewise.
17320 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17321 instead of vec_perm_indices.
17322 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17323 (vect_gen_perm_mask_checked): Likewise,
17324 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17325 (vect_gen_perm_mask_checked): Likewise,
17327 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17329 * optabs-query.h (qimode_for_vec_perm): Declare.
17330 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17331 (qimode_for_vec_perm): ...this new function.
17332 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17334 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17336 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17337 does not have a conditional at the top.
17339 2018-01-02 Richard Biener <rguenther@suse.de>
17341 * ipa-inline.c (big_speedup_p): Fix expression.
17343 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17346 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17349 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17353 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17354 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17355 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17356 cond_taken_branch_cost 3->4.
17358 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17360 PR tree-optimization/83581
17361 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17362 TODO_cleanup_cfg if any changes have been made.
17364 PR middle-end/83608
17365 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17366 convert_modes if target mode has the right side, but different mode
17369 PR middle-end/83609
17370 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17371 last argument when extracting from CONCAT. If either from_real or
17372 from_imag is NULL, use expansion through memory. If result is not
17373 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17374 the parts directly to inner mode, if even that fails, use expansion
17377 PR middle-end/83623
17378 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17379 check for bswap in mode rather than HImode and use that in expand_unop
17382 Copyright (C) 2018 Free Software Foundation, Inc.
17384 Copying and distribution of this file, with or without modification,
17385 are permitted in any medium without royalty provided the copyright
17386 notice and this notice are preserved.