1 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/86231
4 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
5 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7 2018-06-20 Tom de Vries <tdevries@suse.de>
9 PR tree-optimization/86097
10 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11 iv type if signedness of iv type is not the same as that of *nit.
13 2018-06-20 Jakub Jelinek <jakub@redhat.com>
15 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
16 EDGE_EH edges, verify they are all EDGE_EH.
18 2018-06-20 Maya Rashish <coypu@sdf.org>
20 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
22 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
25 * config/aarch64/aarch64.c (xgene1_tunings): Add
26 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
27 (aarch64_mode_valid_for_sched_fusion_p):
29 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
30 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
32 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
34 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
35 * config/aarch64/iterators.md (VQ2): New mode iterator.
37 2018-06-20 Martin Liska <mliska@suse.cz>
39 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
40 Change default ratio from 10 to 8.
42 2018-06-20 Martin Liska <mliska@suse.cz>
44 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
46 (bit_test_cluster::find_bit_tests): Likewise.
47 (switch_decision_tree::analyze_switch_statement): Find clusters.
48 * tree-switch-conversion.h (struct jump_table_cluster): Document
51 2018-06-20 Martin Liska <mliska@suse.cz>
53 * tree-switch-conversion.c (switch_conversion::collect):
54 Record m_uniq property.
55 (switch_conversion::expand): Bail out for special conditions.
56 (group_cluster::~group_cluster): New.
57 (group_cluster::group_cluster): Likewise.
58 (group_cluster::dump): Likewise.
59 (jump_table_cluster::emit): New.
60 (switch_decision_tree::fix_phi_operands_for_edges): New.
61 (struct case_node): Remove struct.
62 (jump_table_cluster::can_be_handled): New.
63 (case_values_threshold): Moved to header.
64 (reset_out_edges_aux): Likewise.
65 (jump_table_cluster::is_beneficial): New.
66 (bit_test_cluster::can_be_handled): Likewise.
67 (add_case_node): Remove.
68 (bit_test_cluster::is_beneficial): New.
69 (case_bit_test::cmp): New.
70 (bit_test_cluster::emit): New.
71 (expand_switch_as_decision_tree_p): Remove.
72 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
73 (fix_phi_operands_for_edge): Likewise.
74 (switch_decision_tree::analyze_switch_statement): New.
75 (compute_cases_per_edge): Move ...
76 (switch_decision_tree::compute_cases_per_edge): ... here.
77 (try_switch_expansion): Likewise.
78 (switch_decision_tree::try_switch_expansion): Likewise.
79 (record_phi_operand_mapping): Likewise.
80 (switch_decision_tree::record_phi_operand_mapping): Likewise.
81 (emit_case_decision_tree): Likewise.
82 (switch_decision_tree::emit): Likewise.
83 (balance_case_nodes): Likewise.
84 (switch_decision_tree::balance_case_nodes): Likewise.
85 (dump_case_nodes): Likewise.
86 (switch_decision_tree::dump_case_nodes): Likewise.
87 (emit_jump): Likewise.
88 (switch_decision_tree::emit_jump): Likewise.
89 (emit_cmp_and_jump_insns): Likewise.
90 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
91 (emit_case_nodes): Likewise.
92 (switch_decision_tree::emit_case_nodes): Likewise.
93 (conditional_probability): Remove.
94 * tree-switch-conversion.h (enum cluster_type): New.
96 (struct cluster): Likewise.
97 (cluster::cluster): Likewise.
98 (struct simple_cluster): Likewise.
99 (simple_cluster::simple_cluster): Likewise.
100 (struct group_cluster): Likewise.
101 (struct jump_table_cluster): Likewise.
102 (struct bit_test_cluster): Likewise.
103 (struct min_cluster_item): Likewise.
104 (struct case_tree_node): Likewise.
105 (case_tree_node::case_tree_node): Likewise.
106 (jump_table_cluster::case_values_threshold): Likewise.
107 (struct case_bit_test): Likewise.
108 (struct switch_decision_tree): Likewise.
109 (struct switch_conversion): Likewise.
110 (switch_decision_tree::reset_out_edges_aux): Likewise.
112 2018-06-20 Martin Liska <mliska@suse.cz>
114 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
115 (hoist_edge_and_branch_if_true): Likewise.
116 (expand_switch_using_bit_tests_p): Likewise.
117 (struct case_bit_test): Likewise.
118 (case_bit_test_cmp): Likewise.
119 (emit_case_bit_tests): Likewise.
120 (switch_conversion::switch_conversion): New class.
121 (struct switch_conv_info): Remove old struct.
122 (collect_switch_conv_info): More to ...
123 (switch_conversion::collect): ... this.
124 (check_range): Likewise.
125 (switch_conversion::check_range): Likewise.
126 (check_all_empty_except_final): Likewise.
127 (switch_conversion::check_all_empty_except_final): Likewise.
128 (check_final_bb): Likewise.
129 (switch_conversion::check_final_bb): Likewise.
130 (create_temp_arrays): Likewise.
131 (switch_conversion::create_temp_arrays): Likewise.
132 (free_temp_arrays): Likewise.
133 (gather_default_values): Likewise.
134 (switch_conversion::gather_default_values): Likewise.
135 (build_constructors): Likewise.
136 (switch_conversion::build_constructors): Likewise.
137 (constructor_contains_same_values_p): Likewise.
138 (switch_conversion::contains_same_values_p): Likewise.
139 (array_value_type): Likewise.
140 (switch_conversion::array_value_type): Likewise.
141 (build_one_array): Likewise.
142 (switch_conversion::build_one_array): Likewise.
143 (build_arrays): Likewise.
144 (switch_conversion::build_arrays): Likewise.
145 (gen_def_assigns): Likewise.
146 (switch_conversion::gen_def_assigns): Likewise.
147 (prune_bbs): Likewise.
148 (switch_conversion::prune_bbs): Likewise.
149 (fix_phi_nodes): Likewise.
150 (switch_conversion::fix_phi_nodes): Likewise.
151 (gen_inbound_check): Likewise.
152 (switch_conversion::gen_inbound_check): Likewise.
153 (process_switch): Use the newly created class.
154 (switch_conversion::expand): New.
155 (switch_conversion::~switch_conversion): New.
156 * tree-switch-conversion.h: New file.
158 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
160 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
161 tree-vect-patterns.c.
162 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
163 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
164 (vect_recog_sad_pattern): Likewise.
165 (vect_recog_widen_sum_pattern): Likewise.
166 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
167 (vect_recog_widen_shift_pattern): Remove the type_in argument.
168 (vect_recog_rotate_pattern): Likewise.
169 (vect_recog_mult_pattern): Likewise.
170 (vect_recog_vector_vector_shift_pattern): Likewise.
171 (vect_recog_divmod_pattern): Likewise.
172 (vect_recog_mixed_size_cond_pattern): Likewise.
173 (vect_recog_bool_pattern): Likewise.
174 (vect_recog_mask_conversion_pattern): Likewise.
175 (vect_try_gather_scatter_pattern): Likewise.
176 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
177 (vect_recog_over_widening_pattern): Likewise.
178 (vect_recog_gather_scatter_pattern): Likewise.
179 (vect_recog_func_ptr): Move from tree-vectorizer.h
180 (vect_vect_recog_func_ptrs): Move further down the file.
181 (vect_recog_func): Likewise. Remove the third argument.
182 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
183 (vect_pattern_recog_1): Expect the pattern function to do any
184 necessary target tests. Also expect it to provide a vector type.
185 Remove the type_in handling.
187 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
189 * tree-vect-patterns.c (vect_pattern_detected): New function.
190 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
191 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
192 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
193 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
194 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
195 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
196 (vect_recog_mask_conversion_pattern)
197 (vect_try_gather_scatter_pattern): Likewise.
199 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
201 * tree-vect-patterns.c (vect_get_internal_def): New function.
202 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
203 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
204 (search_type_for_mask_1): Use it.
206 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
208 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
209 redundant WIDEN_SUM_EXPR handling.
210 (vect_recog_sad_pattern): Likewise.
212 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
214 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
215 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
216 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
217 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
218 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
220 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
222 * tree-vect-stmts.c (vectorizable_call): Make sure that we
223 use the stmt_vec_info of the original bb statement for the
224 new zero assignment, even if the call is part of a pattern.
226 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
228 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
229 that the sequence is attached to the original statement rather
230 than the pattern statement.
231 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
232 PATTERN_DEF_SEQ from the original statement rather than
233 the main pattern statement.
234 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
235 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
236 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
238 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
240 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
241 definition statements before the early exit for statements that aren't
243 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
245 (vect_transform_loop): ...here. Process pattern definition
246 statements without first checking whether the main pattern
247 statement is live or relevant.
249 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
251 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
252 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
254 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
256 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
257 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
258 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
259 (expand_block_compare): Change select_block_compare_mode call.
260 (expand_strncmp_align_check): Use new functions, fix comment.
261 (emit_final_str_compare_gpr): New function.
262 (expand_strn_compare): Refactor and clean up code.
263 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
265 2018-06-19 Tony Reix <tony.reix@atos.com>
266 Damien Bergamini <damien.bergamini@atos.com>
267 David Edelsohn <dje.gcc@gmail.com>
269 * collect2.c (static_obj): New variable.
270 (static_libs): New variable.
271 (is_in_list): Uncomment declaration.
272 (main): Track AIX libraries linked statically.
273 (is_in_list): Uncomment definition.
274 (scan_prog_file): Don't add AIX shared libraries initializer
275 to constructor list if linking statically.
277 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
279 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
281 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
283 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
285 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
288 2018-06-19 Martin Liska <mliska@suse.cz>
290 * config/i386/i386.c (ix86_can_inline_p): Do not use
291 ipa_fn_summaries::get_create.
292 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
294 (devirtualization_time_bonus): Likewise.
295 (ipcp_propagate_stage): Likewise.
296 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
297 (edge_set_predicate): Likewise.
298 (evaluate_conditions_for_known_args): Likewise.
299 (evaluate_properties_for_edge): Likewise.
300 (ipa_call_summary::reset): Tranform to ...
301 (ipa_call_summary::~ipa_call_summary): ... this.
302 (ipa_fn_summary::reset): Transform to ...
303 (ipa_fn_summary::~ipa_fn_summary): ... this.
304 (ipa_fn_summary_t::remove): Rename to ...
305 (ipa_fn_summary_t::remove_callees): ... this.
306 (ipa_fn_summary_t::duplicate): Use placement new
307 instead of memory copy.
308 (ipa_call_summary_t::duplicate): Likewise.
309 (ipa_call_summary_t::remove): Remove.
310 (dump_ipa_call_summary): Change get_create to get.
311 (ipa_dump_fn_summary): Dump only when summary exists.
312 (analyze_function_body): Use symbol_summary::get instead
314 (compute_fn_summary): Likewise.
315 (estimate_edge_devirt_benefit): Likewise.
316 (estimate_edge_size_and_time): Likewise.
317 (inline_update_callee_summaries): Likewise.
318 (remap_edge_change_prob): Likewise.
319 (remap_edge_summaries): Likewise.
320 (ipa_merge_fn_summary_after_inlining): Likewise.
321 (write_ipa_call_summary): Likewise.
322 (ipa_fn_summary_write): Likewise.
323 (ipa_free_fn_summary): Likewise.
324 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
325 (struct ipa_call_summary): Likewise.
326 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
328 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
329 (estimate_size_after_inlining): Likewise.
330 (estimate_growth): Likewise.
331 (growth_likely_positive): Likewise.
332 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
333 (inline_call): Likewise.
334 * ipa-inline.c (caller_growth_limits): Likewise.
335 (can_inline_edge_p): Likewise.
336 (can_inline_edge_by_limits_p): Likewise.
337 (compute_uninlined_call_time): Likewise.
338 (compute_inlined_call_time): Likewise.
339 (want_inline_small_function_p): Likewise.
340 (edge_badness): Likewise.
341 (update_caller_keys): Likewise.
342 (update_callee_keys): Likewise.
343 (inline_small_functions): Likewise.
344 (inline_to_all_callers_1): Likewise.
345 (dump_overall_stats): Likewise.
346 (early_inline_small_functions): Likewise.
347 (early_inliner): Likewise.
348 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
349 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
350 * ipa-pure-const.c (malloc_candidate_p): Likewise.
351 * ipa-split.c (execute_split_functions): Likewise.
352 * symbol-summary.h: Likewise.
353 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
355 2018-06-19 Richard Biener <rguenther@suse.de>
357 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
358 (vectorize_loops): ... here. Fix dbgcnt handling.
359 (try_vectorize_loop): Wrap try_vectorize_loop_1.
361 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
364 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
365 ieee128 argument takes up only one (vector) register, not two (floating
368 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
370 * gimplify.c (gimplify_init_constructor): Really never clear for an
371 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
373 2018-06-19 Richard Biener <rguenther@suse.de>
375 PR tree-optimization/86179
376 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
377 after failed recognition.
379 2018-06-18 Martin Sebor <msebor@redhat.com>
382 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
383 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
384 Handle integer subtraction.
385 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
386 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
388 2018-06-18 David Malcolm <dmalcolm@redhat.com>
390 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
391 param from rtx to rtx_insn *.
392 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
394 (frv_ifcvt_modify_insn): Likwise.
395 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
396 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
397 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
398 as_a <rtx_insn *> cast to local "unprotected_region" once
399 it's been established that it's not NULL or pc_rtx.
400 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
401 param "sethi" from rtx to rtx_insn *.
402 (nds32_group_float_insns): Likewise for param "insn".
403 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
405 (vax_output_int_subtract): Likewise.
406 * config/vax/vax.c (vax_output_int_add): Likewise for param
408 (vax_output_int_subtract): Likewise.
409 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
410 (emit_pattern_after): Likewise for param "after".
411 (emit_insn_after): Likewise.
412 (emit_jump_insn_after): Likewise.
413 (emit_call_insn_after): Likewise.
414 (emit_debug_insn_after): Likewise.
415 (emit_pattern_before): Likewise for param "before".
416 (emit_insn_before): Likewise.
417 (emit_jump_insn_before): Likewise.
418 * final.c (get_insn_template): Likewise for param "insn", removing
420 * output.h (get_insn_template): Likewise for 2nd param.
421 * rtl.h (emit_insn_before): Likewise.
422 (emit_jump_insn_before): Likewise.
423 (emit_debug_insn_before_noloc): Likewise.
424 (emit_insn_after): Likewise.
425 (emit_jump_insn_after): Likewise.
426 (emit_call_insn_after): Likewise.
427 (emit_debug_insn_after): Likewise.
428 (set_insn_deleted): Likewise for param.
430 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
433 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
434 floating point modes, so that IFmode is numerically greater than
435 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
436 to declare the ordering. This prevents IFmode from being
437 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
438 machine. Include rs6000-modes.h to share the fractional values
439 between genmodes* and the rest of the compiler.
443 * config/rs6000/rs6000-modes.h: New file.
444 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
445 meaning of rs6000_long_double_size so that 126..128 selects an
446 appropriate 128-bit floating point type.
447 (rs6000_option_override_internal): Likewise.
448 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
449 (TARGET_LONG_DOUBLE_128): Change the meaning of
450 rs6000_long_double_size so that 126..128 selects an appropriate
451 128-bit floating point type.
452 (LONG_DOUBLE_TYPE_SIZE): Update comment.
453 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
454 source and destination to match the standard usage.
455 (truncifkf2): Likewise.
456 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
457 ISA 2.07 to use an explicit clobber, instead of passing in a
459 (copysign<mode>3_soft): Likewise.
461 2018-06-18 David Malcolm <dmalcolm@redhat.com>
463 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
464 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
465 (vect_slp_analyze_instance_dependence): Likewise.
466 (vect_enhance_data_refs_alignment): Likewise.
467 (vect_analyze_data_refs_alignment): Likewise.
468 (vect_slp_analyze_and_verify_instance_alignment
469 (vect_analyze_data_ref_accesses): Likewise.
470 (vect_prune_runtime_alias_test_list): Likewise.
471 (vect_analyze_data_refs): Likewise.
472 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
473 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
474 (vect_analyze_scalar_cycles_1): Likewise.
475 (vect_get_loop_niters): Likewise.
476 (vect_analyze_loop_form_1): Likewise.
477 (vect_update_vf_for_slp): Likewise.
478 (vect_analyze_loop_operations): Likewise.
479 (vect_analyze_loop): Likewise.
480 (vectorizable_induction): Likewise.
481 (vect_transform_loop): Likewise.
482 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
483 * tree-vect-slp.c (vect_analyze_slp): Likewise.
484 (vect_make_slp_decision): Likewise.
485 (vect_detect_hybrid_slp): Likewise.
486 (vect_slp_analyze_operations): Likewise.
487 (vect_slp_bb): Likewise.
488 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
489 (vectorizable_bswap): Likewise.
490 (vectorizable_call): Likewise.
491 (vectorizable_simd_clone_call): Likewise.
492 (vectorizable_conversion): Likewise.
493 (vectorizable_assignment): Likewise.
494 (vectorizable_shift): Likewise.
495 (vectorizable_operation): Likewise.
496 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
498 2018-06-18 Martin Sebor <msebor@redhat.com>
500 PR tree-optimization/81384
501 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
502 * builtins.c (expand_builtin_strnlen): New function.
503 (expand_builtin): Call it.
504 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
505 * builtins.def (BUILT_IN_STRNLEN): New.
506 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
507 Warn for bounds in excess of maximum object size.
508 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
509 single-value ranges. Handle strnlen.
510 (handle_builtin_strlen): Handle strnlen.
511 (strlen_check_and_optimize_stmt): Same.
512 * doc/extend.texi (Other Builtins): Document strnlen.
514 2018-06-18 Maya Rashish <coypu@sdf.org>
516 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
517 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
518 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
520 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
522 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
524 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
526 * tree.c (escaped_string::escape): Replace cast to char * by
527 const_cast<char *> (unescaped).
529 2018-06-18 Nick Clifton <nickc@redhat.com>
532 * tree.c (escaped_string): New class. Converts an unescaped
533 string into its escaped equivalent.
534 (warn_deprecated_use): Use the new class to convert the
535 deprecation message, if present.
536 (test_escaped_strings): New self test.
537 (test_c_tests): Add test_escaped_strings.
538 * doc/extend.texi (deprecated): Add a note that the
539 deprecation message is affected by the -fmessage-length
540 option, and that control characters will be escaped.
541 (#pragma GCC error): Document this pragma.
542 (#pragma GCC warning): Likewise.
543 * doc/invoke.texi (-fmessage-length): Document this option's
544 effect on the #warning and #error preprocessor directives and
545 the deprecated attribute.
547 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
549 * tree.c (decl_value_expr_lookup): Revert latest change.
550 (decl_value_expr_insert): Likewise.
552 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
554 * gimplify.c (nonlocal_vlas): Delete.
555 (nonlocal_vla_vars): Likewise.
556 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
558 (gimplify_body): Do not create and destroy nonlocal_vlas.
559 * tree-nested.c: Include diagnostic.h.
560 (use_pointer_in_frame): Tweak.
561 (lookup_field_for_decl): Add assertion and declare the transformation.
562 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
563 internal error when the reference is in a wrong context. Do not
564 create a debug decl by default.
565 (note_nonlocal_block_vlas): Delete.
566 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
567 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
568 create a debug decl by default.
569 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
570 call is in a wrong context.
571 (fixup_vla_decls): New function.
572 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
573 debug variables were created.
574 * tree.c (decl_value_expr_lookup): Add checking assertion.
575 (decl_value_expr_insert): Likewise.
577 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
580 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
581 * tree-scalar-evolution.c (interpret_expr): Likewise.
582 (expression_expensive_p): Likewise.
583 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
584 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
585 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
586 (ssa_defined_by_minus_one_stmt_p): New.
588 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
591 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
592 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
593 * dojump.c (do_jump): Likewise.
594 * expr.c (expand_expr_real_2): Check operand type's sign.
595 * fold-const.c (const_unop): Handle ABSU_EXPR.
596 (fold_abs_const): Likewise.
597 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
598 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
599 (strip_sign_op_1): Likesise.
600 * match.pd: Add new pattern to generate ABSU_EXPR.
601 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
602 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
603 * tree-eh.c (operation_could_trap_helper_p): Likewise.
604 * tree-inline.c (estimate_operator_cost): Likewise.
605 * tree-pretty-print.c (dump_generic_node): Likewise.
606 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
607 * tree.def (ABSU_EXPR): New.
609 2018-06-16 Jakub Jelinek <jakub@redhat.com>
612 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
613 documented as preserved for backward compatibility only.
614 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
616 PR rtl-optimization/86108
617 * bb-reorder.c (create_forwarder_block): Renamed to ...
618 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
619 jump from new landing pad to the second part.
620 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
623 2018-06-15 Jakub Jelinek <jakub@redhat.com>
626 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
627 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
628 Only call store_expr for halves if the mode is the same.
631 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
632 Fix up comment formatting.
634 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
636 * typed-splay-tree.h (typed_splay_tree::remove): New function.
637 (typed_splay_tree::closure,
638 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
639 (typed_splay_tree::typed_splay_tree,
640 typed_splay_tree::operator =): Declared private.
641 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
642 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
643 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
644 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
645 typed_splay_tree::splay_tree_splay,
646 typed_splay_tree::splay_tree_foreach_helper,
647 typed_splay_tree::splay_tree_insert,
648 typed_splay_tree::splay_tree_remove,
649 typed_splay_tree::splay_tree_lookup,
650 typed_splay_tree::splay_tree_predecessor,
651 typed_splay_tree::splay_tree_successor,
652 typed_splay_tree::splay_tree_min,
653 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
654 (typed_splay_tree::root, typed_splay_tree::comp,
655 typed_splay_tree::delete_key,
656 typed_splay_tree::delete_value): New data members.
657 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
658 typed_splay_tree::remove.
660 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
662 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
663 -mginv and -mno-ginv to the assembler.
664 * config/mips/mips.opt (-mcrc): New option.
666 * doc/invoke.text (-mcrc): Document.
669 2018-06-15 Nick Clifton <nickc@redhat.com>
672 * tree.c (escaped_string): New class. Converts an unescaped
673 string into its escaped equivalent.
674 (warn_deprecated_use): Use the new class to convert the
675 deprecation message, if present.
676 (test_escaped_strings): New self test.
677 (test_c_tests): Add test_escaped_strings.
678 * doc/extend.texi (deprecated): Add a note that the
679 deprecation message is affected by the -fmessage-length
680 option, and that control characters will be escaped.
681 (#pragma GCC error): Document this pragma.
682 (#pragma GCC warning): Likewise.
683 * doc/invoke.texi (-fmessage-length): Document this option's
684 effect on the #warning and #error preprocessor directives and
685 the deprecated attribute.
687 2018-06-15 Richard Biener <rguenther@suse.de>
689 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
690 here, also noting vector size used.
691 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
692 size used in MSG_OPTIMIZED_LOCATIONS dump.
693 (pass_slp_vectorize::execute): Adjust.
695 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
698 * config/arc/arc.c (arc_return_address_register): Fix
701 2018-06-15 Richard Biener <rguenther@suse.de>
704 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
705 leave useless conversion stripping to force_gimple_operand_gsi.
706 (gimplify_build2): Likewise.
707 (gimplify_build1): Likewise.
709 2018-06-15 Richard Biener <rguenther@suse.de>
712 * tree-cfg.c (move_stmt_op): unshare invariant addresses
713 before adjusting their block.
715 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
717 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
718 multilibs for *-*-rtems*.
719 * config/riscv/t-rtems: New file.
721 2018-06-14 Jakub Jelinek <jakub@redhat.com>
724 * match.pd ((A +- CST1) +- CST2): Punt if last resort
725 unsigned_type_for returns NULL.
728 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
729 subregs of multi-word pseudos unless the float mode has word size.
731 2018-06-14 Richard Biener <rguenther@suse.de>
734 * tree-vect-generic.c (build_word_mode_vector_type): Remove
735 duplicate and harmful type_hash_canon.
736 * tree.c (type_hash_canon): Assert we didn't find ourselves.
738 2018-06-14 Richard Biener <rguenther@suse.de>
741 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
744 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
746 * config/rtems.h (STDINT_LONG32): Define.
748 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
749 Prachi Godbole <prachi.godbole@imgtec.com>
751 * config/mips/mips-cpus.def: Define P6600.
752 * config/mips/mips-tables.opt: Regenerate.
753 * config/mips/mips.c (mips_ucbranch_type): New enum.
754 (mips_rtx_cost_data): Add support for P6600.
755 (mips_issue_rate): Likewise.
756 (mips_multipass_dfa_lookahead): Likewise.
757 (mips_avoid_hazard): Likewise.
758 (mips_reorg_process_insns): Likewise.
759 (mips_classify_branch_p6600): New function.
760 * config/mips/mips.h (TUNE_P6600): New define.
761 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
762 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
763 * config/mips/mips.md: Include p6600.md.
764 (processor): Add p6600.
765 * config/mips/p6600.md: New file.
766 * doc/invoke.texi: Add p6600 to supported architectures.
768 2018-06-13 Martin Sebor <msebor@redhat.com>
770 PR tree-optimization/86114
771 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
773 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
775 2018-06-13 Richard Biener <rguenther@suse.de>
777 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
778 Properly set vector type of the intermediate stmt.
779 * tree-vect-stmts.c (vectorizable_operation): The destination
780 var always has vectype_out type.
782 2018-06-13 Jeff Law <law@redhat.com>
784 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
785 integer 0 for argument to print_rtl_with_bb.
786 (rl78_reorg): Likewise.
788 2018-06-13 David Malcolm <dmalcolm@redhat.com>
790 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
791 from rtx to rtx_insn *.
792 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
794 (add_sched_insns_for_speculation): Likewise for local "target",
795 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
796 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
797 from rtx_insn ** to rtx_code_label **.
798 (reorg_emit_nops): Likewise.
799 (c6x_reorg): Likewise for local "call_labels".
800 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
802 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
803 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
804 the loops over LABEL_REFS.
805 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
807 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
808 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
809 (split_branches): Strengthen local "olabel" from rtx to
810 rtx_insn *, adding a safe_as_a cast.
811 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
813 (add_insn_after): Likewise for first two params.
814 (add_insn_before): Likewise.
815 (remove_insn): Likewise for param.
816 (emit_pattern_before_noloc): Likewise for second and third params.
817 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
818 (emit_call_insn_before_noloc): Likewise.
819 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
821 (emit_barrier_before): Likewise.
822 (emit_label_before): Strengthen "label" param from "rtx" to
823 "rtx_code_label *". Strengthen "before" param from "rtx" to
825 (emit_insn_after_1): Strengthen "after" param from "rtx" to
827 (emit_pattern_after_noloc): Likewise.
828 (emit_insn_after_noloc): Likewise.
829 (emit_jump_insn_after_noloc): Likewise.
830 (emit_call_insn_after_noloc): Likewise.
831 (emit_debug_insn_after_noloc): Likewise.
832 (emit_barrier_after): Likewise.
833 (emit_label_after): Likewise for both params.
834 (emit_pattern_after_setloc): Likewise for "after" param. Convert
835 "loc" param from "int" to "location_t".
836 (emit_insn_after_setloc): Likewise.
837 (emit_jump_insn_after_setloc): Likewise.
838 (emit_call_insn_after_setloc): Likewise.
839 (emit_debug_insn_after_setloc): Likewise.
840 (emit_pattern_before_setloc): Likewise for "before" param. Convert
841 "loc" param from "int" to "location_t".
842 (emit_pattern_before): Convert NULL_RTX to NULL.
843 (emit_insn_before_setloc): Convert "loc" param from "int" to
845 (emit_jump_insn_before_setloc): Likewise.
846 (emit_call_insn_before_setloc): Likewise.
847 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
848 rtx_insn *. Convert "loc" param from "int" to "location_t".
849 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
850 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
851 Convert 3rd param from "int" to "location_t".
852 (emit_barrier_before, emit_barrier_after, next_real_insn):
853 Strengthen param from rtx to rtx_insn *.
854 (emit_label_before): Strengthen 1st param from "rtx" to
855 "rtx_code_label *". Strengthen 2nd param from "rtx" to
857 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
858 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
859 Strengthen 2nd param from "rtx" to "rtx_insn *".
860 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
861 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
862 Likewise. Convert 3rd param from "int" to "location_t".
863 (emit_label_after): Strengthen 1st param from "rtx" to
865 (next_real_insn, remove_insn): Strengthen param from "rtx" to
867 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
868 from "rtx" to "rtx_insn *".
870 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
872 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
873 bodies streamed in with -Q.
874 * dumpfile.c (dump_files): Add lto-stream-out dump file.
875 * dumpfile.h (tree_dump_index): Add lto_stream_out.
876 * gimple-streamer-out.c: Include gimple-pretty-print.h
877 (output_bb): Dump stmts streamed.
878 * lto-section-out.c: Include print-tree.h
879 (lto_begin_section): Dump sections created.
880 (lto_output_decl_index): Dump decl encoded.
881 * lto-streamer-out.c: Include print-tree.h
882 (create_output_block): Dump output block created.
883 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
884 (output_function): Dump function output.
885 (output_constructor): Dump constructor streamed.
886 (write_global_stream): Output indexes encoded.
887 (produce_asm_for_decls): Dump streams encoded.
888 * lto-streamer.c (streamer_dump_file): New global var.
889 * lto-streamer.h (streamer_dump_file): Declare.
890 * passes.c (ipa_write_summaries): Initialize streamer dump.
891 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
894 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
897 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
898 offsets for register save directives. Emit a second batch of save
899 directives, if need be, when the function accesses prior frames.
901 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
903 * config/arc/fpu.md (fmasf4): Force operand to register.
906 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
908 * config/arc/arc-protos.h (arc_pad_return): Remove.
909 * config/arc/arc.c (machine_function): Remove force_short_suffix
911 (arc_print_operand): Adjust printing of '&'.
912 (arc_verify_short): Remove conditional printing of short suffix.
913 (arc_final_prescan_insn): Remove reference to size_reason.
914 (pad_return): New function.
915 (arc_reorg): Call pad_return.
916 (arc_pad_return): Remove.
917 (arc_init_machine_status): Remove reference to force_short_suffix.
918 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
919 (attr length): When attribute iscompact is true force to 2
920 regardless; in the case of maybe check if we want to force the
921 instruction to have 4 bytes length.
922 (nopv): Change it to generate 4 byte long nop as well.
923 (blockage): New pattern.
924 (simple_return): Remove call to arc_pad_return.
925 (p_return_i): Likewise.
927 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
929 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
931 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
933 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
936 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
938 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
939 for ARC700 and ARCv2.
941 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
944 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
945 operands[2] instead of operands[1].
948 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
950 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
951 case, check whether the outer register overlaps an unallocatable
952 register, not just whether it fits the required class.
954 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
956 * poly-int.h (can_div_trunc_p): Add new overload in which all values
958 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
959 (memrefs_conflict_p): Likewise.
960 (init_alias_analysis): Likewise.
961 * cfgexpand.c (expand_debug_expr): Likewise.
962 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
963 * cse.c (fold_rtx): Likewise.
964 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
965 * expr.c (emit_block_move_hints): Likewise.
966 (clear_storage_hints, push_block, emit_push_insn): Likewise.
967 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
968 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
969 (emit_group_store): Likewise.
970 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
971 to read the PRE/POST_MODIFY increment.
972 * calls.c (store_one_arg): Use strip_offset.
973 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
975 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
977 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
978 (simplify_binary_operation_1): Extend CONST_INT handling to
980 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
981 than a HOST_WIDE_INT.
982 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
984 (adjust_mems, add_stores): Update accodingly.
985 (vt_canonicalize_addr): Track polynomial offsets.
986 (emit_note_insn_var_location): Likewise.
987 (vt_add_function_parameter): Likewise.
988 (vt_initialize): Likewise.
990 2018-06-12 Jeff Law <law@redhat.com>
992 * config.gcc (alpha*-*-freebsd*): Remove.
993 * config/alpha/freebsd.h: Remove.
995 2018-06-12 David Malcolm <dmalcolm@redhat.com>
998 * spellcheck-tree.c (levenshtein_distance): Rename to...
999 (get_edit_distance): ...this, and update for underlying renaming.
1000 * spellcheck-tree.h (levenshtein_distance): Rename to...
1001 (get_edit_distance): ...this.
1002 * spellcheck.c (levenshtein_distance): Rename to...
1003 (get_edit_distance): ...this. Convert from Levenshtein distance
1004 to Damerau-Levenshtein distance by supporting transpositions of
1005 adjacent characters. Rename "v1" to "v_next" and "v0" to
1007 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1008 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1009 update for underlying renaming.
1010 (selftest::levenshtein_distance_unit_test): Rename to...
1011 (selftest::test_get_edit_distance_unit): ...this, and update for
1012 underlying renaming.
1013 (selftest::test_find_closest_string): Add example from PR 69968
1014 where transposition helps
1015 (selftest::test_metric_conditions): Update for renaming.
1016 (selftest::test_metric_conditions): Likewise.
1017 (selftest::spellcheck_c_tests): Likewise.
1018 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1019 (get_edit_distance): ...this.
1020 (best_match::consider): Update for renaming.
1022 2018-06-12 Martin Sebor <msebor@redhat.com>
1024 PR tree-optimization/85259
1025 * builtins.c (compute_objsize): Handle constant offsets.
1026 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1027 true iff a warning has been issued.
1028 * gimple.h (gimple_nonartificial_location): New function.
1029 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1030 gimple_nonartificial_location and handle -Wno-system-headers.
1031 (handle_builtin_stxncpy): Same.
1033 2018-06-12 Martin Sebor <msebor@redhat.com>
1036 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1038 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1040 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1041 BUILTIN_VEC_XST entries for pointer to double and long long.
1043 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1046 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1048 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1051 2018-06-12 Martin Liska <mliska@suse.cz>
1053 * doc/options.texi: Document IntegerRange.
1055 2018-06-12 Martin Liska <mliska@suse.cz>
1057 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1058 * opt-functions.awk: Handle Deprecated flag.
1059 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1061 (read_cmdline_option): Report warning for a deprecated option.
1062 * opts.h (struct cl_option): Add new field cl_deprecated.
1063 (CL_ERR_DEPRECATED): New.
1065 2018-06-12 Martin Liska <mliska@suse.cz>
1067 * doc/options.texi: Document Deprecated option flag.
1069 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1071 * config/arc/arc-arch.h (arc_extras): New enum.
1072 (arc_cpu_t):Add field extra.
1073 (arc_cpu_types): Consider the extras.
1074 * config/arc/arc-cpus.def: Add extras info.
1075 * config/arc/arc-opts.h (processor_type): Consider extra field.
1076 * config/arc/arc.c (arc_override_options): Handle extra field.
1078 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1080 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1081 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1083 * config/arc/arc.c (arc_init): Update pic warning.
1084 (irq_range): Update irq range parsing warnings.
1085 (arc_override_options): Update various warning messages.
1086 (arc_handle_aux_attribute): Likewise.
1088 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1090 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1092 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1094 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1095 compiler messages directives.
1097 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1099 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1100 * config/mips/mips-tables.opt: Regenerate.
1101 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1103 * doc/invoke.texi: Document -march=i6500.
1105 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1107 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1108 (i6400_gpmul): Add cpu_unit.
1109 (i6400_gpdiv): Likewise.
1110 (i6400_msa_add_d): Update reservations.
1111 (i6400_msa_int_add) Likewise.
1112 (i6400_msa_short_logic3) Likewise.
1113 (i6400_msa_short_logic2) Likewise.
1114 (i6400_msa_short_logic) Likewise.
1115 (i6400_msa_move) Likewise.
1116 (i6400_msa_cmp) Likewise.
1117 (i6400_msa_short_float2) Likewise.
1118 (i6400_msa_div_d) Likewise.
1119 (i6400_msa_long_logic1) Likewise.
1120 (i6400_msa_long_logic2) Likewise.
1121 (i6400_msa_mult) Likewise.
1122 (i6400_msa_long_float2) Likewise.
1123 (i6400_msa_long_float4) Likewise.
1124 (i6400_msa_long_float5) Likewise.
1125 (i6400_msa_long_float8) Likewise.
1126 (i6400_fpu_fadd): Include frint type.
1127 (i6400_fpu_store): New define_insn_reservation.
1128 (i6400_fpu_load): Likewise.
1129 (i6400_fpu_move): Likewise.
1130 (i6400_fpu_fcmp): Likewise.
1131 (i6400_fpu_fmadd): Likewise.
1132 (i6400_int_mult): Include imul3nc type and update reservation.
1133 (i6400_int_div): Include idiv3 type and update reservation.
1134 (i6400_int_load): Update to check type not move_type.
1135 (i6400_int_store): Likewise.
1136 (i6400_int_prefetch): Set zero latency.
1138 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1140 * gcc.c: Document new %@{...} sequence.
1141 (LINK_COMMAND_SPEC): Use it for the -L switches.
1142 (cpp_unique_options): Use it for the -I switches.
1143 (at_file_argbuf): New global variable.
1144 (in_at_file): Likewise.
1145 (alloc_args): Create at_file_argbuf.
1146 (clear_args): Truncate at_file_argbuf.
1147 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1148 (open_at_file): New function.
1149 (close_at_file): Likewise.
1150 (create_at_file): Delete.
1151 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1154 (validate_switches_from_spec): Deal with %@{...} sequence.
1155 (validate_switches): Likewise.
1156 (driver::finalize): Call clear_args.
1158 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1160 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1162 2018-06-11 Martin Sebor <msebor@redhat.com>
1164 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1165 (Wno-class-memaccess): Add @opindex.
1166 (Wno-templates, Wno-multiple-inheritance): Same.
1167 (Wno-virtual-inheritance, Wno-namespaces): Same.
1168 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1169 (Wno-format-overflow, Wno-format-truncation): Same.
1170 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1171 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1172 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1173 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1174 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1175 (Wno-misspelled-isr): Same.
1177 2018-06-11 Martin Sebor <msebor@redhat.com>
1179 * PR tree-optimization/86083
1180 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1182 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1184 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1186 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1189 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1190 on the correct operand.
1191 (*movdi_internal64): Ditto.
1193 2018-06-11 Martin Liska <mliska@suse.cz>
1195 PR tree-optimization/86089
1196 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1198 2018-06-11 Julia Koval <julia.koval@intel.com>
1200 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1201 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1202 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1204 2018-06-11 Olivier Hainque <hainque@adacore.com>
1206 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1207 for Ada with strict dwarf2.
1209 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1212 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1215 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1217 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1219 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1221 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1224 2018-06-08 Martin Liska <mliska@suse.cz>
1226 * tree-cfg.h (debug_function): Fix argument type to match
1229 2018-06-08 Martin Liska <mliska@suse.cz>
1231 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1232 Remove usage of MPX-related (and removed) fields.
1233 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1235 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1237 * cfg.c (debug): Use TDF_NONE rather than 0.
1238 * cfghooks.c (debug): Likewise.
1239 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1240 (struct dump_option_value_info): Convert to...
1241 (struct kv_pair): ...this template type.
1242 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1244 (optinfo_verbosity_options): Likewise.
1245 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1247 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1248 than int for "optgroup_flags" param.
1249 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1251 (dump_dec): Likewise.
1252 (dump_finish): Use TDF_NONE rather than 0.
1253 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1254 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1255 than 0. Update for change to option_ptr.
1256 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1257 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1258 0. Update for changes to optinfo_verbosity_options and
1260 (opt_info_switch_p): Convert optgroup_flags from int to
1262 (dump_basic_block): Use dump_flags_t rather than int
1263 for "dump_kind" param.
1264 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1265 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1266 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1267 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1268 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1269 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1270 TDF_NONE): Convert from macros to...
1271 (enum dump_flag): ...this new enum.
1272 (dump_flags_t): Update to use enum.
1273 (operator|, operator&, operator~, operator|=, operator&=):
1274 Implement for dump_flags_t.
1275 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1276 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1277 Convert from macros to...
1278 (enum optgroup_flag): ...this new enum.
1279 (optgroup_flags_t): New typedef.
1280 (operator|, operator|=): Implement for optgroup_flags_t.
1281 (struct dump_file_info): Convert field "alt_flags" to
1282 dump_flags_t. Convert field "optgroup_flags" to
1284 (dump_basic_block): Use dump_flags_t rather than int for param.
1285 (dump_generic_expr_loc): Likewise.
1286 (dump_dec): Likewise.
1287 (dump_register): Convert param "optgroup_flags" to
1289 (opt_info_enable_passes): Likewise.
1290 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1292 * gimple-pretty-print.c (debug): Likewise.
1293 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1294 (merged_store_group::apply_stores): Likewise.
1295 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1296 * gimple.c (verify_gimple_pp): Likewise.
1297 * graphite-poly.c (print_pbb_body): Likewise.
1298 * passes.c (pass_manager::register_one_dump_file): Convert
1299 local "optgroup_flags" to optgroup_flags_t.
1300 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1302 (debug_body): Likewise.
1303 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1304 to optgroup_flags_t.
1305 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1307 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1308 (convert_mult_to_fma): Likewise.
1309 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1310 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1311 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1312 "dump_kind" to dump_flags_t.
1314 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1316 * config/rs6000/rs6000.c (min, max): Delete.
1318 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1320 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1321 -mabi=spe and -mabi=no-spe.
1323 2018-06-08 Martin Liska <mliska@suse.cz>
1325 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1326 where we expect an existing summary.
1328 2018-06-08 Martin Liska <mliska@suse.cz>
1330 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1331 * ipa-inline.h (estimate_edge_growth): Likewise.
1333 2018-06-08 Martin Liska <mliska@suse.cz>
1335 * cgraph.c (function_version_hasher::hash): Use
1336 cgraph_node::get_uid ().
1337 (function_version_hasher::equal):
1338 * cgraph.h (cgraph_node::get_uid): New method.
1339 * ipa-inline.c (update_caller_keys): Use
1340 cgraph_node::get_uid ().
1341 (update_callee_keys): Likewise.
1342 * ipa-utils.c (searchc): Likewise.
1343 (ipa_reduced_postorder): Likewise.
1344 * lto-cgraph.c (input_node): Likewise.
1345 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1346 * symbol-summary.h (symtab_insertion): Likewise.
1347 (symtab_removal): Likewise.
1348 (symtab_duplication): Likewise.
1349 * tree-pretty-print.c (dump_function_header): Likewise.
1350 * tree-sra.c (convert_callers_for_node): Likewise.
1352 2018-06-08 Martin Liska <mliska@suse.cz>
1354 * cgraph.c (symbol_table::create_edge): Always assign a new
1356 (symbol_table::free_edge): Do not recycle numbers.
1357 * cgraph.h (cgraph_edge::get): New method.
1358 * symbol-summary.h (symtab_removal): Use it.
1359 (symtab_duplication): Likewise.
1360 (call_summary::hashable_uid): Remove.
1362 2018-06-08 Martin Liska <mliska@suse.cz>
1364 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1365 (initialize_growth_caches): Remove.
1366 (free_growth_caches): Likewise.
1367 (do_estimate_edge_time): Use edge_growth_cache.
1368 (do_estimate_edge_size): Likewise.
1369 (do_estimate_edge_hints): Likewise.
1370 * ipa-inline.c (reset_edge_caches): Likewise.
1371 (recursive_inlining): Likewise.
1372 (inline_small_functions): Likewise.
1373 * ipa-inline.h (initialize_growth_caches): Remove.
1374 (estimate_edge_size): Likewise.
1375 (estimate_edge_time): Likewise.
1376 (estimate_edge_hints): Likewise.
1377 (reset_edge_growth_cache): Likewise.
1378 * symbol-summary.h (call_summary::remove): New method.
1380 2018-06-08 Martin Liska <mliska@suse.cz>
1382 * ipa-cp.c (class edge_clone_summary): New summary.
1383 (grow_edge_clone_vectors): Remove.
1384 (ipcp_edge_duplication_hook): Remove.
1385 (class edge_clone_summary_t): New call_summary class.
1386 (ipcp_edge_removal_hook): Remove.
1387 (edge_clone_summary_t::duplicate): New function.
1388 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1389 (create_specialized_node): Likewise.
1390 (ipcp_driver): Initialize edge_clone_summaries and do not
1393 2018-06-08 Martin Liska <mliska@suse.cz>
1395 * symbol-summary.h (get): New function.
1396 (call_summary::m_initialize_when_cloning): New class member.
1398 2018-06-08 Martin Liska <mliska@suse.cz>
1400 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1401 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1402 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1403 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1404 (remove_cgraph_node_from_order): Use the removed_nodes set.
1405 (do_per_function_toporder): Likwise.
1406 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1407 instead of summary_uid.
1408 (symtab_removal): Likewise.
1409 (symtab_duplication): Likewise.
1411 2018-06-08 Martin Liska <mliska@suse.cz>
1413 * ipa-cp.c (ipcp_store_bits_results): Use
1414 ipcp_transformation_sum.
1415 (ipcp_store_vr_results): Likewise.
1416 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1418 (ipcp_transformation_initialize): ... this.
1419 (ipa_set_node_agg_value_chain):
1420 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1421 (write_ipcp_transformation_info): Likewise.
1422 (read_ipcp_transformation_info): Likewise.
1423 (ipcp_update_bits): Likewise.
1424 (ipcp_update_vr): Likewise.
1425 (ipcp_transform_function): Likewise.
1426 * ipa-prop.h: Rename ipcp_transformation_summary to
1427 ipcp_transformation.
1428 (class ipcp_transformation_t): New function summary.
1429 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1430 (ipa_get_agg_replacements_for_node): Likewise.
1432 2018-06-08 Martin Liska <mliska@suse.cz>
1434 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1436 (class funct_state_summary_t): New function_summary class.
1437 (has_function_state): Remove.
1438 (get_function_state): Likewise.
1439 (set_function_state): Likewise.
1440 (add_new_function): Likewise.
1441 (funct_state_summary_t::insert): New function.
1442 (duplicate_node_data): Remove.
1443 (remove_node_data): Remove.
1444 (funct_state_summary_t::duplicate): New function.
1445 (register_hooks): Create new funct_state_summaries.
1446 (pure_const_generate_summary): Use it.
1447 (pure_const_write_summary): Likewise.
1448 (pure_const_read_summary): Likewise.
1449 (propagate_pure_const): Likewise.
1450 (propagate_nothrow): Likewise.
1451 (dump_malloc_lattice): Likewise.
1452 (propagate_malloc): Likewise.
1453 (execute): Do not register hooks, just remove summary
1455 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1458 2018-06-08 Martin Liska <mliska@suse.cz>
1460 * ipa-reference.c (remove_node_data): Remove.
1461 (duplicate_node_data): Likewise.
1462 (class ipa_ref_var_info_summary_t): New class.
1463 (class ipa_ref_opt_summary_t): Likewise.
1464 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1465 (get_reference_optimization_summary): Use
1466 ipa_ref_opt_sum_summaries.
1467 (set_reference_vars_info): Remove.
1468 (set_reference_optimization_summary): Likewise.
1469 (ipa_init): Create summaries.
1470 (init_function_info): Use function summary.
1471 (ipa_ref_opt_summary_t::duplicate): New function.
1472 (ipa_ref_opt_summary_t::remove): New function.
1473 (get_read_write_all_from_node): Fix GNU coding style.
1474 (propagate): Use function summary.
1475 (write_node_summary_p): Fix GNU coding style.
1476 (stream_out_bitmap): Likewise.
1477 (ipa_reference_read_optimization_summary): Use function summary.
1478 (ipa_reference_c_finalize): Do not release hooks.
1480 2018-06-08 Martin Liska <mliska@suse.cz>
1482 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1483 (analyze_function_body): Extract multiple calls of get_create.
1484 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1485 * ipa-inline.c (recursive_inlining): Use ::get method.
1486 * ipa-inline.h (estimate_edge_growth): Likewise.
1488 2018-06-08 Martin Liska <mliska@suse.cz>
1490 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1492 (hsa_function_summary::hsa_function_summary): Use the new enum
1494 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1495 * hsa-gen.c (hsa_get_host_function): Likewise.
1496 (get_brig_function_name): Likewise.
1497 * ipa-hsa.c (process_hsa_functions): Likewise.
1498 (ipa_hsa_write_summary): Likewise.
1499 * symbol-summary.h (symtab_duplication): Use ::get function/
1500 (get): New function.
1502 2018-06-08 Martin Liska <mliska@suse.cz>
1504 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1506 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1507 (hsa_register_kernel): Likewise.
1508 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1509 * hsa-gen.c (hsa_get_host_function): Likewise.
1510 (get_brig_function_name): Likewise.
1511 (generate_hsa): Likewise.
1512 (pass_gen_hsail::execute): Likewise.
1513 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1514 (devirtualization_time_bonus): Likewise.
1515 (ipcp_propagate_stage): Likewise.
1516 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1517 (edge_set_predicate): Likewise.
1518 (evaluate_conditions_for_known_args): Likewise.
1519 (evaluate_properties_for_edge): Likewise.
1520 (ipa_fn_summary::reset): Likewise.
1521 (ipa_fn_summary_t::duplicate): Likewise.
1522 (dump_ipa_call_summary): Likewise.
1523 (ipa_dump_fn_summary): Likewise.
1524 (analyze_function_body): Likewise.
1525 (compute_fn_summary): Likewise.
1526 (estimate_edge_devirt_benefit): Likewise.
1527 (estimate_edge_size_and_time): Likewise.
1528 (estimate_calls_size_and_time): Likewise.
1529 (estimate_node_size_and_time): Likewise.
1530 (inline_update_callee_summaries): Likewise.
1531 (remap_edge_change_prob): Likewise.
1532 (remap_edge_summaries): Likewise.
1533 (ipa_merge_fn_summary_after_inlining): Likewise.
1534 (ipa_update_overall_fn_summary): Likewise.
1535 (read_ipa_call_summary): Likewise.
1536 (inline_read_section): Likewise.
1537 (write_ipa_call_summary): Likewise.
1538 (ipa_fn_summary_write): Likewise.
1539 (ipa_free_fn_summary): Likewise.
1540 * ipa-hsa.c (process_hsa_functions): Likewise.
1541 (ipa_hsa_write_summary): Likewise.
1542 (ipa_hsa_read_section): Likewise.
1543 * ipa-icf.c (sem_function::merge): Likewise.
1544 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1545 (do_estimate_edge_time): Likewise.
1546 (estimate_size_after_inlining): Likewise.
1547 (estimate_growth): Likewise.
1548 (growth_likely_positive): Likewise.
1549 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1550 (inline_call): Likewise.
1551 * ipa-inline.c (caller_growth_limits): Likewise.
1552 (can_inline_edge_p): Likewise.
1553 (can_inline_edge_by_limits_p): Likewise.
1554 (compute_uninlined_call_time): Likewise.
1555 (compute_inlined_call_time): Likewise.
1556 (want_inline_small_function_p): Likewise.
1557 (edge_badness): Likewise.
1558 (update_caller_keys): Likewise.
1559 (update_callee_keys): Likewise.
1560 (recursive_inlining): Likewise.
1561 (inline_small_functions): Likewise.
1562 (inline_to_all_callers_1): Likewise.
1563 (dump_overall_stats): Likewise.
1564 (early_inline_small_functions): Likewise.
1565 (early_inliner): Likewise.
1566 * ipa-inline.h (estimate_edge_growth): Likewise.
1567 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1568 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1569 * ipa-prop.h (IPA_NODE_REF): Likewise.
1570 (IPA_EDGE_REF): Likewise.
1571 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1572 (propagate_malloc): Likewise.
1573 * ipa-split.c (execute_split_functions): Likewise.
1574 * symbol-summary.h: Rename get to get_create.
1576 (get_create): Likewise.
1577 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1579 2018-06-08 Martin Liska <mliska@suse.cz>
1581 * symbol-summary.h (release): Move definition out of class
1583 (symtab_removal): Likewise.
1584 (symtab_duplication): Likewise.
1586 2018-06-08 Martin Liska <mliska@suse.cz>
1588 * symbol-summary.h (function_summary): Move constructor
1589 implementation out of class declaration.
1590 (release): Likewise.
1591 (symtab_insertion): Likewise.
1592 (symtab_removal): Likewise.
1593 (symtab_duplication): Likewise.
1596 2018-06-08 Martin Liska <mliska@suse.cz>
1598 * Makefile.in: Remove support for MPX (macros, related functions,
1599 fields in cgraph_node, ...).
1600 * builtin-types.def (BT_BND): Likewise.
1601 (BT_FN_BND_CONST_PTR): Likewise.
1602 (BT_FN_CONST_PTR_BND): Likewise.
1603 (BT_FN_VOID_PTR_BND): Likewise.
1604 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1605 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1606 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1607 (expand_builtin_mempcpy_with_bounds): Likewise.
1608 (expand_builtin_memset_with_bounds): Likewise.
1609 (expand_builtin_memset_args): Likewise.
1610 (std_expand_builtin_va_start): Likewise.
1611 (expand_builtin): Likewise.
1612 (expand_builtin_with_bounds): Likewise.
1613 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1614 (DEF_LIB_BUILTIN_CHKP): Likewise.
1615 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1616 (DEF_CHKP_BUILTIN): Likewise.
1617 (BUILT_IN_MEMCPY): Likewise.
1618 (BUILT_IN_MEMMOVE): Likewise.
1619 (BUILT_IN_MEMPCPY): Likewise.
1620 (BUILT_IN_MEMSET): Likewise.
1621 (BUILT_IN_STPCPY): Likewise.
1622 (BUILT_IN_STRCAT): Likewise.
1623 (BUILT_IN_STRCHR): Likewise.
1624 (BUILT_IN_STRCPY): Likewise.
1625 (BUILT_IN_STRLEN): Likewise.
1626 (BUILT_IN_MEMCPY_CHK): Likewise.
1627 (BUILT_IN_MEMMOVE_CHK): Likewise.
1628 (BUILT_IN_MEMPCPY_CHK): Likewise.
1629 (BUILT_IN_MEMSET_CHK): Likewise.
1630 (BUILT_IN_STPCPY_CHK): Likewise.
1631 (BUILT_IN_STRCAT_CHK): Likewise.
1632 (BUILT_IN_STRCPY_CHK): Likewise.
1633 * calls.c (store_bounds): Likewise.
1634 (emit_call_1): Likewise.
1635 (special_function_p): Likewise.
1636 (maybe_warn_nonstring_arg): Likewise.
1637 (initialize_argument_information): Likewise.
1638 (finalize_must_preallocate): Likewise.
1639 (compute_argument_addresses): Likewise.
1640 (expand_call): Likewise.
1641 * cfgexpand.c (expand_call_stmt): Likewise.
1642 (expand_return): Likewise.
1643 (expand_gimple_stmt_1): Likewise.
1644 (pass_expand::execute): Likewise.
1645 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1646 (cgraph_node::remove): Likewise.
1647 (cgraph_node::dump): Likewise.
1648 (cgraph_node::verify_node): Likewise.
1649 * cgraph.h (chkp_function_instrumented_p): Likewise.
1650 (symtab_node::get_alias_target): Likewise.
1651 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1652 (cgraph_local_p): Likewise.
1653 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1654 (cgraph_edge::rebuild_references): Likewise.
1655 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1656 (walk_polymorphic_call_targets): Likewise.
1657 (cgraph_node::expand_thunk): Likewise.
1658 (symbol_table::output_weakrefs): Likewise.
1659 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1660 (ix86_handle_option): Likewise.
1661 * config/i386/constraints.md: Likewise.
1662 * config/i386/i386-builtin-types.def (BND): Likewise.
1666 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1667 (BDESC_FIRST): Likewise.
1669 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1670 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1671 * config/i386/i386.c (enum reg_class): Likewise.
1672 (ix86_target_string): Likewise.
1673 (ix86_option_override_internal): Likewise.
1674 (ix86_conditional_register_usage): Likewise.
1675 (ix86_valid_target_attribute_inner_p): Likewise.
1676 (ix86_set_indirect_branch_type): Likewise.
1677 (ix86_set_current_function): Likewise.
1678 (ix86_function_arg_regno_p): Likewise.
1679 (init_cumulative_args): Likewise.
1680 (ix86_function_arg_advance): Likewise.
1681 (ix86_function_arg): Likewise.
1682 (ix86_pass_by_reference): Likewise.
1683 (ix86_function_value_regno_p): Likewise.
1684 (ix86_function_value_1): Likewise.
1685 (ix86_function_value_bounds): Likewise.
1686 (ix86_return_in_memory): Likewise.
1687 (ix86_setup_incoming_vararg_bounds): Likewise.
1688 (ix86_va_start): Likewise.
1689 (indirect_thunk_need_prefix): Likewise.
1690 (print_reg): Likewise.
1691 (ix86_print_operand): Likewise.
1692 (ix86_expand_call): Likewise.
1693 (ix86_output_function_return): Likewise.
1694 (reg_encoded_number): Likewise.
1695 (BDESC_VERIFYS): Likewise.
1696 (ix86_init_mpx_builtins): Likewise.
1697 (ix86_init_builtins): Likewise.
1698 (ix86_emit_cmove): Likewise.
1699 (ix86_emit_move_max): Likewise.
1700 (ix86_expand_builtin): Likewise.
1701 (ix86_builtin_mpx_function): Likewise.
1702 (ix86_get_arg_address_for_bt): Likewise.
1703 (ix86_load_bounds): Likewise.
1704 (ix86_store_bounds): Likewise.
1705 (ix86_load_returned_bounds): Likewise.
1706 (ix86_store_returned_bounds): Likewise.
1707 (ix86_class_likely_spilled_p): Likewise.
1708 (ix86_hard_regno_mode_ok): Likewise.
1709 (x86_order_regs_for_local_alloc): Likewise.
1710 (ix86_mitigate_rop): Likewise.
1711 (ix86_bnd_prefixed_insn_p): Likewise.
1712 (ix86_mpx_bound_mode): Likewise.
1713 (ix86_make_bounds_constant): Likewise.
1714 (ix86_initialize_bounds): Likewise.
1715 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1716 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1717 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1718 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1719 (TARGET_CHKP_BOUND_MODE): Likewise.
1720 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1721 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1722 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1723 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1724 * config/i386/i386.h (TARGET_MPX): Likewise.
1725 (TARGET_MPX_P): Likewise.
1726 (VALID_BND_REG_MODE): Likewise.
1727 (FIRST_BND_REG): Likewise.
1728 (LAST_BND_REG): Likewise.
1729 (enum reg_class): Likewise.
1730 (BND_REG_P): Likewise.
1731 (BND_REGNO_P): Likewise.
1732 (BNDmode): Likewise.
1733 (ADJUST_INSN_LENGTH): Likewise.
1734 * config/i386/i386.md: Likewise.
1735 * config/i386/i386.opt: Likewise.
1736 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1737 (defined): Likewise.
1738 (LINK_MPX): Likewise.
1739 (MPX_SPEC): Likewise.
1740 (LIBMPX_SPEC): Likewise.
1741 (LIBMPXWRAPPERS_SPEC): Likewise.
1742 (CHKP_SPEC): Likewise.
1743 * config/i386/predicates.md: Likewise.
1744 * dbxout.c (dbxout_type): Likewise.
1745 * doc/extend.texi: Likewise.
1746 * doc/invoke.texi: Likewise.
1747 * doc/md.texi: Likewise.
1748 * doc/tm.texi: Likewise.
1749 * doc/tm.texi.in: Likewise.
1750 * dwarf2out.c (is_base_type): Likewise.
1751 (gen_formal_types_die): Likewise.
1752 (gen_subprogram_die): Likewise.
1753 (gen_type_die_with_usage): Likewise.
1754 (gen_decl_die): Likewise.
1755 (dwarf2out_late_global_decl): Likewise.
1756 * expr.c (expand_assignment): Likewise.
1757 (emit_storent_insn): Likewise.
1758 (store_expr_with_bounds): Likewise.
1759 (store_expr): Likewise.
1760 (expand_expr_real_1): Likewise.
1761 * expr.h (store_expr_with_bounds): Likewise.
1762 * function.c (use_register_for_decl): Likewise.
1763 (struct bounds_parm_data): Likewise.
1764 (assign_parms_augmented_arg_list): Likewise.
1765 (assign_parm_find_entry_rtl): Likewise.
1766 (assign_parm_is_stack_parm): Likewise.
1767 (assign_parm_load_bounds): Likewise.
1768 (assign_bounds): Likewise.
1769 (assign_parms): Likewise.
1770 (expand_function_start): Likewise.
1771 * gcc.c (CHKP_SPEC): Likewise.
1772 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1773 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1774 (wrestrict_dom_walker::check_call): Likewise.
1775 * gimple.c (gimple_build_call_from_tree): Likewise.
1776 * gimple.h (enum gf_mask): Likewise.
1777 (gimple_call_with_bounds_p): Likewise.
1778 (gimple_call_set_with_bounds): Likewise.
1779 * gimplify.c (gimplify_init_constructor): Likewise.
1780 * ipa-cp.c (initialize_node_lattices): Likewise.
1781 (propagate_constants_across_call): Likewise.
1782 (find_more_scalar_values_for_callers_subset): Likewise.
1783 * ipa-hsa.c (process_hsa_functions): Likewise.
1784 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1785 * ipa-icf.c (sem_function::merge): Likewise.
1786 * ipa-inline.c (early_inliner): Likewise.
1787 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1788 (warn_function_cold): Likewise.
1789 (propagate_pure_const): Likewise.
1790 * ipa-ref.h (enum GTY): Likewise.
1791 * ipa-split.c (find_retbnd): Likewise.
1792 (consider_split): Likewise.
1793 (split_function): Likewise.
1794 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1795 * ipa.c (walk_polymorphic_call_targets): Likewise.
1796 (symbol_table::remove_unreachable_nodes): Likewise.
1797 (process_references): Likewise.
1798 (cgraph_build_static_cdtor_1): Likewise.
1799 * lto-cgraph.c (lto_output_node): Likewise.
1800 (output_refs): Likewise.
1801 (compute_ltrans_boundary): Likewise.
1802 (input_overwrite_node): Likewise.
1803 (input_node): Likewise.
1804 (input_cgraph_1): Likewise.
1805 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1806 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1807 (class pass_chkp_instrumentation_passes): Likewise.
1808 (make_pass_chkp_instrumentation_passes): Likewise.
1809 * passes.def: Likewise.
1810 * rtl.h (struct GTY): Likewise.
1811 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1812 * stor-layout.c (layout_type): Likewise.
1813 * symtab.c: Likewise.
1814 * target.def: Likewise.
1815 * targhooks.c (default_chkp_bound_type): Likewise.
1816 (default_chkp_bound_mode): Likewise.
1817 (default_builtin_chkp_function): Likewise.
1818 (default_chkp_function_value_bounds): Likewise.
1819 (default_chkp_make_bounds_constant): Likewise.
1820 (default_chkp_initialize_bounds): Likewise.
1821 * targhooks.h (default_chkp_bound_type): Likewise.
1822 (default_chkp_bound_mode): Likewise.
1823 (default_builtin_chkp_function): Likewise.
1824 (default_chkp_function_value_bounds): Likewise.
1825 (default_chkp_make_bounds_constant): Likewise.
1826 (default_chkp_initialize_bounds): Likewise.
1827 * toplev.c (compile_file): Likewise.
1828 (process_options): Likewise.
1829 * tree-core.h (DEF_BUILTIN): Likewise.
1830 (DEF_BUILTIN_CHKP): Likewise.
1831 * tree-inline.c (declare_return_variable): Likewise.
1832 (remap_gimple_stmt): Likewise.
1833 (copy_bb): Likewise.
1834 (initialize_inlined_parameters): Likewise.
1835 (expand_call_inline): Likewise.
1836 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1837 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1838 (make_pass_ipa_chkp_produce_thunks): Likewise.
1839 (make_pass_chkp): Likewise.
1840 (make_pass_chkp_opt): Likewise.
1841 (make_pass_chkp_instrumentation_passes): Likewise.
1842 * tree-pretty-print.c (dump_generic_node): Likewise.
1843 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1844 * tree-ssa-dce.c (propagate_necessity): Likewise.
1845 (eliminate_unnecessary_stmts): Likewise.
1846 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1847 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1848 * tree-ssa-sccvn.h: Likewise.
1849 * tree-ssa-strlen.c (get_string_length): Likewise.
1850 (valid_builtin_call): Likewise.
1851 (adjust_last_stmt): Likewise.
1852 (handle_builtin_strchr): Likewise.
1853 (handle_builtin_strcpy): Likewise.
1854 (handle_builtin_stxncpy): Likewise.
1855 (handle_builtin_memcpy): Likewise.
1856 (handle_builtin_strcat): Likewise.
1857 (strlen_check_and_optimize_stmt): Likewise.
1858 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1859 * tree-streamer-in.c: Likewise.
1860 * tree-streamer.c (record_common_node): Likewise.
1861 * tree.c (tree_code_size): Likewise.
1862 (wide_int_to_tree_1): Likewise.
1863 (type_contains_placeholder_1): Likewise.
1864 (build_common_tree_nodes): Likewise.
1865 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1866 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1867 (POINTER_BOUNDS_P): Likewise.
1868 (BOUNDED_TYPE_P): Likewise.
1869 (BOUNDED_P): Likewise.
1870 (CALL_WITH_BOUNDS_P): Likewise.
1871 (pointer_bounds_type_node): Likewise.
1872 * value-prof.c (gimple_ic): Likewise.
1873 * var-tracking.c (vt_add_function_parameters): Likewise.
1874 * varasm.c (make_decl_rtl): Likewise.
1875 (assemble_start_function): Likewise.
1876 (output_constant): Likewise.
1877 (maybe_assemble_visibility): Likewise.
1878 * varpool.c (ctor_for_folding): Likewise.
1879 * chkp-builtins.def: Remove.
1880 * ipa-chkp.c: Remove.
1881 * ipa-chkp.h: Remove.
1882 * rtl-chkp.c: Remove.
1883 * rtl-chkp.h: Remove.
1884 * tree-chkp-opt.c: Remove.
1885 * tree-chkp.c: Remove.
1886 * tree-chkp.h: Remove.
1888 2018-06-07 Carl Love <cel@us.ibm.com>
1890 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1891 vextract_fp_from_shortl): Add BE support.
1893 2018-06-07 Paul Koning <ni1d@arrl.net>
1895 * compare-elim.c (try_merge_compare): Don't merge compare if
1896 address contains a side effect.
1897 (try_eliminate_compare): Likewise.
1899 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1901 * config.gcc: Support "tremont".
1902 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1903 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1905 * config/i386/i386.c (m_TREMONT): Define.
1906 (processor_target_table): Add "tremont".
1907 (PTA_TREMONT): Define.
1908 (ix86_lea_outperforms): Add TARGET_TREMONT.
1909 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1910 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1911 and M_INTEL_GOLDMONT_PLUS.
1912 (fold_builtin_cpu): Add "tremont".
1913 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1914 (ix86_option_override_internal): Add "tremont".
1915 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1916 (processor_type): Add PROCESSOR_TREMONT.
1917 * config/i386/x86-tune.def: Add m_TREMONT.
1918 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1920 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1922 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1923 symbol defined for msp430i* devices to be lower case.
1925 gcc/testsuite/gcc.target/msp430/
1926 * msp430i-device-symbol.c: New test.
1927 * msp430f-device-symbol.c: New test.
1928 * msp430.h: New test header file.
1930 2018-06-07 Richard Biener <rguenther@suse.de>
1932 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1933 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1934 Properly wrap signed arithmetic if overflow wraps.
1936 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1938 PR tree-optimization/69615
1939 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1940 of a cast from a same precision integral SSA_NAME in a bb dominated
1941 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1942 cast to utype if rhs2 has already a compatible type.
1944 2018-06-07 Richard Biener <rguenther@suse.de>
1946 PR tree-optimization/85935
1947 * graphite-scop-detection.c (find_params_in_bb): Analyze
1948 condition operands with respect to the correct loop. Assert
1949 the analysis doesn't fail.
1951 2018-06-04 Carl Love <cel@us.ibm.com>
1953 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1954 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1955 as it is slightly cheaper.
1956 (first_match_or_eos_index_<mode>):
1957 Calculate index using natural element order.
1958 (first_match_index_<mode>):
1959 Calculate index using natural element order.
1960 (first_match_or_eos_index_<mode>):
1961 Calculate index using natural order.
1962 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1963 for BE and LE modes.
1964 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1965 P9V_BUILTIN_VCLZLSBB_V16QI.
1966 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1969 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1971 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1972 indentation and line wrap for many prototypes. Add missing
1973 @smallexample directives around block of prototypes for vec_xl and
1976 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1978 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1979 track if we pass or return IEEE 128-bit floating point.
1980 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1981 C++ mangling that is compatible with GCC 8.1.
1982 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1983 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1984 floating point types.
1985 (rs6000_function_arg_advance_1): Likewise.
1986 (rs6000_mangle_type): Optionally generate mangled names that match
1987 what GCC 8.1 generated for IEEE 128-bit floating point types.
1988 (rs6000_globalize_decl_name): If we have an external function that
1989 passes or returns IEEE 128-bit types, generate a weak reference
1990 from the mangled name used in GCC 8.1 to the current mangled
1992 (rs6000_init_builtins): Make __ibm128 use the long double type if
1993 long double is IBM extended double. Make __float128 use the long
1994 double type if long double is IEEE 128-bit.
1997 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1998 macro for __ibm128 built-in functions.
1999 (PACK_IF): Add __ibm128 pack/unpack functions.
2000 (UNPACK_IF): Likewise.
2001 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2002 enable long double built-in functions if long double is IEEE
2003 128-bit floating point.
2004 (rs6000_invalid_builtin): Update long double built-in function
2006 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2007 functions, adjust the built-in function to use the long double
2008 built-in function if __ibm128 and long double are the same type.
2009 * doc/extend.texi (PowerPC builtins): Update documention for
2010 __builtin_{,un}pack_longdouble. Add documentation for
2011 __builtin_{,un}pack_ibm128.
2013 2018-06-06 Jim Wilson <jimw@sifive.com>
2015 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2016 (struct machine_function): New field interrupt_mode.
2017 (riscv_handle_type_attribute): New function. Add forward declaration.
2018 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2019 (riscv_expand_epilogue): Check interrupt_mode field.
2020 (riscv_set_current_function): Check interrupt attribute args and
2021 set interrupt_mode field.
2022 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2023 (riscv_sret, riscv_uret): New.
2024 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2025 new arguments to interrupt attribute.
2027 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2030 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2031 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2033 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2035 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2036 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2037 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2038 several redundant entries.
2040 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2042 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2043 type from "rtx" to "rtx_insn *".
2044 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2045 for local "call_insn", removing cast.
2046 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2048 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2050 PR tree-optimization/86066
2051 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2052 for BIT_INSERT_EXPR stores.
2054 2018-06-06 Richard Biener <rguenther@suse.de>
2056 PR tree-optimization/86062
2057 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2058 component refs ontop
2059 of to be offsetted base.
2061 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2063 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2064 to be static and remove check on interrupt attribute name.
2066 gcc/testsuite/gcc.target/msp430/
2067 * function-attributes-4.c: New test.
2068 * static-interrupts.c: New test.
2070 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2072 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2073 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2075 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2078 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2080 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2081 Remove second argument, change how error is called.
2082 (aarch64_layout_arg): Remove second argument from
2083 aarch64_err_no_fpadvsimd call.
2084 (aarch64_init_cumulative_args): Ditto.
2085 (aarch64_gimplify_va_arg_expr): Ditto.
2086 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2088 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2090 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2091 (*simple_return_indirect_internal<mode>): Rename from
2092 simple_return_indirect_internal. Use W mode iterator.
2093 (rstorssp): New expander.
2094 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2095 (clrssbsy): New expander.
2096 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2098 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2100 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2102 (cmse_check_pointed_object): Likewise.
2104 2018-06-05 Martin Liska <mliska@suse.cz>
2106 PR gcov-profile/47618
2107 * doc/invoke.texi: Document how -fprofile-dir format
2110 2018-06-05 Richard Biener <rguenther@suse.de>
2112 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2113 removal pretend DOM info isn't available so we do not update
2114 it and only remove edges, not dominated blocks. Actually free
2115 DOM info in case we removed something. Remove unreachable blocks.
2116 (mfb_keep_latches): Work with either DOM info or marked backedges.
2117 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2118 first. Mark backedges if DOM info isn't available.
2119 (Re-)compute DOM info after cleanup_control_flow_pre.
2121 2018-06-05 Richard Biener <rguenther@suse.de>
2123 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2124 (locus_discrim_hasher::hash): Adjust.
2125 (locus_discrim_hasher::equal): Likewise.
2126 (next_discriminator_for_locus): Work on line directly.
2127 (same_line_p): Pass in expanded locus1 as well.
2128 (assign_discriminators): Avoid redundant location expansions.
2130 2018-06-05 Richard Biener <rguenther@suse.de>
2132 PR tree-optimization/86046
2133 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2134 if required after clearing TREE_ADDRESSABLE.
2136 2018-06-05 Richard Biener <rguenther@suse.de>
2138 PR tree-optimization/86047
2139 * tree-ssa-loop.c (for_each_index): Glob handling of all
2140 decls and constants and really handle all of them.
2142 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2145 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2146 qualifier_void_pointer and qualifier_const_void_pointer.
2147 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2148 (arm_init_builtins): Handle the above.
2149 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2150 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2151 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2154 2018-06-05 Martin Liska <mliska@suse.cz>
2156 * auto-profile.c (read_autofdo_file): Do not use
2157 gcov_ctr_summary struct.
2158 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2159 * coverage.c (struct counts_entry): Likewise.
2160 (read_counts_file): Read just single summary entry.
2161 (get_coverage_counts): Use gcov_summary struct.
2162 * coverage.h (get_coverage_counts): Likewise.
2163 * gcov-dump.c (dump_working_sets): Likewise.
2164 (tag_summary): Dump just single summary.
2165 * gcov-io.c (gcov_write_summary): Write just histogram
2167 (gcov_read_summary): Read just single summary.
2168 (compute_working_sets): Use gcov_summary struct.
2169 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2170 of GCOV_COUNTERS_SUMMABLE.
2171 (GCOV_COUNTERS_SUMMABLE): Remove.
2172 (GCOV_FIRST_VALUE_COUNTER): Replace with
2173 GCOV_COUNTER_V_INTERVAL.
2174 (struct gcov_ctr_summary): Remove.
2175 (struct gcov_summary): Directly use fields of former
2177 (compute_working_sets): Use gcov_summary struct.
2178 * gcov.c (read_count_file): Do not use ctrs fields.
2179 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2181 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2183 * profile.c: Likewise.
2184 * profile.h: Likewise.
2186 2018-06-05 Martin Liska <mliska@suse.cz>
2188 PR gcov-profile/84846
2189 * gcov.c (output_lines): Print working directory only
2190 in intermediate format.
2192 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2194 * config/s390/s390-builtin-types.def: Add void function type.
2195 * config/s390/s390-builtins.def: Use the function type for the
2198 2018-06-04 Jim Wilson <jimw@sifive.com>
2200 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2202 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2203 and maybe_eh_return. Change regno to unsigned int. Use new args to
2204 handle EH_RETURN_DATA_REGNO registers properly.
2205 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2206 (riscv_expand_epilogue): Update comment. Change argument name and
2207 type. Update code to use new name and type. Pass new args to
2208 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2210 * config/riscv/riscv.md (NORMAL_RETURN): New.
2211 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2212 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2213 (eh_return): Call gen_eh_return_internal and emit barrier.
2214 (eh_return_internal): Call riscv_expand_epilogue.
2216 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2218 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2219 bit_insertion field and declare can_be_merged_into method.
2220 (merged_store_group::can_be_merged_into): New method.
2221 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2222 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2223 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2225 2018-06-04 Richard Biener <rguenther@suse.de>
2227 PR tree-optimization/85955
2228 * builtins.c (fold_builtin_sincos): Convert pointers to
2229 destination to appropriate type before dereferencing.
2231 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2233 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2235 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2237 * expr.c (expand_expr_real_1): Force the operand into memory if
2238 its TYPE_MODE is BLKmode and if there is no integer mode for
2239 the number of bits being extracted.
2241 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2245 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2246 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2248 2018-06-04 Richard Biener <rguenther@suse.de>
2250 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2251 (cleanup_tree_cfg_noloop): ... single caller. Do
2252 start_recording_case_labels later.
2254 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2256 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2257 to _IMMINTRIN_H_INCLUDED.
2258 * config/i386/pconfigintrin.h: Ditto.
2259 * config/i386/waitpkgintrin.h: Ditto.
2260 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2261 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2262 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2263 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2264 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2265 waitpkgintrin.h and cldemoteintrin.h.
2267 2018-06-04 Richard Biener <rguenther@suse.de>
2269 PR tree-optimization/86038
2270 * tracer.c (find_best_successor): Check probability for
2271 being initialized, bail out if not.
2273 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2276 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2277 of bits to ignore when comparing architectures.
2279 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2281 PR tree-optimization/69615
2282 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2283 maximum or minimum of the type, try to merge it also as if
2284 range1 is + [-, x - 1] or + [x + 1, -].
2287 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2289 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2291 PR tree-optimization/86034
2292 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2293 the unsigned bitfield type in a bit insertion sequence if it does not
2294 have a larger precision than the bitfield size.
2295 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2297 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2299 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2301 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2303 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2304 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2305 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2306 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2308 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2310 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2311 Disable -fdelete-null-pointer-checks for ELF toolchain.
2313 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2314 Kito Cheng <kito.cheng@gmail.com>
2316 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2317 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2318 (nds32*-*-*): Add glibc and uclibc conditions.
2319 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2320 (TARGET_EXCEPT_UNWIND_INFO): Define.
2321 * config/nds32/elf.h: New file.
2322 * config/nds32/linux.h: New file.
2323 * config/nds32/nds32-elf.opt: New file.
2324 * config/nds32/nds32-linux.opt: New file.
2325 * config/nds32/nds32-fp-as-gp.c
2326 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2327 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2329 (nds32_asm_file_end): Ditto.
2330 (nds32_print_operand): Ditto.
2331 (nds32_insert_attributes): Ditto.
2332 (nds32_init_libfuncs): New function.
2333 (TARGET_HAVE_TLS): Define.
2334 (TARGET_INIT_LIBFUNCS): Define.
2335 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2337 (TARGET_ELF): Apply different mcmodel setting.
2338 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2339 been migrated into elf.h and linux.h files.
2340 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2341 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2342 (mcmodel): The content has been migrated into nds32-elf.opt and
2343 nds32-linux.opt files.
2344 * config/nds32/t-elf: New file.
2345 * config/nds32/t-linux: New file.
2347 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2348 Shiva Chen <shiva0217@gmail.com>
2350 * config/nds32/constants.md (unspec_volatile_element): Add
2351 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2352 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2354 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2355 (make_pass_nds32_fp_as_gp): Declare.
2356 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2358 (nds32_asm_function_end_prologue): Remove unused asm output.
2359 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2360 (nds32_asm_file_start): Output necessary fp_as_gp information.
2361 (nds32_option_override): Adjust register usage.
2362 (nds32_expand_prologue): Consider fp_as_gp situation.
2363 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2364 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2367 (simple_return): Ditto.
2368 (omit_fp_begin): Output special directive for fp_as_gp.
2369 (omit_fp_end): Output special directive for fp_as_gp.
2370 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2371 mforbid-fp-as-gp): New options.
2373 2018-06-01 Mark Wielaard <mark@klomp.org>
2375 * dwarf2out.c (dwarf2out_finish): Remove generation of
2376 DW_AT_loclists_base.
2378 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2380 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2381 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2382 (struct merged_store_group): Add bit_insertion field.
2383 (dump_char_array): Use standard hexadecimal format.
2384 (merged_store_group::merged_store_group): Set bit_insertion to false.
2385 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2386 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2387 also print the mask in the dump file.
2388 (pass_store_merging::gate): Minor tweak.
2389 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2390 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2391 stores with INTEGER_CST stores.
2392 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2393 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2394 and use it throughout. Generate bit insertion sequences if need be.
2395 (pass_store_merging::process_store): Remove redundant condition.
2396 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2398 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2400 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2401 the 128-bit floating point types. Fix function comment.
2403 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2405 * config/aarch64/aarch64-simd.md
2406 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2408 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2411 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2413 PR tree-optimization/85989
2414 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2416 (backprop::intersect_uses): Check it when deciding whether this
2417 is a backedge reference.
2418 (backprop::process_block): Add each phi to m_visited_phis
2419 after visiting it, then clear it at the end.
2421 2018-06-01 Richard Biener <rguenther@suse.de>
2423 * tree-vectorizer.h (vect_dr_stmt): New function.
2424 (vect_get_load_cost): Adjust.
2425 (vect_get_store_cost): Likewise.
2426 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2427 Use vect_dr_stmt instead of DR_SMTT.
2428 (vect_record_base_alignments): Likewise.
2429 (vect_calculate_target_alignment): Likewise.
2430 (vect_compute_data_ref_alignment): Likewise and make static.
2431 (vect_update_misalignment_for_peel): Likewise.
2432 (vect_verify_datarefs_alignment): Likewise.
2433 (vector_alignment_reachable_p): Likewise.
2434 (vect_get_data_access_cost): Likewise. Pass down
2435 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2436 (vect_get_peeling_costs_all_drs): Likewise.
2437 (vect_peeling_hash_get_lowest_cost): Likewise.
2438 (vect_enhance_data_refs_alignment): Likewise.
2439 (vect_find_same_alignment_drs): Likewise.
2440 (vect_analyze_data_refs_alignment): Likewise.
2441 (vect_analyze_group_access_1): Likewise.
2442 (vect_analyze_group_access): Likewise.
2443 (vect_analyze_data_ref_access): Likewise.
2444 (vect_analyze_data_ref_accesses): Likewise.
2445 (vect_vfa_segment_size): Likewise.
2446 (vect_small_gap_p): Likewise.
2447 (vectorizable_with_step_bound_p): Likewise.
2448 (vect_prune_runtime_alias_test_list): Likewise.
2449 (vect_analyze_data_refs): Likewise.
2450 (vect_supportable_dr_alignment): Likewise.
2451 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2452 (vect_gen_prolog_loop_niters): Likewise.
2453 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2454 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2456 (vect_recog_mask_conversion_pattern): Likewise.
2457 (vect_try_gather_scatter_pattern): Likewise.
2458 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2459 to vect_get_store_cost.
2460 (vect_get_store_cost): Get stmt_info instead of DR.
2461 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2462 (vect_get_load_cost): Get stmt_info instead of DR.
2464 2018-06-01 Richard Biener <rguenther@suse.de>
2467 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2468 vars wrapped in MEM_REFs.
2470 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2472 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2473 Fix subreg tests so that we only return a choice between
2474 GENERAL_REGS and FP_REGS if the original classes included both.
2476 2018-06-01 Richard Biener <rguenther@suse.de>
2479 * tree-ssa-structalias.c (get_function_part_constraint):
2480 Handle NULL fi->decl.
2481 (find_func_aliases_for_call): Properly handle indirect
2482 fi from direct call.
2483 (find_func_clobbers): Likewise.
2484 (ipa_pta_execute): Likewise.
2485 (create_variable_info_for): For functions that are ifunc_resolver
2486 resolve to a varinfo that contains the result of the resolver call.
2487 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2490 2018-05-31 Michael Collison <michael.collison@arm.com>
2492 * config/aarch64/aarch64.md:
2493 (*fix_to_zero_extenddfdi2): New pattern.
2494 * gcc.target/aarch64/fix_extend1.c: New testcase.
2496 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2500 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2501 and BUILT_IN_STRNCMP_EQ.
2502 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2503 BUILT_IN_STRNCMP_EQ.
2504 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2505 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2506 (gimple_fold_builtin): Likewise.
2507 * tree-ssa-strlen.c (compute_string_length): New function.
2508 (determine_min_obsize): New function.
2509 (handle_builtin_string_cmp): New function to handle calls to
2510 string compare functions.
2511 (strlen_optimize_stmt): Add handling to builtin string compare
2513 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2514 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2515 * tree.c (build_common_builtin_nodes): Add new defines of
2516 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2518 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2521 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2522 functions with naked attribute.
2524 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2526 * config/i386/sse.md (avx_vec_concat<mode>):
2527 Substitute concat_tg_mode mode attribute with xtg_mode.
2528 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2529 (concat_tg_mode): Remove mode attribute.
2531 2018-05-31 Martin Sebor <msebor@redhat.com>
2534 * calls.c (alloc_max_size): Correct a logic error/typo.
2535 Treat excessive arguments as infinite. Warn for invalid arguments.
2536 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2538 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2541 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2542 and movx for Haswell.
2544 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2545 Cesar Philippidis <cesar@codesourcery.com>
2548 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2549 when emitting error on private/firstprivate reductions.
2550 * omp-low.c (lower_omp_target): Avoid reference-type processing
2551 on pointers for firstprivate clause.
2553 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2555 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2558 * config/aarch64/aarch64-simd.md
2559 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2560 (aarch64_ld1_x3_<mode>): Likewise
2561 (aarch64_st1x2<VALLDIF:mode>): Likewise
2562 (aarch64_st1_x2_<mode>): Likewise
2563 (aarch64_st1x3<VALLDIF:mode>): Likewise
2564 (aarch64_st1_x3_<mode>): Likewise
2565 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2566 (vld1_s8_x3): Likewise.
2567 (vld1_u16_x3): Likewise.
2568 (vld1_s16_x3): Likewise.
2569 (vld1_u32_x3): Likewise.
2570 (vld1_s32_x3): Likewise.
2571 (vld1_u64_x3): Likewise.
2572 (vld1_s64_x3): Likewise.
2573 (vld1_f16_x3): Likewise.
2574 (vld1_f32_x3): Likewise.
2575 (vld1_f64_x3): Likewise.
2576 (vld1_p8_x3): Likewise.
2577 (vld1_p16_x3): Likewise.
2578 (vld1_p64_x3): Likewise.
2579 (vld1q_u8_x3): Likewise.
2580 (vld1q_s8_x3): Likewise.
2581 (vld1q_u16_x3): Likewise.
2582 (vld1q_s16_x3): Likewise.
2583 (vld1q_u32_x3): Likewise.
2584 (vld1q_s32_x3): Likewise.
2585 (vld1q_u64_x3): Likewise.
2586 (vld1q_s64_x3): Likewise.
2587 (vld1q_f16_x3): Likewise.
2588 (vld1q_f32_x3): Likewise.
2589 (vld1q_f64_x3): Likewise.
2590 (vld1q_p8_x3): Likewise.
2591 (vld1q_p16_x3): Likewise.
2592 (vld1q_p64_x3): Likewise.
2593 (vst1_s64_x2): Likewise.
2594 (vst1_u64_x2): Likewise.
2595 (vst1_f64_x2): Likewise.
2596 (vst1_s8_x2): Likewise.
2597 (vst1_p8_x2): Likewise.
2598 (vst1_s16_x2): Likewise.
2599 (vst1_p16_x2): Likewise.
2600 (vst1_s32_x2): Likewise.
2601 (vst1_u8_x2): Likewise.
2602 (vst1_u16_x2): Likewise.
2603 (vst1_u32_x2): Likewise.
2604 (vst1_f16_x2): Likewise.
2605 (vst1_f32_x2): Likewise.
2606 (vst1_p64_x2): Likewise.
2607 (vst1q_s8_x2): Likewise.
2608 (vst1q_p8_x2): Likewise.
2609 (vst1q_s16_x2): Likewise.
2610 (vst1q_p16_x2): Likewise.
2611 (vst1q_s32_x2): Likewise.
2612 (vst1q_s64_x2): Likewise.
2613 (vst1q_u8_x2): Likewise.
2614 (vst1q_u16_x2): Likewise.
2615 (vst1q_u32_x2): Likewise.
2616 (vst1q_u64_x2): Likewise.
2617 (vst1q_f16_x2): Likewise.
2618 (vst1q_f32_x2): Likewise.
2619 (vst1q_f64_x2): Likewise.
2620 (vst1q_p64_x2): Likewise.
2621 (vst1_s64_x3): Likewise.
2622 (vst1_u64_x3): Likewise.
2623 (vst1_f64_x3): Likewise.
2624 (vst1_s8_x3): Likewise.
2625 (vst1_p8_x3): Likewise.
2626 (vst1_s16_x3): Likewise.
2627 (vst1_p16_x3): Likewise.
2628 (vst1_s32_x3): Likewise.
2629 (vst1_u8_x3): Likewise.
2630 (vst1_u16_x3): Likewise.
2631 (vst1_u32_x3): Likewise.
2632 (vst1_f16_x3): Likewise.
2633 (vst1_f32_x3): Likewise.
2634 (vst1_p64_x3): Likewise.
2635 (vst1q_s8_x3): Likewise.
2636 (vst1q_p8_x3): Likewise.
2637 (vst1q_s16_x3): Likewise.
2638 (vst1q_p16_x3): Likewise.
2639 (vst1q_s32_x3): Likewise.
2640 (vst1q_s64_x3): Likewise.
2641 (vst1q_u8_x3): Likewise.
2642 (vst1q_u16_x3): Likewise.
2643 (vst1q_u32_x3): Likewise.
2644 (vst1q_u64_x3): Likewise.
2645 (vst1q_f16_x3): Likewise.
2646 (vst1q_f32_x3): Likewise.
2647 (vst1q_f64_x3): Likewise.
2648 (vst1q_p64_x3): Likewise.
2650 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2652 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2653 user_label_prefix to name.
2655 * tree-core.h: Update comment about the format of NAME string
2656 passed to handler in attribute_spec.
2658 * config/msp430/msp430.md: Remove erroneous subreg expression from
2659 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2660 zero_extend{q,h}isi2.
2662 2018-05-30 Borislav Petkov <bp@suse.de>
2664 * doc/extend.texi: Document some architecture specific
2665 constraints and sort entries.
2667 2018-05-30 Martin Sebor <msebor@redhat.com>
2670 * builtins.c (expand_builtin_stpcpy_1): New function.
2671 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2672 only if the former succeeds.
2674 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2676 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2679 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2681 * doc/invoke.texi (-flinker-output): Document
2683 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2685 * passes.c (ipa_write_summaries): Only modify statements if body
2687 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2688 incrementally linking.
2689 (ipa_passes): Likewise.
2690 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2691 pass down resolution info.
2692 * common.opt (flag_incremental_link): Update info.
2693 * gcc.c (plugin specs): Turn flinker-output=* to
2694 -plugin-opt=-linker-output-known
2695 * toplev.c (compile_file): Also cut compilation when doing incremental
2697 * flag-types. (enum lto_partition_model): Add
2698 LTO_LINKER_OUTPUT_NOLTOREL.
2699 (invoke.texi): Add -flinker-output docs.
2700 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2701 link same way as WPA; do not stream in dead initializers.
2703 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2704 darf2out_register_external_decl): Support incremental link.
2706 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2708 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2710 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2712 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2713 it down to simple_object_copy_lto_debug_sections.
2714 (run_gcc): Determine incremental LTO link time and configure
2715 lto1 into non-wpa mode, disable renaming of debug sections.
2717 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2719 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2720 descriptions of various incorrectly documented functions.
2722 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2725 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2726 address check not strict.
2728 2018-05-30 Richard Biener <rguenther@suse.de>
2730 PR tree-optimization/85964
2731 * tracer.c (better_p): Drop initialized count check, we only
2732 call the function with initialized counts now.
2733 (find_best_successor): Do find a best edge if one
2734 has uninitialized count.
2735 (find_best_predecessor): Likewise. Do BB frequency check only
2736 if count is initialized.
2738 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2740 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2741 (aarch64_ldrstr_offset_compare): New.
2742 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2743 load/store orderings.
2744 (aarch64_gen_adjusted_ldpstp): Likewise.
2746 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2748 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2749 Check for subset of GENERAL_REGS and FP_REGS.
2750 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2753 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2755 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2756 and wi::to_poly_offset. Add the current offset and then check
2757 whether the sum fits, rather than using an unchecked addition of
2758 a checked term. Check for a shwi rather than a uhwi.
2759 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2760 (store_constructor): Use poly_int_tree_p.
2761 (expand_expr_real_1): Likewise.
2762 * function.c (assign_temp): Likewise.
2763 * fold-const.c (const_binop): Use poly_int_tree_p and
2765 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2767 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2768 to_poly_offset for MEM offsets.
2769 * ipa-icf.c (sem_variable::equals): Likewise.
2770 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2771 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2772 wi::to_poly_offset for BIT_FIELD_REF offsets.
2773 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2775 * var-tracking.c (emit_note_insn_var_location): Use
2776 tree_to_poly_uint64.
2778 2018-05-29 Jim Wilson <jimw@sifive.com>
2780 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2782 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2785 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2786 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2788 (sse4_1_round<mode>2): Use nonimmediate_operand
2789 for operand 1 predicate.
2791 2018-05-29 Martin Sebor <msebor@redhat.com>
2792 Richard Biener <rguenther@suse.de>
2795 * calls.c (get_size_range): Call determine_value_range instead
2796 of get_value_range..
2797 * tree-vrp.h (determine_value_range): Declared new function.
2798 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2800 2018-05-29 Richard Biener <rguenther@suse.de>
2802 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2803 sure to use non-pattern stmts for get_earlier_stmt arguments.
2804 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2805 called on pattern stmts.
2806 (get_later_stmt): Likewise.
2808 2018-05-29 Martin Liska <mliska@suse.cz>
2810 PR gcov-profile/85759
2811 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2814 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2816 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2818 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2822 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2823 VEC_PACK_FLOAT_EXPR): New tree codes.
2824 * tree-pretty-print.c (op_code_prio): Handle
2825 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2826 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2827 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2828 * tree-inline.c (estimate_operator_cost): Likewise.
2829 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2830 * fold-const.c (const_binop): Likewise.
2831 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2832 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2833 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2834 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2835 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2836 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2837 * expr.c (expand_expr_real_2): Likewise.
2838 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2839 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2840 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2842 * optabs.c (expand_widen_pattern_expr): For
2843 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2844 sign from result type rather than operand's type.
2845 (expand_binop_directly): For vec_packu_float_optab and
2846 vec_packs_float_optab allow result type to be different from operand's
2848 * optabs-tree.c (optab_for_tree_code): Handle
2849 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2850 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2851 * tree-vect-generic.c (expand_vector_operations_1): Handle
2852 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2853 VEC_PACK_FLOAT_EXPR.
2854 * tree-vect-stmts.c (supportable_widening_operation): Handle
2856 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2857 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2858 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2859 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2860 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2862 (vec_pack<floatprefix>_float_<mode>): New expander.
2863 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2865 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2866 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2867 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2868 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2869 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2871 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2872 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2873 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2874 VEC_PACK_FLOAT_EXPR): Document.
2876 2018-05-29 Richard Biener <rguenther@suse.de>
2878 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2880 (stmt_vec_info_vec): Make pointer.
2881 (init_stmt_vec_info_vec): Remove.
2882 (free_stmt_vec_info_vec): Likewise.
2883 (set_stmt_vec_info_vec): New function.
2884 (free_stmt_vec_infos): Likewise.
2885 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2886 (set_vinfo_for_stmt): Likewise.
2887 (get_earlier_stmt): Likewise.
2888 (get_later_stmt): Likewise.
2889 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2890 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2891 (vec_info::~vec_info): Free stmt_vec_infos.
2892 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2893 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2894 (pass_slp_vectorize::execute): Likewise.
2895 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2896 (free_stmt_vec_info_vec): Likewise.
2897 (set_stmt_vec_info_vec): New function.
2898 (free_stmt_vec_infos): Likewise.
2899 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2900 the global stmt_vec_info_vec.
2901 * tree-parloops.c (gather_scalar_reductions): Use
2902 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2905 2018-05-29 Richard Biener <rguenther@suse.de>
2907 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2909 2018-05-29 Martin Liska <mliska@suse.cz>
2910 David Malcolm <dmalcolm@redhat.com>
2912 * vec.c (test_reverse): New.
2913 (vec_c_tests): Add new test.
2914 * vec.h (vl_ptr>::reverse): New function.
2916 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2918 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2920 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2923 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2925 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2927 2018-05-28 Richard Biener <rguenther@suse.de>
2929 PR tree-optimization/85933
2930 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2931 look at stmts marked as vectorizable.
2933 2018-05-28 Richard Biener <rguenther@suse.de>
2935 PR tree-optimization/85934
2936 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2937 vector boolean check before scalar optimization.
2939 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2941 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2944 2018-05-28 Mark Wielaard <mark@klomp.org>
2946 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2947 if it is an expression containing a minus sign.
2949 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2951 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2953 2018-05-27 Paul Koning <ni1d@arrl.net>
2955 * config/pdp11/pdp11.md (truncsihi2): Remove.
2957 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2958 Chung-Ju Wu <jasonwucj@gmail.com>
2960 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2962 (unaligned_store_dw): Ditto.
2963 * config/nds32/nds32-memory-manipulation.c
2964 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2965 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2966 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2967 (emit_setmem_word_loop): Rename to ...
2968 (emit_setmem_doubleword_loop): ... this.
2969 (nds32_gen_dup_4_byte_to_word_value): New function.
2970 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2971 (nds32_expand_setmem_loop): Refine implementation.
2972 (nds32_expand_setmem_loop_v3m): Ditto.
2973 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2976 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2978 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2980 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2982 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2983 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2984 machine->attr_no_prologue_p.
2985 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2986 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2987 (nds32_expand_epilogue): Consider attr_naked_p.
2988 (nds32_expand_epilogue_v3pop): Likewise.
2989 (nds32_can_use_return_insn): Likewise.
2990 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2991 attr_no_prologue_p fields.
2992 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2994 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2997 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2999 * config/i386/sse.md
3000 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3002 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3004 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3006 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3008 (*<floatsuffix>floatv2div2sf2): Rename to ...
3009 (*float<floatunssuffix>v2div2sf2): ... this.
3010 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3011 (float<floatunssuffix>v2div2sf2_mask): ... this.
3012 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3013 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3014 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3016 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3018 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3020 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3022 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3024 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3026 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3027 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3028 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3029 gen_ufix_truncv8dfv8si2.
3030 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3031 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3032 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3033 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3034 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3035 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3036 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3037 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3039 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3043 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3045 2018-05-25 Jim Wilson <jimw@sifive.com>
3047 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3048 * config/riscv/riscv.c (struct machine_function): Add
3049 interrupt_handler_p and attribute_checked_p fields.
3050 (riscv_attribute_table): Add interrupt.
3051 (riscv_interrupt_type_p): New.
3052 (riscv_save_reg_p): Save extra regs for interrupt handler.
3053 (riscv_use_save_libcall): Return false for interrupt handler.
3054 (riscv_first_stack_step): Add forward declaration.
3055 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3056 for interrupt handler with large frame. Use it for saved reg list.
3057 (riscv_expand_prologue): Move flag_stack_usage_info support to
3058 eliminate duplication.
3059 (riscv_expand_epilogue): Generate mret for interrupt handler.
3060 (riscv_epilogue_uses): New.
3061 (riscv_can_use_return_insn): Return false for interrupt handler.
3062 (riscv_function_ok_for_sibcall): Likewise.
3063 (riscv_set_current_function): Add interrupt handler support.
3064 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3065 * config/riscv/riscv.md (UNSPECV_MRET): New.
3067 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3069 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3071 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3073 PR tree-optimization/85712
3074 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3075 this candidate has already been replaced in-situ by a copy.
3077 2018-05-25 Jason Merrill <jason@redhat.com>
3079 PR c++/80485 - inline function non-zero address.
3080 * symtab.c (nonzero_address): Check DECL_COMDAT.
3082 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3085 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3086 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3087 extension of SImode operation. Use const123_operand predicate.
3088 (*saddsi_1): Remove.
3089 (*saddl_se_1): Ditto.
3091 (*ssubl_se_1): Ditto.
3092 * config/alpha/predicates.md (const123_operand): New predicate.
3093 * config/alpha/constraints.md (P): Use IN_RANGE.
3095 2018-05-25 Richard Biener <rguenther@suse.de>
3097 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3099 (ref_maybe_used_by_stmt_p): Likewise.
3100 (stmt_may_clobber_ref_p): Likewise.
3101 (stmt_may_clobber_ref_p_1): Likewise.
3102 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3104 (ref_maybe_used_by_stmt_p): Likewise.
3105 (stmt_may_clobber_ref_p): Likewise.
3106 (stmt_may_clobber_ref_p_1): Likewise.
3107 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3108 the alias oracle to disambiguate DRs with stmts DR analysis
3110 (vect_analyze_data_refs): Do not give up on not analyzable
3111 DRs for BB vectorization. Remove code truncating the dataref
3114 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3117 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3118 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3119 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3121 2018-05-25 Richard Biener <rguenther@suse.de>
3123 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3124 function, combining stmt data ref gathering and fatal analysis
3126 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3127 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3128 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3129 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3130 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3132 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3134 PR tree-optimization/85720
3135 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3136 SCC if all partitions are builtins.
3137 (version_loop_by_alias_check): New parameter. Generate cancelable
3138 runtime alias check if all partitions are builtins.
3139 (distribute_loop): Update call to above function.
3141 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3143 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3144 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3145 (parm_default_def_partition_arg): Ditto.
3146 (set_parm_default_def_partition): Ditto.
3147 (get_parm_default_def_partitions): Ditto and make it static.
3148 (get_undefined_value_partitions): Ditto and make it static.
3149 (remove_ssa_form): Refactor call to init_var_map here.
3150 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3151 computation for loop region.
3152 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3153 (register_default_def): Delete.
3154 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3155 (parm_default_def_partition_arg): Ditto.
3156 (set_parm_default_def_partition): Ditto.
3157 (get_parm_default_def_partitions): Ditto and make it static.
3158 (get_undefined_value_partitions): Ditto and make it static.
3159 (coalesce_with_default, coalesce_with_default): Update comment.
3160 (create_coalesce_list_for_region): New func factored out from
3161 create_outofssa_var_map.
3162 (populate_coalesce_list_for_outofssa): New func factored out from
3163 create_outofssa_var_map and coalesce_ssa_name.
3164 (create_outofssa_var_map): Delete.
3165 (coalesce_ssa_name): Refactor to support live range computation.
3166 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3167 (get_parm_default_def_partitions): Delete.
3168 (get_undefined_value_partitions): Ditto.
3169 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3170 computation for loop region.
3171 (new_tree_live_info, loe_visit_block): Ditto.
3172 (live_worklist, set_var_live_on_entry): Ditto.
3173 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3174 * tree-ssa-live.h (struct _var_map): New fields.
3175 (init_var_map): Change decl.
3176 (region_contains_p): New.
3178 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3180 * tree-ssa-live.h (live_merge_and_clear): Delete.
3182 2018-05-25 Richard Biener <rguenther@suse.de>
3185 * tree-dump.c (dequeue_and_dump): Remove access to removed
3186 operand 2 of a SWITCH_EXPR.
3188 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3190 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3191 multiplication and division.
3192 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3193 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3194 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3195 (cond_udiv_optab, cond_umod_optab): New optabs.
3196 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3197 (IFN_COND_RDIV): New internal functions.
3198 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3199 TRUNC_MOD_EXPR and RDIV_EXPR.
3200 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3201 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3203 (SVE_INT_BINARY): Include mult.
3204 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3205 (optab, sve_int_op): Handle mult.
3206 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3208 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3209 for SVE_INT_BINARY_SD.
3211 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3213 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3214 (optab, sve_int_op): Handle div and udiv.
3215 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3216 for SVE_INT_BINARY_SD.
3217 (*<optab><mode>3): New insn for the same.
3219 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3221 * tree-vect-patterns.c: Include predict.h.
3222 (vect_recog_divmod_pattern): Restrict check for division support
3223 to when optimizing for size.
3225 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3227 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3228 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3229 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3230 (gimple_match_op::set_op): Likewise.
3231 (gimple_resimplify4): Declare.
3232 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3233 (expr::gen_transform): Likewise.
3234 (decision_tree::gen): Generate a simplification routine for 4 operands.
3235 * gimple-match-head.c (gimple_simplify): Add an overload for
3236 4 operands. In the top-level function, handle up to 4 call
3237 arguments and call gimple_resimplify4.
3238 (gimple_resimplify4): New function.
3239 (build_call_internal): Pass a fourth operand.
3240 (maybe_push_to_seq): Likewise.
3241 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3242 Fold VEC_COND_EXPRs of an operation and a default value into
3243 an IFN_COND_* function if possible.
3244 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3246 (SVE_COND_FP_BINARY): Include them.
3247 (optab, sve_fp_op): Handle them.
3248 (SVE_INT_BINARY_REV): New code iterator.
3249 (SVE_COND_FP_BINARY_REV): New int iterator.
3250 (commutative): New int attribute.
3251 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3253 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3255 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3256 (*cond_<optab><mode>): New patterns for reversed operands.
3258 2018-05-25 Richard Biener <rguenther@suse.de>
3260 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3261 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3262 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3263 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3264 * tree-vect-data-refs.c (everywhere): Adjust users.
3265 * tree-vect-loop.c (everywhere): Likewise.
3266 * tree-vect-slp.c (everywhere): Likewise.
3267 * tree-vect-stmts.c (everywhere): Likewise.
3268 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3270 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3272 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3274 (gcc_cv_as_section_exclude): ... this.
3275 Try Solaris as #exclude syntax.
3276 * configure: Regenerate.
3277 * config.in: Regenerate.
3278 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3280 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3281 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3283 * varasm.c (default_elf_asm_named_section): Don't check if
3284 HAVE_GAS_SECTION_EXCLUDE is defined.
3286 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3288 * doc/md.texi: Update the documentation of the cond_* optabs
3289 to mention the new final operand. Fix GET_MODE_NUNITS call.
3290 Describe the scalar case too.
3291 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3292 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3294 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3295 (get_conditional_internal_fn): Update comment.
3296 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3297 accumulator value as a final argument to conditional functions.
3298 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3299 a define_expand and add an "else" operand. Assert for now that
3300 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3301 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3302 (*cond_<optab><mode>): New patterns.
3303 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3304 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3305 (UNSPEC_COND_EOR): Delete.
3306 (optab): Remove associated mappings.
3307 (SVE_INT_BINARY): New code iterator.
3308 (sve_int_op): Remove int attribute and add "minus" to the code
3310 (SVE_COND_INT_OP): Delete.
3311 (SVE_COND_FP_OP): Rename to...
3312 (SVE_COND_FP_BINARY): ...this.
3314 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3316 * optabs.c (can_reuse_operands_p): New function.
3317 (maybe_legitimize_operands): Try to reuse the results for
3320 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3322 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3323 Add {q} suffix to insn mnemonic.
3325 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3327 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3328 (msp430_warn_func_return): New.
3330 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3332 * fold-const.c (tree_nonzero_bits): New function.
3333 * fold-const.h (tree_nonzero_bits): Likewise.
3334 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3335 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3337 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3341 * varasm.c (assemble_alias): Check ifunc_resolver only on
3344 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3347 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3348 when memory input operand is handled.
3350 2018-05-24 Luis Machado <luis.machado@linaro.org>
3352 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3354 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3356 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3358 * match.pd: Delay FMA folds until after vectorization.
3360 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3363 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3364 address check not strict.
3366 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3368 * gimple-match.h (gimple_match_op): New class.
3369 (mprts_hook): Replace parameters with a gimple_match_op *.
3370 (maybe_build_generic_op): Likewise.
3371 (gimple_simplified_result_is_gimple_val): Replace parameters with
3372 a const gimple_match_op *.
3373 (gimple_simplify): Replace code_helper * and tree * parameters with
3374 a gimple_match_op * parameter.
3375 (gimple_resimplify1): Replace code_helper *, tree and tree *
3376 parameters with a gimple_match_op * parameter.
3377 (gimple_resimplify2): Likewise.
3378 (gimple_resimplify3): Likewise.
3379 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3380 parameters with a gimple_match_op * parameter.
3381 * gimple-match-head.c (gimple_simplify): Change prototypes of
3382 auto-generated functions to take a gimple_match_op * instead of
3383 separate code_helper * and tree * parameters. Make the same
3384 change in the top-level overload and update calls to the
3385 gimple_resimplify routines. Update calls to the auto-generated
3386 functions and to maybe_push_res_to_seq in the publicly-facing
3387 operation-specific gimple_simplify overloads.
3388 (gimple_match_op::MAX_NUM_OPS): Define.
3389 (gimple_resimplify1): Replace rcode and ops with a single res_op
3390 parameter. Update call to gimple_simplify.
3391 (gimple_resimplify2): Likewise.
3392 (gimple_resimplify3): Likewise.
3393 (mprts_hook): Replace parameters with a gimple_match_op *.
3394 (maybe_build_generic_op): Likewise.
3395 (build_call_internal): Replace type, nargs and ops with
3396 a gimple_match_op *.
3397 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3398 with a single gimple_match_op *. Update calls to mprts_hook,
3399 build_call_internal and gimple_simplified_result_is_gimple_val.
3400 Factor out code that is common to the tree_code and combined_fn cases.
3401 * genmatch.c (expr::gen_transform): Replace tem_code and
3402 tem_ops with a gimple_match_op called tem_op. Update calls
3403 to the gimple_resimplify functions and maybe_push_res_to_seq.
3404 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3405 res_ops. Update call to the gimple_resimplify functions.
3406 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3407 (decision_tree::gen): Make the functions take a gimple_match_op *
3408 called res_op instead of separate res_code and res_ops parameters.
3409 Update call accordingly.
3410 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3411 and ops with a single res_op parameter. Update calls to
3412 maybe_build_generic_op and maybe_push_res_to_seq.
3413 (fold_stmt_1): Update calls to gimple_simplify and
3414 replace_stmt_with_simplification.
3415 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3416 and gimple_simplified_result_is_gimple_val.
3417 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3419 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3420 with a gimple_match_op *.
3421 (vn_nary_build_or_lookup): Likewise. Update call to
3422 vn_nary_build_or_lookup_1.
3423 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3424 gimple_match_op *. Update calls to the gimple_resimplify routines
3425 and to gimple_simplified_result_is_gimple_val.
3426 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3427 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3428 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3429 (visit_nary_op): Likewise.
3430 (visit_reference_op_load): Likewise.
3432 2018-05-23 Luis Machado <luis.machado@linaro.org>
3434 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3435 modifier for printing the step amount.
3437 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3440 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3443 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3445 * doc/sourcebuild.texi (Endianness): New subsubsection.
3447 2018-05-23 Luis Machado <luis.machado@linaro.org>
3449 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3450 <prefetch_dynamic_strides>: New const bool field.
3451 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3452 prefetch_dynamic_strides.
3453 (exynosm1_prefetch_tune): Likewise.
3454 (thunderxt88_prefetch_tune): Likewise.
3455 (thunderx_prefetch_tune): Likewise.
3456 (thunderx2t99_prefetch_tune): Likewise.
3457 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3459 (aarch64_override_options_internal): Update to set
3460 PARAM_PREFETCH_DYNAMIC_STRIDES.
3461 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3462 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3463 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3464 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3465 prefetch-dynamic-strides setting.
3467 2018-05-23 Luis Machado <luis.machado@linaro.org>
3469 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3470 <minimum_stride>: New const int field.
3471 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3472 minimum_stride field defaulting to -1.
3473 (exynosm1_prefetch_tune): Likewise.
3474 (thunderxt88_prefetch_tune): Likewise.
3475 (thunderx_prefetch_tune): Likewise.
3476 (thunderx2t99_prefetch_tune): Likewise.
3477 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3478 <default_opt_level>: Set to 3.
3479 (aarch64_override_options_internal): Update to set
3480 PARAM_PREFETCH_MINIMUM_STRIDE.
3481 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3482 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3483 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3484 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3485 stride is constant and is below the minimum stride threshold.
3487 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3489 * config/arm/arm-cpus.in (mode26): Delete.
3490 (armv4): Delete mode26 reference.
3491 * config/arm/arm.c (arm_configure_build_target): Delete use of
3494 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3496 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3498 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3499 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3501 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3502 (floatunsdidf2): Ditto.
3504 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3506 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3507 (fixuns_trunc<mode>si2_avx512f): Ditto.
3508 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3509 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3510 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3512 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3514 PR rtl-optimization/79985
3515 * df-scan.c (df_insn_refs_collect): Remove special case for
3516 global registers and asm statements.
3518 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3520 * extend.texi (Global Register Variables): Rewrite the bullet list.
3521 Note that the register is available for allocation. Note that access
3522 via inline asm must use constraints. Add note about async-signal
3523 handlers. Remove paragraph about automagic register selection.
3525 2018-05-23 Richard Biener <rguenther@suse.de>
3527 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3528 of fixed offset from memset VN.
3530 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3532 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3534 (alloc_cand_and_find_basis): Initialize first_interp field.
3535 (slsr_process_mul): Modify first_interp field.
3536 (slsr_process_add): Likewise.
3537 (slsr_process_cast): Modify first_interp field for each new
3539 (slsr_process_copy): Likewise.
3540 (dump_candidate): Dump first_interp field.
3541 (replace_mult_candidate): Process all interpretations, not just
3543 (replace_rhs_if_not_dup): Likewise.
3544 (replace_one_candidate): Likewise.
3546 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3548 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3550 (aarch64_needs_frame_chain): New function.
3551 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3553 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3556 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3557 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3558 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3559 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3560 as true for strict-align.
3561 (aarch64_can_inline_p): Perform checks even when callee has no
3562 attributes to check for strict alignment.
3563 * doc/extend.texi (AArch64 Function Attributes): Document
3565 * doc/invoke.texi: (AArch64 Options): Likewise.
3567 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3569 PR tree-optimization/85853
3570 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3571 the handling of the root of the node to...
3572 (vect_slp_analyze_node_operations_1): ...this new function,
3573 and run the whole thing with the child nodes' def types
3574 set according to their SLP node's def type.
3576 2018-05-23 Richard Biener <rguenther@suse.de>
3579 * tree-data-ref.c (create_runtime_alias_checks): Defer
3580 and ignore overflow warnings.
3582 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3584 PR tree-optimization/85822
3585 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3588 2018-05-23 Richard Biener <rguenther@suse.de>
3590 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3591 memset constants via native_interpret_expr.
3593 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3596 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3598 (cgraph_node::create_alias): Likewise.
3599 (cgraph_node::get_availability): Check ifunc_resolver instead
3600 of looking up ifunc attribute.
3601 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3602 * varasm.c (do_assemble_alias): Likewise.
3603 (assemble_alias): Likewise.
3604 (default_binds_local_p_3): Likewise.
3605 * cgraph.h (cgraph_node): Add ifunc_resolver.
3606 (cgraph_node::only_called_directly_or_aliased_p): Return false
3608 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3610 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3611 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3612 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3613 instead of looking up ifunc attribute.
3615 2018-05-22 Luis Machado <luis.machado@linaro.org>
3617 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3619 2018-05-22 Martin Sebor <msebor@redhat.com>
3622 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3623 only when expasion succeeds.
3624 (expand_builtin_strcmp): Same.
3625 (expand_builtin_strncmp): Same.
3627 2018-05-22 Martin Sebor <msebor@redhat.com>
3629 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3631 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3632 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3634 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3635 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3637 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3638 Add check for legitimate_address.
3639 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3640 (aarch64_swap_ldrstr_operands): New.
3641 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3644 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3645 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3647 * config/aarch64/aarch64.md: New patterns to generate stp
3649 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3650 single words and double words.
3651 (load_pair_sw, load_pair_dw): Likewise.
3652 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3654 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3656 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3657 for different mode ldpstp and add peephole for merged zero stores.
3659 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3661 (aarch64_gen_store_pair): Rename calls to match new patterns.
3662 (aarch64_gen_load_pair): Rename calls to match new patterns.
3663 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3664 (load_pair<DREG:mode><DREG2:mode>): ... This.
3665 (store_pair<mode>): Rename to...
3666 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3667 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3669 (V_INT_EQUIV): Handle SImode.
3670 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3673 2018-05-22 Martin Sebor <msebor@redhat.com>
3676 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3677 or ajust the presumed bound on an operation to avoid unnecessary
3680 2018-05-22 Martin Sebor <msebor@redhat.com>
3682 PR tree-optimization/85826
3683 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3684 assuming that a DECL necesarily has a constant size.
3686 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3689 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3691 2018-05-22 Richard Biener <rguenther@suse.de>
3693 PR tree-optimization/85834
3694 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3695 non-constant and non-zero memset arguments.
3697 2018-05-22 Martin Liska <mliska@suse.cz>
3700 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3702 2018-05-22 Richard Biener <rguenther@suse.de>
3704 PR tree-optimization/85863
3705 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3706 comparisons when vectype is specified.
3707 (vectorizable_condition): Do not specify vectype for
3708 vect_is_simple_cond when SLP vectorizing.
3710 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3713 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3714 define __ibm128 as long double.
3715 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3716 as a distinct type when IEEE 128-bit support is enabled.
3717 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3718 128-bit types to use the correct functions.
3719 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3720 convert between 128-bit floating point types that have different
3721 modes but the same representation, instead of using gen_lowpart to
3723 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3725 (IFKF_reg): New attributes to give the register constraints for
3727 (extend<mode>tf2_internal): New insns to mark an explicit
3728 conversion between 128-bit floating point types that have a
3729 different mode but share the same representation.
3731 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3733 PR tree-optimization/85814
3734 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3735 a null return from get_strinfo when unsharing the next
3736 strinfo in the chain.
3738 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3741 * varasm.c (weak_finish): Clean up weak_decls.
3743 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3745 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3746 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3747 UNSPEC_UADALP values.
3748 * config/aarch64/iterators.md (ABAL): New int iterator.
3751 (sur): Add mappings for the above.
3752 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3754 (aarch64_<sur>abal<mode>_4): Likewise.
3755 (aarch64_<sur>adalp<mode>_3): Likewise.
3756 (<sur>sadv16qi): New define_expand.
3758 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3760 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3761 (*movdf_internal): Ditto.
3762 (*rcpsf2_sse): Ditto.
3763 (*rsqrtsf2_sse): Ditto.
3764 (*sqrt<mode>2_sse): Ditto.
3766 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3768 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3770 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3771 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3772 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3773 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3775 * config/aarch64/arm_neon.h: Likewise.
3776 * config/aarch64/iterators.md (VQ_I): New.
3778 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3780 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3782 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3785 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3787 * config/nds32/constraints.md (S): New constraint.
3788 * config/nds32/nds32.md (call_internal): Use constraint S.
3789 (call_value_internal): Likewise.
3790 (sibcall_internal): Likewise.
3791 (sibcall_value_internal): Likewise.
3793 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3794 Chung-Ju Wu <jasonwucj@gmail.com>
3796 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3799 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3800 Chung-Ju Wu <jasonwucj@gmail.com>
3802 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3803 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3804 (nds32_rtx_costs_impl): Simplify.
3805 (nds32_address_cost_impl): Simplify.
3806 (nds32_init_rtx_costs): New function.
3807 (nds32_rtx_costs_speed_prefer): Likewise.
3808 (nds32_rtx_costs_size_prefer): Likewise.
3809 (nds32_address_cost_speed_prefer): Likewise.
3810 (nds32_address_cost_speed_fwprop): Likewise.
3811 (nds32_address_cost_size_prefer): Likewise.
3812 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3813 * config/nds32/nds32.c (nds32_option_override): Use
3814 nds32_init_rtx_costs function.
3816 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3818 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3819 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3820 (TARGET_PIPELINE_N8): Likewise.
3821 (TARGET_PIPELINE_N10): Likewise.
3822 (TARGET_PIPELINE_N13): Likewise.
3823 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3825 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3827 * config/nds32/nds32-fpu.md: Update copyright year.
3829 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3833 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3835 * config/nds32/nds32.c
3836 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3837 * config/nds32/nds32.opt (minline-asm-r15): New option.
3839 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3841 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3843 * config/nds32/nds32.md (abssi2): New pattern.
3845 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3847 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3848 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3849 Merge insn pattern from sse_cvtsi2ss<round_name> and
3850 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3851 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3852 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3853 using SWI48 mode iterator.
3854 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3855 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3856 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3857 pattern from sse_cvttss2si<round_saeonly_name>
3858 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3859 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3860 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3861 using SWI48 mode iterator.
3862 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3863 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3864 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3865 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3866 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3867 using SWI48 mode iterator.
3868 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3869 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3870 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3871 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3872 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3873 SWI48 mode iterator.
3874 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3875 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3876 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3877 pattern from sse_cvttsd2si<round_saeonly_name>
3878 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3880 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3882 * config/nds32/nds32-md-auxiliary.c
3883 (nds32_valid_smw_lwm_base_p): Refine.
3884 (nds32_output_smw_single_word): Refine.
3885 (nds32_output_smw_double_word): New.
3886 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3888 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3890 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3891 (nds32_output_stack_pop): Refine.
3892 (nds32_expand_unaligned_load): Refine.
3893 (nds32_expand_unaligned_store): Refine.
3895 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3896 Chung-Ju Wu <jasonwucj@gmail.com>
3898 * config/nds32/constants.md: Add TP_REGNUM constant.
3899 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3900 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3902 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3903 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3904 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3905 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3906 and PIC code generation.
3907 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3909 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3911 * config/nds32/nds32.md: Support TLS and PIC.
3912 * config/nds32/nds32.c: Support TLS and PIC.
3913 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3914 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3917 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3919 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3920 mode with E_ prefix.
3922 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3923 Chung-Ju Wu <jasonwucj@gmail.com>
3925 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3926 * config/nds32/nds32-md-auxiliary.c
3927 (symbolic_reference_mentioned_p): New.
3928 (nds32_legitimize_ict_address): New.
3929 (nds32_expand_ict_move): New.
3930 (nds32_indirect_call_referenced_p): New.
3931 (nds32_symbol_binds_local_p): Delete.
3932 (nds32_long_call_p): Modify.
3933 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3934 * config/nds32/nds32-protos.h
3935 (symbolic_reference_mentioned_p): Declare.
3936 (nds32_legitimize_ict_address): Declare.
3937 (nds32_expand_ict_move): Declare.
3938 (nds32_indirect_call_referenced_p): Declare.
3939 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3940 (nds32_relax_group): Use nds32_ict_const_p as condition.
3941 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3942 (nds32_asm_file_start): Output ict_model directive in asm code.
3943 (nds32_legitimate_address_p): Consider indirect call.
3944 (nds32_print_operand): Consider indirect call.
3945 (nds32_print_operand_address): Consider indirect call.
3946 (nds32_insert_attributes): Handle "indirect_call" attribute.
3947 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3948 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3949 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3950 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3951 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3952 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3953 (TARGET_ICT_MODEL_SMALL): Define.
3954 (TARGET_ICT_MODEL_LARGE): Define.
3955 * config/nds32/nds32.md (movsi): Consider ict model.
3956 (call, call_value): Consider ict model.
3957 (sibcall, sibcall_value): Consider ict model.
3958 * config/nds32/nds32.opt (mict-model): New option.
3959 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3962 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3963 Monk Chiang <sh.chiang04@gmail.com>
3964 Jim Wilson <jimw@sifive.com>
3966 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3967 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3968 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3969 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3970 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3971 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3972 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3973 compute save_libcall_adjustment properly.
3974 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3975 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3976 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3977 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3978 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3979 (ABI_SPEC): Handle mabi=ilp32e.
3980 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3981 (RVE): Add RVE mask.
3982 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3983 <-march>: Add rv32e as an example.
3985 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3988 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3989 (intra_create_variable_infos): Handle C++ constructors.
3991 2018-05-18 Martin Liska <mliska@suse.cz>
3993 * passes.def: Remove a redundant pass.
3995 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3998 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4000 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4002 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4004 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4006 (armv2, armv2a, armv3, armv3m): Delete architectures.
4007 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4008 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4009 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4011 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4012 (*mulsidi3adddi): Likewise.
4013 (mulsidi3): Likewise.
4014 (*mulsidi3_nov6): Likewise.
4015 (umulsidi3): Likewise.
4016 (umulsidi3_nov6): Likewise.
4017 (umaddsidi4): Likewise.
4018 (*umulsidi3adddi): Likewise.
4019 (smulsi3_highpart): Likewise.
4020 (*smulsi3_highpart_nov6): Likewise.
4021 (umulsi3_highpart): Likewise.
4022 (*umulsi3_highpart_nov6): Likewise.
4023 * config/arm/arm.h (arm_arch3m): Delete.
4024 * config/arm/arm.c (arm_arch3m): Delete.
4025 (arm_option_override_internal): Update armv3-related comment.
4026 (arm_configure_build_target): Delete use of isa_bit_mode32.
4027 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4028 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4029 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4031 (mulusa3): Likewise.
4032 * config/arm/arm-protos.h (arm_arch3m): Delete.
4033 * config/arm/arm-tables.opt: Regenerate.
4034 * config/arm/arm-tune.md: Likewise.
4035 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4036 deleted architectures.
4038 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4040 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4041 (armv5t, armv5te): New features.
4042 (ARMv5, ARMv5e): Delete fgroups.
4043 (ARMv5t, ARMv5te): Adjust for above changes.
4045 (armv5, armv5e): Delete arches.
4046 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4048 (*call_reg_arm): Likewise.
4049 (*call_value_reg_armv5): Likewise.
4050 (*call_value_reg_arm): Likewise.
4051 (*call_symbol): Likewise.
4052 (*call_value_symbol): Likewise.
4053 (*sibcall_insn): Likewise.
4054 (*sibcall_value_insn): Likewise.
4056 (prefetch): Likewise.
4057 (define_split and define_peephole2 dependent on arm_arch5):
4059 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4061 (TARGET_ARM_QBIT): Likewise.
4062 (TARGET_DSP_MULTIPLY): Likewise.
4063 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4064 (arm_arch5, arm_arch5e): Delete.
4065 (arm_arch5t, arm_arch5te): Declare.
4066 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4067 (arm_arch5t): Declare.
4068 (arm_option_reconfigure_globals): Update for the above.
4069 (arm_options_perform_arch_sanity_checks): Update comment, replace
4070 use of arm_arch5 with arm_arch5t.
4071 (use_return_insn): Likewise.
4072 (arm_emit_call_insn): Likewise.
4073 (output_return_instruction): Likewise.
4074 (arm_final_prescan_insn): Likewise.
4075 (arm_coproc_builtin_available): Likewise.
4076 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4077 arm_arch5e with arm_arch5t and arm_arch5te.
4078 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4079 (arm_arch5t, arm_arch5te): Declare.
4080 * config/arm/arm-tables.opt: Regenerate.
4081 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4082 * config/arm/t-multilib: Likewise.
4083 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4084 instead of arm_arch5.
4085 (*call_reg_thumb1): Likewise.
4086 (*call_value_reg_thumb1_v5): Likewise.
4087 (*call_value_reg_thumb1): Likewise.
4088 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4090 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4092 2018-05-18 Martin Liska <mliska@suse.cz>
4094 PR gcov-profile/84846
4095 * doc/gcov.texi: Document -t option of gcov tool.
4097 2018-05-18 Martin Liska <mliska@suse.cz>
4099 PR gcov-profile/84846
4100 * gcov.c (print_usage): Add new -t option.
4101 (process_args): Handle the option.
4102 (generate_results): Use stdout as output when requested by
4105 2018-05-18 Martin Liska <mliska@suse.cz>
4107 PR gcov-profile/84846
4108 * coverage.c (coverage_init): Write PWD to .gcno file.
4109 * doc/gcov.texi: Document how working directory is printed.
4110 * gcov-dump.c (dump_gcov_file): Print PWD.
4111 * gcov.c (output_intermediate_file): Likewise.
4112 (read_graph_file): Read PWD string.
4113 (output_lines): Print PWD.
4115 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4118 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4119 for retval and return false if all args to phi are zero.
4121 2018-05-18 Richard Biener <rguenther@suse.de>
4123 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4125 (evrp_dom_walker::before_dom_children): Call it.
4127 2018-05-18 Richard Biener <rguenther@suse.de>
4129 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4130 results when processing array refs with variable index.
4132 2018-05-18 Toon Moene <toon@moene.org>
4134 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4135 directly after that of -floop-interchange. Indicate that both
4136 options are enabled by default when specifying -O3.
4138 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4140 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4141 iterator. Delete separate integer-mode vec_set<mode> expander.
4142 (aarch64_simd_vec_setv2di): Delete.
4143 (vec_setv2di): Delete.
4144 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4145 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4146 the "w, r" alternative.
4148 2018-05-18 Martin Liska <mliska@suse.cz>
4150 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4151 * tree-pass.h (make_pass_lower_switch_O0): New function.
4152 * tree-switch-conversion.c (node_has_low_bound): Remove.
4153 (node_has_high_bound): Likewise.
4154 (node_is_bounded): Likewise.
4155 (class pass_lower_switch): Make it a template type and create
4157 (pass_lower_switch::execute): Add template argument.
4158 (make_pass_lower_switch): New function.
4159 (make_pass_lower_switch_O0): New function.
4160 (do_jump_if_equal): Remove.
4161 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4162 all the hard work to tree optimization passes.
4164 2018-05-18 Martin Liska <mliska@suse.cz>
4166 * dbgcnt.c (limit_low): Renamed from limit.
4167 (limit_high): New variable.
4168 (dbg_cnt_is_enabled): Check for upper limit.
4169 (dbg_cnt): Adjust dumping.
4170 (dbg_cnt_set_limit_by_index): Add new argument for high
4172 (dbg_cnt_set_limit_by_name): Likewise.
4173 (dbg_cnt_process_single_pair): Parse new format.
4174 (dbg_cnt_process_opt): Use strtok.
4175 (dbg_cnt_list_all_counters): Remove 'value' and add
4177 * doc/invoke.texi: Document changes.
4179 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4181 * doc/sourcebuild.texi (scalar_all_fma): Document.
4182 * tree.def (FMA_EXPR): Delete.
4183 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4184 * internal-fn.c (ternary_direct): New macro.
4185 (expand_ternary_optab_fn): Likewise.
4186 (direct_ternary_optab_supported_p): Likewise.
4187 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4188 * builtins.c (fold_builtin_fma): Delete.
4189 (fold_builtin_3): Don't call it.
4190 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4191 * expr.c (expand_expr_real_2): Likewise.
4192 * fold-const.c (operand_equal_p): Likewise.
4193 (fold_ternary_loc): Likewise.
4194 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4195 * gimple.c (DEFTREECODE): Likewise.
4196 * gimplify.c (gimplify_expr): Likewise.
4197 * optabs-tree.c (optab_for_tree_code): Likewise.
4198 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4199 * tree-eh.c (operation_could_trap_p): Likewise.
4200 (stmt_could_throw_1_p): Likewise.
4201 * tree-inline.c (estimate_operator_cost): Likewise.
4202 * tree-pretty-print.c (dump_generic_node): Likewise.
4203 (op_code_prio): Likewise.
4204 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4205 * tree-ssa-operands.c (get_expr_operands): Likewise.
4206 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4207 * fold-const-call.h (fold_fma): Delete.
4208 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4209 CFN_FNMA and CFN_FNMS.
4211 * genmatch.c (combined_fn): New enum.
4212 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4213 (commutative_op): New function.
4214 (commutate): Use it. Handle more than 2 operands.
4215 (dt_operand::gen_gimple_expr): Use commutative_op.
4216 (parser::parse_expr): Allow :c to be used with non-binary
4217 operators if the commutative operand is known.
4218 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4219 CFN_FMS, CFN_FNMA and CFN_FNMS.
4220 (backprop::process_assign_use): Remove FMA_EXPR handling.
4221 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4222 (gen_hsa_fma): New function.
4223 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4224 IFN_FNMA and IFN_FNMS.
4225 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4226 * gimple-fold.h (follow_all_ssa_edges): Declare.
4227 * gimple-fold.c (follow_all_ssa_edges): New function.
4228 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4229 gimple_build interface and use follow_all_ssa_edges to fold the result.
4230 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4231 instead of checking for optabs directly.
4232 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4233 rather than FMA_EXPRs.
4234 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4235 call to IFN_FMA instead of an FMA_EXPR.
4237 2018-05-17 Jim Wilson <jimw@sifive.com>
4239 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4240 sign extended promoted subreg, and the range does not have the sign bit
4241 set, then do a sign extend.
4243 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4244 test, check for sign extended subreg and/or constant operands, and
4245 do a sign extend in that case.
4247 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4249 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4250 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4252 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4253 Change logics_shift_reg to logics_shift_imm.
4254 (thunderx2t99_fp_loadpair_basic): Delete.
4255 (thunderx2t99_fp_storepair_basic): Delete.
4256 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4257 (thunderx2t99_asimd_polynomial): Delete.
4258 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4259 and neon_fp_mul_d_scalar_q.
4260 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4261 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4262 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4263 (thunderx2t99_asimd_lut): Add missing tbl types.
4264 (thunderx2t99_asimd_ext): Delete.
4265 (thunderx2t99_asimd_load1_1_mult): Delete.
4266 (thunderx2t99_asimd_load1_2_mult): Delete.
4267 (thunderx2t99_asimd_load1_ldp): New.
4268 (thunderx2t99_asimd_load1): New.
4269 (thunderx2t99_asimd_load2): Add missing *load2* types.
4270 (thunderx2t99_asimd_load3): New.
4271 (thunderx2t99_asimd_load4): New.
4272 (thunderx2t99_asimd_store1_1_mult): Delete.
4273 (thunderx2t99_asimd_store1_2_mult): Delete.
4274 (thunderx2t99_asimd_store2_mult): Delete.
4275 (thunderx2t99_asimd_store2_onelane): Delete.
4276 (thunderx2t99_asimd_store_stp): New.
4277 (thunderx2t99_asimd_store1): New.
4278 (thunderx2t99_asimd_store2): New.
4279 (thunderx2t99_asimd_store3): New.
4280 (thunderx2t99_asimd_store4): New.
4282 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4284 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4285 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4287 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4288 Segher Boessenkool <segher@kernel.crashing.org>
4291 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4294 2018-05-17 Richard Biener <rguenther@suse.de>
4296 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4297 for pruning loop and prune defs feeding only already visited PHIs.
4299 2018-05-17 Richard Biener <rguenther@suse.de>
4301 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4303 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4304 Richard Biener <rguenther@suse.de>
4306 PR tree-optimization/85793
4307 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4308 for VMAT_ELEMENTWISE.
4310 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4312 * internal-fn.h (lookup_internal_fn): Declare
4313 * internal-fn.c (lookup_internal_fn): New function.
4314 * gimple.c (gimple_build_call_from_tree): Handle calls to
4316 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4317 internal function names.
4318 * tree-pretty-print.c (dump_generic_node): Likewise.
4319 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4321 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4323 * gimple-fold.h (gimple_build): Make the function forms take
4324 combined_fn rather than built_in_function.
4325 (gimple_simplify): Likewise.
4326 * gimple-match-head.c (gimple_simplify): Likewise.
4327 * gimple-fold.c (gimple_build): Likewise.
4328 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4329 rather than gimple_build_call_internal.
4330 (get_initial_defs_for_reduction): Likewise.
4331 (vect_create_epilog_for_reduction): Likewise.
4332 (vectorizable_live_operation): Likewise.
4334 2018-05-17 Martin Liska <mliska@suse.cz>
4336 * gimple-ssa-sprintf.c (format_directive): Do not use
4337 space in between 'G_' and '('.
4339 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4342 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4343 even if the mask is not all ones.
4346 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4348 (ix86_gimple_fold_builtin): Likewise.
4351 * config/i386/i386.c: Include tree-vector-builder.h.
4352 (ix86_vector_shift_count): New function.
4353 (ix86_fold_builtin): Fold shift builtins by scalar count.
4354 (ix86_gimple_fold_builtin): Likewise.
4356 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4357 _mm512_setzero): New intrinsics.
4359 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4360 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4362 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4363 code generation for cases where splatting a value is not useful.
4364 * simplify-rtx.c (simplify_ternary_operation): Simplify
4365 vec_merge across a vec_duplicate and a paradoxical subreg forming
4366 a vector mode to a vec_concat.
4368 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4370 * config.gcc: Support "goldmont-plus".
4371 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4373 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4374 PROCESSOR_GOLDMONT_PLUS.
4375 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4376 (processor_target_table): Add "goldmont-plus".
4377 (PTA_GOLDMONT_PLUS): Define.
4378 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4379 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4380 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4381 (fold_builtin_cpu): Add "goldmont-plus".
4382 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4383 (ix86_option_override_internal): Add "goldmont-plus".
4384 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4385 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4386 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4387 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4389 2018-05-17 Richard Biener <rguenther@suse.de>
4391 PR tree-optimization/85757
4392 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4393 remove defs that only feed that PHI from further processing.
4395 2018-05-16 Jim Wilson <jimw@sifive.com>
4397 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4399 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4401 2018-05-16 Mark Wielaard <mark@klomp.org>
4403 * dwarf2out.c (count_index_strings): New function.
4404 (output_indirect_strings): Call count_index_strings and generate
4405 header for dwarf_version >= 5.
4407 2018-05-16 Mark Wielaard <mark@klomp.org>
4409 * dwarf2out.c (dwarf_FORM): New function.
4410 (set_indirect_string): Use dwarf_FORM.
4411 (reset_indirect_string): Likewise.
4412 (size_of_die): Likewise.
4413 (value_format): Likewise.
4414 (output_die): Likewise.
4415 (add_skeleton_AT_string): Likewise.
4416 (output_macinfo_op): Likewise.
4417 (index_string): Likewise.
4418 (output_index_string_offset): Likewise.
4419 (output_index_string): Likewise.
4420 (count_index_strings): Likewise.
4422 2018-05-16 Carl Love <cel@us.ibm.com>
4424 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4425 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4427 2018-05-16 Martin Jambor <mjambor@suse.cz>
4429 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4430 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4432 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4434 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4435 (fnma<mode>4): Likewise.
4436 (fms<mode>4): Likewise.
4437 (fnms<mode>4): Likewise.
4438 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4439 (aarch64_fnma<mode>4): Likewise.
4440 (aarch64_fms<mode>4): Likewise.
4441 (aarch64_fnms<mode>4): Likewise.
4442 (aarch64_fnmadd<mode>4): Likewise.
4444 2018-05-16 Jason Merrill <jason@redhat.com>
4446 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4448 2018-05-16 Richard Biener <rguenther@suse.de>
4450 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4451 (dump_stmt_cost): Declare.
4452 (add_stmt_cost): Dump cost we add.
4453 (add_stmt_costs): New function.
4454 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4456 (vect_analyze_stmt): Adjust prototype.
4457 (vectorizable_condition): Likewise.
4458 (vectorizable_live_operation): Likewise.
4459 (vectorizable_reduction): Likewise.
4460 (vectorizable_induction): Likewise.
4461 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4462 cost vector to pass to vectorizable_ and record afterwards.
4463 (vect_model_reduction_cost): Take cost vector argument and adjust.
4464 (vect_model_induction_cost): Likewise.
4465 (vectorizable_reduction): Likewise.
4466 (vectorizable_induction): Likewise.
4467 (vectorizable_live_operation): Likewise.
4468 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4469 SLP_TREE_NUMBER_OF_VEC_STMTS.
4470 (vect_analyze_slp_cost_1): Remove.
4471 (vect_analyze_slp_cost): Likewise.
4472 (vect_slp_analyze_node_operations): Take visited args and
4473 a target cost vector. Avoid processing already visited stmt sets.
4474 (vect_slp_analyze_operations): Use a local cost vector to gather
4475 costs and register those of non-discarded instances.
4476 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4477 (vect_schedule_slp_instance): Remove copying of
4478 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4480 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4481 adding cost. Record cost entry location.
4482 (vect_prologue_cost_for_slp_op): Function to compute cost of
4483 a constant or invariant generated for SLP vect in the prologue,
4484 split out from vect_analyze_slp_cost_1.
4485 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4486 (vect_model_promotion_demotion_cost): Likewise.
4487 (vect_model_store_cost): Likewise, make static.
4488 (vect_model_load_cost): Likewise.
4489 (vectorizable_bswap): Add cost vector arg and adjust.
4490 (vectorizable_call): Likewise.
4491 (vectorizable_simd_clone_call): Likewise.
4492 (vectorizable_conversion): Likewise.
4493 (vectorizable_assignment): Likewise.
4494 (vectorizable_shift): Likewise.
4495 (vectorizable_operation): Likewise.
4496 (vectorizable_store): Likewise.
4497 (vectorizable_load): Likewise.
4498 (vectorizable_condition): Likewise.
4499 (vectorizable_comparison): Likewise.
4500 (can_vectorize_live_stmts): Likewise.
4501 (vect_analyze_stmt): Likewise.
4502 (vect_transform_stmt): Adjust calls to vectorizable_*.
4503 * tree-vectorizer.c: Include gimple-pretty-print.h.
4504 (dump_stmt_cost): New function.
4506 2018-05-16 Richard Biener <rguenther@suse.de>
4508 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4509 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4510 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4511 (check_name): New callback.
4512 (dse_classify_store): Track cycles via a visited bitmap of PHI
4513 defs and simplify handling of in-loop and across loop dead stores
4514 and properly fail for loop-variant refs. Handle byte-tracking with
4515 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4518 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4520 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4521 (vect_get_mask_type_for_stmt): Likewise.
4522 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4524 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4525 to determine the statement's vector type and the vector type that
4526 should be used for calculating nunits. Deal with cases in which
4527 the type has to be deferred.
4528 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4529 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4530 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4531 (vect_determine_vf_for_stmt): New functions, split out from...
4532 (vect_determine_vectorization_factor): ...here.
4533 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4534 (vect_get_mask_type_for_stmt): New functions, split out from
4535 vect_determine_vectorization_factor.
4537 2018-05-16 Richard Biener <rguenther@suse.de>
4539 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4540 verify the [VEC_]COND_EXPR embedded comparison.
4542 2018-05-15 Martin Sebor <msebor@redhat.com>
4544 PR tree-optimization/85753
4545 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4546 RECORD_TYPE in addition to ARRAY_TYPE.
4548 2018-05-15 Martin Sebor <msebor@redhat.com>
4551 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4553 2018-05-15 Richard Biener <rguenther@suse.de>
4555 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4556 add by_clobber_p one. Change algorithm to collect all defs
4557 representing uses we need to walk and try reducing them to
4558 a single one before failing.
4559 (dse_dom_walker::dse_optimize_stmt): Adjust.
4561 2018-05-13 Mark Wielaard <mark@klomp.org>
4563 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4564 (size_of_loc_descr): Likewise.
4565 (output_loc_operands): Likewise.
4566 (output_loc_operands_raw): Likewise.
4567 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4568 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4569 (hash_loc_operands): Likewise.
4570 (compare_loc_operands): Likewise.
4572 2018-05-14 Mark Wielaard <mark@klomp.org>
4574 * dwarf2out.c (count_index_addrs): New function.
4575 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4577 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4579 PR tree-optimization/83648
4580 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4581 return value as malloc candidate.
4583 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4586 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4587 param as true in call to suggest_attribute.
4589 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4591 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4592 -mreadonly-in-sdata.
4594 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4596 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4598 (aarch64_crypto_aesd_fused): Likewise.
4600 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4602 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4603 (movsi_aarch64): Likewise.
4604 (load_pairsi): Likewise.
4605 (load_pairdi): Likewise.
4606 (store_pairsi): Likewise.
4607 (store_pairdi): Likewise.
4608 (load_pairsf): Likewise.
4609 (load_pairdf): Likewise.
4610 (store_pairsf): Likewise.
4611 (store_pairdf): Likewise.
4612 (zero_extend): Likewise.
4613 (trunc): Swap alternatives.
4614 (fcvt_target): Add '?' to prefer w over r.
4616 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4619 * config/i386/i386.md: Disallow non-commutative arithmetics in
4620 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4621 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4622 in the peephole2 before it.
4624 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4626 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4627 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4628 (ix86_handle_option): Handle -mcldemote.
4629 * config.gcc: New header.
4630 * config/i386/cldemoteintrin.h: New file.
4631 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4632 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4634 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4635 OPTION_MASK_ISA_CLDEMOTE.
4636 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4637 (ix86_valid_target_attribute_inner_p): Ditto.
4638 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4639 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4640 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4641 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4642 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4644 * config/i386/i386.opt: Add -mcldemote.
4645 * config/i386/x86intrin.h: New header.
4646 * doc/invoke.texi: Add -mcldemote.
4648 2018-05-14 Richard Biener <rguenther@suse.de>
4650 * doc/match-and-simplify.texi: Adjust :s documentation.
4652 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4654 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4655 intended memcpy size.
4656 (REORDER_45): Likewise.
4658 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4660 * sort.cc: New file.
4661 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4662 * vec.c (qsort_chk): Use gcc_qsort.
4663 * Makefile.in (OBJS-libcommon): Add sort.o.
4664 (build/sort.o): New target. Use it...
4665 (BUILD_RTL): ... here, and...
4666 (build/gencfn-macros): ... here, and...
4667 (build/genmatch): ... here.
4669 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4670 Chung-Ju Wu <jasonwucj@gmail.com>
4672 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4673 * config/nds32/nds32-graywolf.md: New file.
4674 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4675 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4677 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4678 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4679 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4680 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4681 * config/nds32/pipelines.md: Include n15 settings.
4683 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4684 Chung-Ju Wu <jasonwucj@gmail.com>
4686 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4687 * config/nds32/nds32-n13.md: New file.
4688 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4689 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4691 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4692 * config/nds32/nds32.md (pipeline_model): Add n13.
4693 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4694 * config/nds32/pipelines.md: Include n13 settings.
4696 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4697 Chung-Ju Wu <jasonwucj@gmail.com>
4699 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4700 * config/nds32/nds32-n10.md: New file.
4701 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4702 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4704 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4705 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4706 * config/nds32/nds32.md (pipeline_model): Add n10.
4707 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4708 * config/nds32/pipelines.md: Include n10 settings.
4710 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4711 Kito Cheng <kito.cheng@gmail.com>
4712 Chung-Ju Wu <jasonwucj@gmail.com>
4714 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4715 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4716 Add enum values for DSP extension instructions.
4717 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4719 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4720 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4722 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4723 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4724 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4725 * config/nds32/nds32-intrinsic.md: Likewise.
4726 * config/nds32/nds32_intrinsic.h: Likewise.
4727 * config/nds32/nds32-md-auxiliary.c: Likewise.
4728 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4729 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4730 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4731 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4732 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4733 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4734 TYPE_DMAC in switch statement.
4735 * config/nds32/nds32.c: New checking and implementation for DSP
4736 extension instructions.
4737 * config/nds32/nds32.h: Likewise.
4738 * config/nds32/nds32.md: Likewise.
4739 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4740 * config/nds32/predicates.md: Implement new predicates for DSP
4743 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4745 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4746 Reformat alternatives and attributes so it is easier to identify
4747 which constraints/attributes go with which instruction.
4748 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4749 (mov<mode>_softfloat32, FMOVE64): Likewise.
4750 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4751 (mov<mode>_softfloat64, FMOVE64): Likewise.
4753 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4755 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4757 (Basic PowerPC Built-in Functions): The new name of the
4758 subsection previously known as "PowerPC Built-in Functions".
4759 (Basic PowerPC Built-in Functions Available on all Configurations):
4761 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4762 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4763 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4764 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4766 2018-05-11 Martin Jambor <mjambor@suse.cz>
4769 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4772 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4775 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4777 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4779 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4780 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4781 (ix86_handle_option): Handle -mwaitpkg.
4782 * config.gcc: New header.
4783 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4784 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4785 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4787 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4788 OPTION_MASK_ISA_WAITPKG.
4789 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4790 (ix86_option_override_internal): Add PTA_WAITPKG.
4791 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4792 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4793 IX86_BUILTIN_TPAUSE.
4794 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4795 __builtin_ia32_umwait and __builtin_ia32_tpause.
4796 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4797 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4798 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4799 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4800 UNSPECV_TPAUSE): New.
4801 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4802 * config/i386/i386.opt: Add -mwaitpkg.
4803 * config/i386/waitpkgintrin.h: New file.
4804 * config/i386/x86intrin.h: New header.
4805 * doc/invoke.texi: Add -mwaitpkg.
4807 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4810 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4812 (cortex-m0): Use armv6s-m isa.
4813 (cortex-m0plus): Likewise.
4814 (cortex-m1): Likewise.
4815 (cortex-m0.small-multiply): Likewise.
4816 (cortex-m0plus.small-multiply): Likewise.
4817 (cortex-m1.small-multiply): Likewise.
4819 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4820 Jakub Jelinek <jakub@redhat.com>
4822 PR tree-optimization/85692
4823 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4824 source permute as well.
4826 2018-05-11 Martin Liska <mliska@suse.cz>
4829 * doc/extend.texi: Document LLVM style format for no_sanitize
4832 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4834 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4835 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4836 (mode_supports_vsx_dform_quad): Likewise.
4837 (mode_supports_vmx_dform): Move these functions to be next to the
4838 other mode_supports functions.
4839 (mode_supports_dq_form): Likewise.
4840 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4841 mode_supports_dq_form.
4842 (reg_offset_addressing_ok_p): Likewise.
4843 (offsettable_ok_by_alignment): Likewise.
4844 (rs6000_legitimate_offset_address_p): Likewise.
4845 (legitimate_lo_sum_address_p): Likewise.
4846 (rs6000_legitimize_address): Likewise.
4847 (rs6000_legitimize_reload_address): Likewise.
4848 (rs6000_secondary_reload_inner): Likewise.
4849 (rs6000_preferred_reload_class): Likewise.
4850 (rs6000_output_move_128bit): Likewise.
4852 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4854 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4855 Generate SImode target register for null target.
4856 <case IX86_BUILTIN_XGETBV>: Ditto.
4857 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4858 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4860 2018-05-10 Carl Love <cel@us.ibm.com>
4862 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4863 dcbtt and dcbtstt if operands[2] is 0.
4865 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4868 * config/i386/sse.md (usadv64qi): New expander.
4870 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4872 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4873 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4874 -maltivec=be support.
4875 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4876 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4877 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4878 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4879 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4880 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4881 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4882 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4883 altivec_vsumsws): Adjust.
4884 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4885 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4886 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4888 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4889 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4890 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4891 (altivec_lve<VI_char>x): Delete expand.
4892 (*altivec_lve<VI_char>x_internal): Rename to...
4893 (altivec_lve<VI_char>x): ... this.
4894 (altivec_lvxl_<mode>): Delete expand.
4895 (*altivec_lvxl_<mode>_internal): Rename to ...
4896 (altivec_lvxl_<mode>): ... this.
4897 (altivec_stvxl_<mode>): Delete expand.
4898 (*altivec_stvxl_<mode>_internal): Rename to ...
4899 (altivec_stvxl_<mode>): ... this.
4900 (altivec_stve<VI_char>x): Delete expand.
4901 (*altivec_stve<VI_char>x_internal): Rename to ...
4902 (altivec_stve<VI_char>x): ... this.
4903 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4904 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4905 reduc_plus_scal_<mode>): Adjust.
4906 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4908 (rs6000_cpu_cpp_builtins): Adjust.
4909 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4910 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4911 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4912 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4913 -maltivec=be support.
4914 (rs6000_split_vec_extract_var): Adjust.
4915 (rs6000_split_v4si_init): Adjust.
4916 (swap_selector_for_mode): Delete.
4917 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4918 altivec_expand_stvex_be): Delete.
4919 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4920 -maltivec=be support.
4921 (rs6000_gimple_fold_builtin): Ditto.
4922 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4924 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4925 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4926 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4927 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4928 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4929 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4930 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4931 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4932 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4933 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4934 anonymous split): Adjust.
4935 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4936 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4938 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4940 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4941 when --with-gxx-include-dir is also specified.
4942 * configure: Regenerate.
4944 2018-05-09 Jim Wilson <jimw@sifive.com>
4947 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4948 * config/riscv/t-withmultilib: New.
4949 * config/riscv/withmultilib.h: New.
4950 * doc/install.texi: Document RISC-V --with-multilib-list support.
4952 2018-05-09 Richard Biener <rguenther@suse.de>
4954 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4956 (vect_bb_vectorization_profitable_p): Adjust. Compute
4957 actual scalar cost using the cost vector and the add_stmt_cost
4960 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4962 PR rtl-optimization/85645
4963 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4964 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4966 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4968 PR rtl-optimization/85645
4969 * shrink-wrap.c (spread_components): Return a boolean saying if
4970 anything was changed.
4971 (try_shrink_wrapping_separate): Iterate spread_components until
4972 nothing changes anymore.
4974 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4976 PR rtl-optimization/85645
4977 * regrename.c (build_def_use): Also kill the chains that include the
4978 destination of a REG_CFA_REGISTER note.
4980 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4982 PR rtl-optimization/85645
4983 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4984 insn that has a REG_CFA_REGISTER note.
4986 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4988 * cfgexpand.c (expand_clobber): New function.
4989 (expand_gimple_stmt_1): Use it.
4990 * tree-vect-stmts.c (vect_clobber_variable): New function,
4992 (vectorizable_simd_clone_call): ...here.
4993 (vectorizable_store): Emit a clobber either side of an
4994 IFN_STORE_LANES sequence.
4995 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4997 2018-05-09 Tom de Vries <tom@codesourcery.com>
5000 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5001 (define_insn "trap_if_false"): Add exit after trap.
5003 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5005 PR rtl-optimization/85638
5006 * bb-reorder.c: Include common/common-target.h.
5007 (create_forwarder_block): New function extracted from...
5008 (fix_up_crossing_landing_pad): ...here. Rename into...
5009 (dw2_fix_up_crossing_landing_pad): ...this.
5010 (sjlj_fix_up_crossing_landing_pad): New function.
5011 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5012 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5013 from both partitions and exit the loop after one iteration.
5015 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5018 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5020 (Basic PowerPC Built-in Functions): The new name of the
5021 subsection previously known as "PowerPC Built-in Functions".
5022 (Basic PowerPC Built-in Functions Available on all Configurations):
5024 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5026 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5028 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5030 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5033 2018-05-08 Jim Wilson <jimw@sifive.com>
5035 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5036 (LD_EMUL_SUFFIX): New.
5037 (LINK_SPEC): Use it.
5039 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5041 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5043 (Basic PowerPC Built-in Functions): The new name of the
5044 subsection previously known as "PowerPC Built-in Functions".
5045 (Basic PowerPC Built-in Functions Available on all Configurations):
5047 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5049 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5051 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5053 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5056 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5059 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5060 after cmpelim optimization.
5062 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5064 * config.gcc: Support "goldmont".
5065 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5066 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5068 * config/i386/i386.c (m_GOLDMONT): Define.
5069 (processor_target_table): Add "goldmont".
5070 (PTA_GOLDMONT): Define.
5071 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5072 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5073 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5074 (fold_builtin_cpu): Add "goldmont".
5075 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5076 (ix86_option_override_internal): Add "goldmont".
5077 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5078 (processor_type): Add PROCESSOR_GOLDMONT.
5079 * config/i386/i386.md: Add CPU "glm".
5080 * config/i386/glm.md: New file.
5081 * config/i386/x86-tune.def: Add m_GOLDMONT.
5082 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5084 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5087 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5089 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5090 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5091 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5094 * config/i386/i386.c (ix86_fold_builtin): Handle
5095 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5098 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5099 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5101 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5104 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5105 (check_arch): Likewise.
5106 (check_fpu): Return the result rather than printing it.
5107 (end arch): Fix operator precedence.
5108 (end cpu): Likewise.
5109 (END): Print the result from check_fpu.
5111 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5112 Alan Hayward <alan.hayward@arm.com>
5113 David Sherwood <david.sherwood@arm.com>
5115 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5116 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5117 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5118 (*fcmuo<mode>_and): New patterns.
5120 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5122 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5123 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5124 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5125 (cmp_op, sve_imm_con): New code attributes.
5126 (SVE_COND_INT_CMP, imm_con): Delete.
5127 (cmp_op): Remove above unspecs from int attribute.
5128 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5130 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5131 comparison-specific unspecs.
5132 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5133 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5134 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5135 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5136 (*vec_fcm<cmp_op><mode>): Rename to...
5137 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5138 (*vec_fcmuo<mode>): Rename to...
5139 (*fcmuo<mode>): ...this and adjust likewise.
5140 (*pred_fcm<cmp_op><mode>): New pattern.
5141 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5142 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5144 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5146 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5147 (aarch64_emit_sve_predicated_cond): New function.
5148 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5149 (aarch64_emit_unspec_cond_or): Replace with...
5150 (aarch64_emit_sve_or_conds): ...this new function. Use
5151 aarch64_emit_sve_ptrue_op for the individual comparisons and
5152 aarch64_emit_binop to OR them together.
5153 (aarch64_emit_inverted_unspec_cond): Replace with...
5154 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5155 aarch64_emit_sve_ptrue_op for the comparison and
5156 aarch64_emit_unop to invert the result.
5157 (aarch64_expand_sve_vec_cmp_float): Update after the above
5158 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5160 2018-05-07 Nathan Sidwell <nathan@acm.org>
5162 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5163 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5164 (Backwards Compatibility): Likewise.
5166 2018-05-07 Luis Machado <luis.machado@linaro.org>
5170 2018-05-07 Luis Machado <luis.machado@linaro.org>
5172 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5173 <prefetch_dynamic_strides>: New const bool field.
5174 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5175 prefetch_dynamic_strides.
5176 (exynosm1_prefetch_tune): Likewise.
5177 (thunderxt88_prefetch_tune): Likewise.
5178 (thunderx_prefetch_tune): Likewise.
5179 (thunderx2t99_prefetch_tune): Likewise.
5180 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5182 (aarch64_override_options_internal): Update to set
5183 PARAM_PREFETCH_DYNAMIC_STRIDES.
5184 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5185 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5186 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5187 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5188 prefetch-dynamic-strides setting.
5190 2018-05-07 Luis Machado <luis.machado@linaro.org>
5192 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5193 <minimum_stride>: New const int field.
5194 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5195 minimum_stride field.
5196 (exynosm1_prefetch_tune): Likewise.
5197 (thunderxt88_prefetch_tune): Likewise.
5198 (thunderx_prefetch_tune): Likewise.
5199 (thunderx2t99_prefetch_tune): Likewise.
5200 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5201 (aarch64_override_options_internal): Update to set
5202 PARAM_PREFETCH_MINIMUM_STRIDE.
5203 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5204 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5205 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5206 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5207 stride is constant and is below the minimum stride threshold.
5209 2018-05-07 Luis Machado <luis.machado@linaro.org>
5211 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5214 2018-05-07 Luis Machado <luis.machado@linaro.org>
5216 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5217 <prefetch_dynamic_strides>: New const bool field.
5218 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5219 prefetch_dynamic_strides.
5220 (exynosm1_prefetch_tune): Likewise.
5221 (thunderxt88_prefetch_tune): Likewise.
5222 (thunderx_prefetch_tune): Likewise.
5223 (thunderx2t99_prefetch_tune): Likewise.
5224 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5226 (aarch64_override_options_internal): Update to set
5227 PARAM_PREFETCH_DYNAMIC_STRIDES.
5228 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5229 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5230 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5231 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5232 prefetch-dynamic-strides setting.
5234 2018-05-07 Luis Machado <luis.machado@linaro.org>
5236 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5237 <minimum_stride>: New const int field.
5238 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5239 minimum_stride field.
5240 (exynosm1_prefetch_tune): Likewise.
5241 (thunderxt88_prefetch_tune): Likewise.
5242 (thunderx_prefetch_tune): Likewise.
5243 (thunderx2t99_prefetch_tune): Likewise.
5244 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5245 (aarch64_override_options_internal): Update to set
5246 PARAM_PREFETCH_MINIMUM_STRIDE.
5247 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5248 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5249 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5250 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5251 stride is constant and is below the minimum stride threshold.
5253 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5256 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5257 the type is addressable. Don't force op into register if it has
5260 2018-05-05 Roland McGrath <mcgrathr@google.com>
5263 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5264 any section for which we don't know a specific type it should have,
5265 regardless of name. Previously this was done only for the exact
5266 names ".init_array", ".fini_array", and ".preinit_array".
5267 (default_elf_asm_named_section): Add comment about
5268 relationship with default_section_type_flags and SECTION_NOTYPE.
5269 (get_section): Don't consider it a type conflict if one side has
5270 SECTION_NOTYPE and the other doesn't, as long as neither has the
5271 SECTION_BSS et al used in the default_section_type_flags logic.
5273 2018-05-05 Tom de Vries <tom@codesourcery.com>
5276 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5277 (workaround_barsyncs): New function.
5278 (nvptx_reorg): Use workaround_barsyncs.
5279 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5280 (define_expand "nvptx_membar_cta"): New define_expand.
5281 (define_insn "*nvptx_membar_cta"): New insn.
5283 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5285 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5286 To improve optimization opportunities.
5287 * builtin-types.def: The new needed builtin types for the above.
5289 2018-05-04 Richard Biener <rguenther@suse.de>
5291 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5292 * gimple-ssa-store-merging.c
5293 (imm_store_chain_info::output_merged_store): Remove redundant create,
5294 release split_store vector contents on failure.
5295 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5296 scalar stmt vector on cache hit.
5298 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5300 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5302 * config.gcc (powerpc-xilinx-eabi*): Remove.
5303 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5305 (fusion_addis_mem_combo_load): Ditto.
5306 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5308 (rs6000_cpu_cpp_builtins): Ditto.
5309 * config/rs6000/rs6000-linux.c
5310 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5311 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5312 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5314 (rs6000_setup_reg_addr_masks): Ditto.
5315 (rs6000_init_hard_regno_mode_ok): Ditto.
5316 (rs6000_option_override_internal): Ditto.
5317 (legitimate_lo_sum_address_p): Ditto.
5318 (rs6000_legitimize_address): Ditto.
5319 (rs6000_legitimize_reload_address): Ditto.
5320 (rs6000_legitimate_address_p): Ditto.
5321 (abi_v4_pass_in_fpr): Ditto.
5322 (setup_incoming_varargs): Ditto.
5323 (rs6000_gimplify_va_arg): Ditto.
5324 (rs6000_split_multireg_move): Ditto.
5325 (rs6000_savres_strategy): Ditto.
5326 (rs6000_emit_prologue_components): Ditto.
5327 (rs6000_emit_epilogue_components): Ditto.
5328 (rs6000_emit_prologue): Ditto.
5329 (rs6000_emit_epilogue): Ditto.
5330 (rs6000_elf_file_end): Ditto.
5331 (rs6000_function_value): Ditto.
5332 (rs6000_libcall_value): Ditto.
5333 * config/rs6000/rs6000.h: Ditto.
5334 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5335 (TARGET_MINMAX): ... this. New.
5336 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5337 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5338 (*movsi_internal1_single): Delete.
5339 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5340 mfpu=, mxilinx-fpu): Delete.
5341 * config/rs6000/singlefp.h: Delete.
5342 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5343 * config/rs6000/t-rs6000: Ditto.
5344 * config/rs6000/t-xilinx: Delete.
5345 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5346 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5347 (VStype_simple): Delete.
5348 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5349 * config/rs6000/xfpu.h: Delete.
5350 * config/rs6000/xfpu.md: Delete.
5351 * config/rs6000/xilinx.h: Delete.
5352 * config/rs6000/xilinx.opt: Delete.
5353 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5354 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5356 2018-05-04 Tom de Vries <tom@codesourcery.com>
5359 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5362 2018-05-04 Richard Biener <rguenther@suse.de>
5365 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5366 (expand_complex_div_wide): Likewise.
5367 (expand_complex_operations_1): Likewise.
5368 (expand_complex_libcall): Preserve EH info of the original stmt.
5369 (tree_lower_complex): Handle removed blocks.
5370 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5371 on complex multiplication and division libcall builtins.
5373 2018-05-04 Richard Biener <rguenther@suse.de>
5376 * fold-const.c (negate_expr_p): Restrict negation of operand
5377 zero of a division to when we know that can happen without
5379 (fold_negate_expr_1): Likewise.
5381 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5384 * real.h (real_nextafter): Declare.
5385 * real.c (real_nextafter): New function.
5386 * fold-const-call.c (fold_const_nextafter): New function.
5387 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5388 CASE_CFN_NEXTTOWARD.
5389 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5390 even when arg1_mode is different from arg0_mode.
5392 2018-05-03 Nathan Sidwell <nathan@acm.org>
5394 * doc/extend.texi (Deprecated Features): Remove
5396 (Backwards Compatibility): Likewise.
5397 * doc/invoke.texi (C++ Language Options): Likewise.
5398 (C++ Dialect Options): Likewise.
5400 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5403 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5404 _mm512_mask_mullox_epi64): New intrinsics.
5406 2018-05-03 Tom de Vries <tom@codesourcery.com>
5409 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5410 dump files): Add offload-tree.
5412 2018-05-03 Richard Biener <rguenther@suse.de>
5414 PR tree-optimization/85615
5415 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5416 to loops not nested in BBs loop father to avoid creating multi-entry
5419 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5421 PR tree-optimization/70291
5422 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5423 arguments. Change return type to tree. Emit libcall as a new
5424 statement rather than replacing existing one when inplace_p is true.
5425 (expand_complex_multiplication_components): New function.
5426 (expand_complex_multiplication): Expand floating-point complex
5427 multiplication using the above.
5428 (expand_complex_division): Rename inner_type parameter to type.
5429 Update expand_complex_libcall call-site.
5430 (expand_complex_operations_1): Update expand_complex_multiplication
5431 and expand_complex_division call-sites.
5433 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5436 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5437 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5438 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5439 the highest significant bit of the shift count mask is clear. In
5440 check whether and[sq]i3 is needed verify that all significant bits
5441 of the shift count other than the highest are set.
5443 2018-05-02 Tom de Vries <tom@codesourcery.com>
5446 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5447 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5448 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5449 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5450 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5451 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5452 __builtin_goacc_parlevel_size.
5454 2018-05-02 Richard Biener <rguenther@suse.de>
5456 PR tree-optimization/85597
5457 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5458 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5461 2018-05-02 Tom de Vries <tom@codesourcery.com>
5464 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5465 dump files): Add ltrans-tree.
5467 2018-05-02 Tom de Vries <tom@codesourcery.com>
5470 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5471 dump files): Add wpa-ipa.
5473 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5475 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5476 powerpc*-*-linux*paired* target.
5477 * config/rs6000/750cl.h: Delete.
5478 * config/rs6000/paired.h: Delete.
5479 * config/rs6000/paired.md: Delete.
5480 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5482 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5483 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5484 comment. Remove paired float support.
5485 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5486 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5488 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5489 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5491 * config/rs6000/rs6000.c: Remove paired float support.
5492 (paired_expand_vector_init, paired_expand_vector_move,
5493 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5494 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5495 paired_expand_builtin, paired_expand_predicate_builtin,
5496 paired_init_builtins): Delete.
5497 * config/rs6000/rs6000.h: Remove paired float support.
5498 * config/rs6000/rs6000.md: Remove paired float support.
5499 (move_from_CR_ov_bit): Delete.
5500 * config/rs6000/rs6000.opt (mpaired): Delete.
5501 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5502 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5504 2018-05-02 Richard Biener <rguenther@suse.de>
5507 * gimplify.c (gimplify_save_expr): When in SSA form allow
5508 SAVE_EXPRs to compute to SSA vars.
5510 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5513 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5514 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5515 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5516 clobber operands[2], instead use a new pseudo. Formatting fixes.
5518 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5520 PR tree-optimization/85586
5521 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5522 exit early for statements in the same group if the accesses are
5525 2018-05-02 Tom de Vries <tom@codesourcery.com>
5528 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5531 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5533 PR tree-optimization/85143
5534 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5536 2018-05-01 Tom de Vries <tom@codesourcery.com>
5539 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5540 not found" error message.
5542 2018-05-01 Tom de Vries <tom@codesourcery.com>
5545 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5546 * vec.c (test_ordered_remove_if): New function.
5547 (vec_c_tests): Call test_ordered_remove_if.
5548 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5549 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5550 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5551 VEC_ORDERED_REMOVE_IF.
5553 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5555 PR tree-optimization/82665
5556 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5557 pointer subtraction where arguments come from a memchr call.
5559 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5561 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5562 --push-state --as-needed and --pop-state instead of --as-needed and
5563 --no-as-needed if ld supports it.
5564 * configure: Regenerated.
5567 * doc/install.texi2html: Replace _002d with - and _002a with * in
5568 generated html files using sed.
5570 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5573 * gcc-rich-location.c (blank_line_before_p): New function.
5574 (use_new_line): New function.
5575 (gcc_rich_location::add_fixit_insert_formatted): New function.
5576 * gcc-rich-location.h
5577 (gcc_rich_location::add_fixit_insert_formatted): New function.
5579 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5581 * selftest.c (assert_streq): Rename "expected" and "actual" to
5582 "val1" and "val2". Extend NULL-handling to cover both inputs
5583 symmetrically, while still requiring both to be non-NULL for a pass.
5584 * selftest.h (assert_streq): Rename "expected" and "actual" to
5586 (ASSERT_EQ): Likewise.
5587 (ASSERT_EQ_AT): Likewise.
5588 (ASSERT_KNOWN_EQ): Likewise.
5589 (ASSERT_KNOWN_EQ_AT): Likewise.
5590 (ASSERT_NE): Likewise.
5591 (ASSERT_MAYBE_NE): Likewise.
5592 (ASSERT_MAYBE_NE_AT): Likewise.
5593 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5594 the assertion to pass.
5595 (ASSERT_STREQ_AT): Likewise.
5597 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5599 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5600 interaction with -pie.
5602 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5604 * selftest.h: Fix alphabetization of per-source-file selftest
5607 2018-04-30 Jason Merrill <jason@redhat.com>
5609 PR c++/61982 - dead stores to destroyed objects.
5610 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5613 2018-04-30 Jason Merrill <jason@redhat.com>
5615 * tree.c (build_clobber): New.
5616 * tree.h: Declare it.
5617 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5619 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5621 * diagnostic-show-locus.c (layout::layout): Update for
5622 location_get_source_line returning a char_span.
5623 (struct char_span): Move to input.h.
5624 (struct correction): Update for fields in char_span becoming
5626 (struct source_line): Update for location_get_source_line
5627 returning a char_span.
5628 (layout::print_line): Likewise.
5629 * edit-context.c (edited_file::print_content): Likewise.
5630 (edited_file::print_diff_hunk): Likewise.
5631 (edited_file::print_run_of_changed_lines): Likewise.
5632 (edited_file::get_num_lines): Likewise.
5633 (edited_line::edited_line): Likewise.
5634 * final.c (asm_show_source): Likewise.
5635 * input.c (location_get_source_line): Convert return type
5636 from const char * to char_span, losing the final "line_len"
5638 (dump_location_info): Update for the above.
5639 (get_substring_ranges_for_loc): Likewise. Use a char_span
5640 when handling the literal within the line.
5641 (test_reading_source_line): Update for location_get_source_line
5642 returning a char_span.
5643 * input.h (class char_span): Move here from
5644 diagnostic-show-locus.c, converting from a struct to a class.
5645 Make data members private.
5646 (char_span::operator bool): New.
5647 (char_span::length): New.
5648 (char_span::get_buffer): New.
5649 (char_span::operator[]): New.
5650 (char_span::subspan): Make const.
5651 (char_span::xstrdup): New.
5652 (location_get_source_line): Convert return type from const char *
5653 to char_span, losing the final "line_size" param.
5655 2018-04-30 Jan Hubicka <jh@suse.cz>
5657 * lto-wrapper.c (ltrans_priorities): New static var.
5658 (cmp_priority): New.
5659 (run_gcc): Read priorities and if doing parallel build order
5660 the Makefile by them.
5662 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5664 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5666 2018-04-30 Richard Biener <rguenther@suse.de>
5668 * tree-cfg.c (verify_address): Remove base argument, add
5669 flag whether to check TREE_ADDRESSABLE and do that.
5670 (verify_expr): Remove.
5671 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5672 (verify_gimple_assign_single): Likewise.
5673 (verify_gimple_switch): Likewise.
5674 (verify_expr_location_1): Dereference tp once. Add (disabled)
5675 piece from verify_expr.
5676 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5678 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5680 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5682 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5684 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5685 (small_data_pattern): Likewise.
5686 (arc_rewrite_small_data): Likewise.
5687 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5688 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5689 (get_symbol_alignment): New function.
5690 (legitimate_small_data_address_p): Likewise.
5691 (legitimate_scaled_address): Update, call
5692 legitimate_small_data_address_p.
5693 (output_sdata): New static variable.
5694 (arc_print_operand): Update how we handle small data operands.
5695 (arc_print_operand_address): Likewise.
5696 (arc_legitimate_address_p): Update, use
5697 legitimate_small_data_address_p.
5698 (arc_rewrite_small_data_p): Remove.
5699 (arc_rewrite_small_data_1): Likewise.
5700 (arc_rewrite_small_data): Likewise.
5701 (small_data_pattern): Likewise.
5702 (compact_sda_memory_operand): Update to use
5703 legitimate_small_data_address_p and get_symbol_alignment.
5704 (prepare_move_operands): Don't rewite sdata pattern.
5705 (prepare_extend_operands): Remove.
5706 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5708 (zero_extendqisi2): Likewise.
5709 (zero_extendhisi2): Likewise.
5710 (extendqihi2): Likewise.
5711 (extendqisi2): Likewise.
5712 (extendhisi2): Likewise.
5716 * config/arc/constraints.md (Usd): Change it to memory constraint.
5718 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5720 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5721 as source of std instructions.
5722 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5723 allow 6-bit constants as source for store instructions.
5724 (movdi_insn): Update instruction pattern to allow 6-bit constants
5725 as source for store instructions.
5727 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5729 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5731 2018-04-30 Nathan Sidwell <nathan@acm.org>
5732 Sandra Loosemore <sandra@codesourcery.com>
5734 * dumpfile.c (dump_open): Allow '-' for stdout.
5735 * doc/invoke.texi (Developer Options): Document dump filename
5736 determination early. Document stdin/stdout selection.
5738 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5740 Microblaze Target: PIC data text relative
5742 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5743 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5745 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5746 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5747 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5748 New addressing mode for data-text relative position indepenedent code.
5749 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5750 'ADDRESS_SYMBOLIC_TXT_REL'.
5751 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5752 (microblaze_legitimate_pic_operand): Exclude function calls from
5753 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5754 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5756 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5757 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5758 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5759 for 'address + offset'.
5760 (microblaze_expand_prologue): Add new function prologue call for
5762 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5763 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5764 table in case of TARGET_PIC_DATA_TEXT_REL.
5765 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5766 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5767 Add new macros 'UNSPEC_TEXT',
5768 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5769 + exclude function calls from 'UNSPEC_PLT' in case of data text
5771 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5772 new target hook for generating address diff vector tables in case of
5774 * doc/tm.texi : Regenerate.
5775 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5776 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5777 of addr diff vector generation.
5778 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5779 target hook definition.
5780 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5781 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5782 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5783 option for data text relative.
5785 2018-04-30 Richard Biener <rguenther@suse.de>
5787 * tree-chrec.h (evolution_function_is_constant_p): Remove
5789 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5791 2018-04-30 Richard Biener <rguenther@suse.de>
5794 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5796 2018-04-30 Richard Biener <rguenther@suse.de>
5798 PR tree-optimization/28364
5799 PR tree-optimization/85275
5800 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5801 copying first exit test.
5803 2018-04-28 Mark Wielaard <mark@klomp.org>
5805 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5807 (dwarf_AT): Handle DW_AT_addr_base.
5808 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5810 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5813 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5814 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5815 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5816 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5818 2018-04-28 Richard Biener <rguenther@suse.de>
5820 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5821 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5822 to reflect use. Only add interesting stmts.
5824 2018-04-27 Martin Jambor <mjambor@suse.cz>
5827 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5828 the jump function allows for passing through aggregate values.
5830 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5832 * input.h (in_system_header_at): Convert from macro to inline
5834 (from_macro_expansion_at): Likewise.
5835 (from_macro_definition_at): Likewise.
5837 2018-04-27 Jeff Law <law@redhat.com>
5839 * config.gcc: Mark tile* targets as deprecated/obsolete.
5841 2018-04-27 Richard Biener <rguenther@suse.de>
5843 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5846 2018-04-27 Richard Biener <rguenther@suse.de>
5848 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5850 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5852 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5853 with Yd constraint. Set "preferred_for_speed" attribute from
5854 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5856 (*movdi_internal): Ditto.
5857 (movti_interunit splitters): Remove
5858 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5859 (movdi_interunit splitters): Ditto.
5860 * config/i386/constraints.md (Ye): Remove.
5861 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5863 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5866 * config/aarch64/constraints.md (Usg): Limit to 31.
5869 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5871 PR tree-optimization/85529
5872 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5873 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5874 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5875 zero extension or masking of the MSB bit.
5876 (optimize_range_tests): Add FIRST_BB argument, pass it through
5877 to optimize_range_tests_var_bound.
5878 (maybe_optimize_range_tests, reassociate_bb): Adjust
5879 optimize_range_tests callers.
5881 2018-04-26 Richard Biener <rguenther@suse.de>
5882 Jakub Jelinek <jakub@redhat.com>
5884 * cgraph.h (symbol_table): Just declare debug method here.
5885 * symtab.c (symbol_table::debug): Define.
5887 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5889 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5891 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5893 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5894 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5895 (*movdi_internal): Substitute Yi and Yj constraint with x
5896 and Ym and Yn constraint with y constraint. Update "isa"
5897 attribute and set "preferred_for_speed" attribute from
5898 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5899 (*movsi_internal): Ditto.
5900 (*movdf_internal): Ditto.
5901 (*movsf_internal): Ditto.
5902 (*zero_extendsidi2): Ditto.
5903 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5904 (sse2_loadld): Ditto.
5905 (*vec_extract<ssevecmodelower>_0): Ditto.
5906 (*vec_extractv4si_0_zext_sse4): Ditto.
5907 (vec_concatv2di): Ditto.
5908 (*vec_dup<mode>): Ditto.
5909 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5910 * config/i386/constraints.md (Yi): Remove.
5915 2018-04-26 Nathan Sidwell <nathan@acm.org>
5917 * dumpfile.c (dump_open): New.
5918 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5919 (dump_finish): Detect stdio/stderr by value not name.
5921 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5923 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5925 2018-04-26 Tom de Vries <tom@codesourcery.com>
5928 * config/nvptx/nvptx.c (verify_neutering_jumps)
5929 (verify_neutering_labels): New function
5930 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5932 2018-04-26 Tom de Vries <tom@codesourcery.com>
5935 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5936 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5939 2018-04-26 Richard Biener <rguenther@suse.de>
5940 Tom de Vries <tom@codesourcery.com>
5943 * lto-streamer-out.c (output_function): Fixup loops if required to match
5944 discovery done in the reader.
5946 2018-04-26 Richard Biener <rguenther@suse.de>
5948 PR tree-optimization/85116
5949 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5950 have a loop exit from the single latch predecessor. Remove
5951 case of header with just condition.
5952 (ch_base::copy_headers): Exclude infinite loops from any
5954 (pass_ch::execute): Record exits.
5956 2018-04-26 Richard Biener <rguenther@suse.de>
5958 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5959 prologue cost vector and pass it to vect_get_load_cost.
5960 (vect_get_peeling_costs_all_drs): Likewise.
5961 (vect_peeling_hash_get_lowest_cost): Likewise.
5962 (vect_enhance_data_refs_alignment): Likewise.
5964 2018-04-26 Richard Biener <rguenther@suse.de>
5967 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5968 checking of integer<->pointer conversions.
5969 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5970 sign-/zero-extending pointer types.
5971 (expand_omp_for_static_chunk): Likewise.
5973 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5974 Jean Lee <xiaoyur347@gmail.com>
5976 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5977 (TARGET_ASAN_SHADOW_OFFSET): Define.
5978 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5979 true for -fsanitize=address.
5981 2018-04-25 Mark Wielaard <mark@klomp.org>
5983 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5986 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5988 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5989 than "alu", remove explicit "memory" and "imm_disp" attributes.
5990 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5993 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5994 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5995 gen_lowpart_no_emit.
5997 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6000 * config/i386/i386.c (ix86_expand_builtin): Change memory
6001 operand to XI, extend p0 to Pmode.
6002 * config/i386/i386.md: Change unspec volatile and operand
6003 1 mode to XI, change operand 0 mode to P.
6005 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6007 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6008 GET_MODE_MASK before any checking.
6009 (nds32_can_use_bset_p): Likewise.
6010 (nds32_can_use_btgl_p): Likewise.
6012 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6014 * config/nds32/nds32-doubleword.md: New define_split pattern for
6015 illegal register number.
6017 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6019 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6021 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6023 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6025 2018-04-25 Richard Biener <rguenther@suse.de>
6027 * lto-streamer.h (LTO_major_version): Bump to 8.
6029 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6031 * BASE-VER: Set to 9.0.0.
6033 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6035 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6036 in __abskf2 and __powikf2.
6038 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6041 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6042 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6043 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6044 Use the above on operand 2. Reindent.
6045 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6047 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6050 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6052 * config/i386/i386.opt (mcet): Removed.
6053 * doc/install.texi: Remove -mcet documentation.
6054 * doc/invoke.texi: Likewise.
6056 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6059 * doc/install.texi: Remove -mcet from bootstrap-cet.
6061 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6064 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6065 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6069 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6070 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6071 containing a CONST_VECTOR.
6073 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6075 * doc/install.texi: Update newlib dependency for nvptx.
6077 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6080 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6081 instead of INTVAL when shifting x left.
6083 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6085 PR tree-optimization/85478
6086 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6087 vect_grouped_store_supported for single element vectors.
6089 2018-04-24 Richard Biener <rguenther@suse.de>
6092 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6093 load cost increase to the case of non-constant step.
6095 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6098 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6099 destination if any_malformed_asm.
6101 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6104 * expr.c (store_field): In the bitfield case, if the value comes from
6105 a function call and is returned in registers by means of a PARALLEL,
6106 do not change the mode of the temporary unless BLKmode and VOIDmode.
6108 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6110 PR rtl-optimization/85423
6111 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6112 dependencies to debug insns when the previous insn is non-debug.
6114 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6116 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6117 enums into a single definition.
6118 (fls): Fix predicates and printing.
6121 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6123 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6124 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6125 and short u6 immediate.
6126 (check_if_valid_sleep_operand): Remove.
6127 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6129 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6131 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6132 flag_always_save_lp condition.
6133 * config/nds32/nds32.opt (malways-save-lp): New option.
6135 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6137 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6138 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6139 * config/nds32/nds32.h
6140 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6141 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6143 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6145 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6146 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6148 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6149 Chung-Ju Wu <jasonwucj@gmail.com>
6151 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6153 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6154 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6156 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6158 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6160 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6162 * config/nds32/nds32-protos.h (nds32_data_alignment,
6163 nds32_local_alignment): Declare.
6164 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6165 nds32_local_alignment): New functions.
6166 (TARGET_CONSTANT_ALIGNMENT): Define.
6167 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6169 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6171 * config/nds32/nds32.c
6172 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6173 (TARGET_MODES_TIEABLE_P): Likewise.
6175 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6177 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6180 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6181 Chung-Ju Wu <jasonwucj@gmail.com>
6183 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6184 for unaligned access.
6185 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6187 * config/nds32/nds32-intrinsic.md: Likewise.
6188 * config/nds32/nds32_intrinsic.h: Likewise.
6189 * config/nds32/nds32.h (nds32_builtins): Likewise.
6190 * config/nds32/nds32.opt (munaligned-access): New option.
6191 * config/nds32/nds32.c (nds32_asm_file_start): Display
6192 flag_unaligned_access status.
6194 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6196 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6197 -mno-relax is present.
6198 * config/riscv/linux.h (LINK_SPEC): Ditto.
6200 2018-04-20 Martin Sebor <msebor@redhat.com>
6203 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6205 (gimple_fold_builtin_stxcpy_chk): Same.
6206 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6208 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6211 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6212 __powikf2 when long double is IEEE 128-bit.
6214 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6216 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6217 step to make sure stack always aligned.
6219 2018-04-20 Carl Love <cel@us.ibm.com>
6222 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6223 size check for arg0.
6225 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6226 Tom de Vries <tom@codesourcery.com>
6229 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6230 Emit insns for calls too.
6231 (nvptx_find_par): Always look for worker-level predecessor insn.
6232 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6234 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6235 (nvptx_process_pars): Propagate frames for calls.
6237 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6240 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6242 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6243 (ix86_handle_option): Don't handle OPT_mibt.
6244 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6246 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6248 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6249 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6250 (ix86_target_macros): Define __CET__ with flag_cf_protection
6251 for -fcf-protection.
6252 * config/i386/i386.c (isa2_opts): Remove -mibt.
6253 * config/i386/i386.h (TARGET_IBT): Removed.
6254 (TARGET_IBT_P): Likewise.
6255 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6256 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6257 * config/i386/i386.opt (mcet): Update help message.
6260 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6261 -mcet as an alias for -mshstk.
6263 2018-04-20 Richard Biener <rguenther@suse.de>
6266 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6267 complexity by forcing a single use of the multiply operand.
6269 2018-04-20 Martin Jambor <mjambor@suse.cz>
6272 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6273 recursion dependency to only apply to non-clones.
6275 2018-04-20 Martin Jambor <mjambor@suse.cz>
6278 * ipa-cp.c (create_specialized_node): Check that clones of
6279 self-recursive edges exist during IPA-CP.
6281 2018-04-19 Toon Moene <toon@moene.org>
6283 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6286 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6288 PR tree-optimization/85467
6289 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6290 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6291 VECTOR_CST element to type.
6293 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6296 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6297 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6298 (builtin_longjmp): Likewise.
6299 (save_stack_nonlocal): New pattern.
6300 (restore_stack_nonlocal): Likewise.
6302 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6305 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6306 Replace ASM_OUTPUT_LABEL with fprintf.
6308 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6311 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6312 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6313 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6314 define __IBT__ and __SHSTK__ for -fcf-protection.
6315 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6317 (ix86_trampoline_init): Likewise.
6318 (x86_output_mi_thunk): Likewise.
6319 (ix86_notrack_prefixed_insn_p): Likewise.
6320 (ix86_option_override_internal): Don't disallow -fcf-protection.
6321 * config/i386/i386.md (rdssp<mode>): Also enable for
6323 (incssp<mode>): Likewise.
6324 (nop_endbr): Likewise.
6325 * config/i386/i386.opt (mcet): Change help message to built-in
6329 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6330 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6331 enable CET built-in functions.
6333 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6335 * common/config/i386/i386-common.c
6336 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6337 OPTION_MASK_ISA_MOVDIRI_UNSET,
6338 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6339 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6340 * config.gcc (movdirintrin.h): New header.
6341 * config/i386/cpuid.h (bit_MOVDIRI,
6342 bit_MOVDIR64B): New bits.
6343 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6345 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6346 (VOID, PVOID, PCVOID)): New function types.
6347 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6348 __builtin_ia32_directstoreu_u64,
6349 __builtin_ia32_movdir64b): New builtins.
6350 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6351 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6353 (ix86_valid_target_attribute_inner_p): Ditto.
6354 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6355 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6356 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6357 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6358 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6359 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6360 (movdiri<mode>, movdir64b_<mode>): New.
6361 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6362 * config/i386/immintrin.h: Include movdirintrin.h.
6363 * config/i386/movdirintrin.h: New file.
6364 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6366 2018-04-19 Richard Biener <rguenther@suse.de>
6369 * cfg.c (clear_bb_flags): When loop state says we have
6370 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6372 2018-04-19 Richard Biener <rguenther@suse.de>
6374 PR tree-optimization/84737
6375 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6376 copying restrict info.
6377 (vect_setup_realignment): Use it.
6378 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6379 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6380 the first DR to all generated stores.
6381 (vectorizable_load): Likewise for loads.
6383 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6385 PR tree-optimization/85446
6386 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6387 the integral and pointer types to have the same precision.
6389 * doc/install.texi: Document --disable-cet being the default and
6392 2018-04-18 Martin Liska <mliska@suse.cz>
6394 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6397 2018-04-18 Martin Liska <mliska@suse.cz>
6400 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6403 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6404 arguments if they are comparable.
6406 2018-04-18 Martin Liska <mliska@suse.cz>
6409 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6412 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6415 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6418 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6419 ENDBR after calling __morestack.
6421 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6424 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6425 by using gcc_base_ver to generate a gcc_driver_version, and use
6426 it when generating GCC_DRIVER_NAME.
6427 * configure: Regenerate.
6429 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6432 * config.gcc: Obsolete powerpc*-*-*spe*.
6434 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6437 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6438 (stabstr_D): Change type of unum from unsigned int to
6439 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6442 2018-04-17 Jim Wilson <jimw@sifive.com>
6445 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6446 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6447 Set arg_pointer_offset after using pretend_args_size.
6449 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6451 PR rtl-optimization/85431
6452 * dse.c (record_store): Ignore zero width stores.
6455 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6456 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6457 __builtin_stack_restore rather than after it.
6458 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6459 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6460 argument instead of virtual_dynamic_stack_rtx.
6462 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6464 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6466 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6467 Add note to error message to explain internal mapping of overloaded
6468 built-in function name to non-overloaded built-in function name.
6469 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6472 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6475 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6476 where the inputs overlap with the output.
6478 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6480 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6481 (=v, v) alternative and explicit "memory" attribute.
6482 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6483 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6485 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6486 "sselog1" type instead of "sselog".
6487 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6488 "sselog". Remove explicit "memory" attribute.
6489 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6490 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6492 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6493 "sselog1" type instead of "sselog". Remove explicit "memory"
6495 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6496 use "sselog1" type instead of "sselog". Remove explicit "memory"
6498 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6499 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6501 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6502 "sselog1" type instead of "sselog". Remove explicit "memory"
6504 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6505 use "sselog1" type instead of "sselog". Remove explicit "memory"
6509 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6512 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6515 2018-04-17 Martin Jambor <mjambor@suse.cz>
6518 * ipa-cp.c (create_specialized_node): Call
6519 expand_all_artificial_thunks if necessary.
6521 2018-04-17 Martin Liska <mliska@suse.cz>
6524 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6525 in message, remote space in between '_G' and '('.
6527 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6530 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6531 avx512f_vmcmp<mode>3<round_saeonly_name>,
6532 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6533 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6534 avx512f_rndscale<mode><round_saeonly_name>,
6535 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6536 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6537 Use %<iptr>2 instead of %2 for -masm=intel.
6538 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6539 avx512f_vcvttss2usi<round_saeonly_name>,
6540 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6542 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6543 avx512f_vcvttsd2usi<round_saeonly_name>,
6544 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6545 Use %q1 instead of %1 for -masm=intel.
6546 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6547 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6548 of %3 for -masm=intel.
6549 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6551 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6553 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6555 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6556 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6558 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6560 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6561 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6562 %g1 and one with %0 and %1.
6563 (avx512er_vmrcp28<mode><round_saeonly_name>,
6564 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6566 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6567 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6568 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6569 of %0 and %{%4%} for -masm=intel.
6570 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6571 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6572 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6573 order of %0 and %{%5%}%{z%} for -masm=intel.
6575 2018-04-17 Jan Hubicka <jh@suse.cz>
6578 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6580 2018-04-17 Martin Liska <mliska@suse.cz>
6583 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6584 for target_clone error message. Make default implementation
6585 clone to be a local declaration.
6586 (separate_attrs): Add new argument and check for an empty
6588 (expand_target_clones): Handle it.
6589 (ipa_target_clone): Make redirection just for target_clones
6592 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6593 Tom de Vries <tom@codesourcery.com>
6596 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6597 tiled basic blocks without omp continue statements.
6599 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6602 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6603 vec_extract expression as having side effects to make sure it gets
6606 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6609 * config/i386/i386.c (get_builtin_code_for_version): Check
6612 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6615 * gcc/config.gcc: Support "skylake".
6616 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6618 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6619 (processor_target_table): Add "skylake".
6620 (ix86_option_override_internal): Add "skylake".
6621 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6622 PROCESSOR_CANNONLAKE.
6623 (get_builtin_code_for_version): Fix priority for
6624 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6625 PROCESSOR_SKYLAKE-AVX512.
6626 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6627 (processor_type): Add PROCESSOR_SKYLAKE.
6629 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6630 Jason Merrill <jason@redhat.com>
6633 * convert.c (convert_to_integer_1): Use direct recursion for
6634 enumeral types and types with a precision less than the number
6635 of bits in their mode.
6637 2018-04-16 Julia Koval <julia.koval@intel.com>
6640 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6641 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6643 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6646 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6647 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6648 and -mno-direct-move.
6650 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6653 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6654 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6655 Ensure negative shifts result in {0}.
6657 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6659 PR rtl-optimization/79916
6660 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6661 regs (if any) to define how to gnerate SD moves when LRA is in
6664 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6666 PR rtl-optimization/85393
6667 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6668 * except.c (expand_dw2_landing_pad_for_region): Make static.
6669 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6670 a label and unconditional jump to old_bb, rather than
6671 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6674 PR rtl-optimization/85376
6675 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6676 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6677 instead of a specific value.
6679 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6680 Bin Cheng <bin.cheng@arm.com>
6682 PR tree-optimization/82965
6683 PR tree-optimization/83991
6684 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6685 by_profile_only parameter.
6686 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6687 information if the loop was predicted to iterate too many times.
6688 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6690 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6693 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6696 2018-04-13 Martin Liska <mliska@suse.cz>
6697 Jakub Jelinek <jakub@redhat.com>
6700 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6701 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6702 * builtins.c (expand_builtin_memory_copy_args): Use
6703 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6704 handle dest_addr == pc_rtx.
6706 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6709 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6710 asked to not generate direct moves.
6711 (fix_trunc<mode>si2_stfiwx): Similar.
6712 (fix_trunc<mode>si2_internal): Similar.
6714 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6717 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6718 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6719 lookup if dest in some wider mode is known to be const0_rtx and
6720 if so, record permanent equivalence for it to be ZERO_EXTEND of
6721 the narrower mode destination.
6723 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6725 * lto-streamer-out.c (output_function): Revert 259346.
6726 * omp-expand.c (expand_oacc_for): Likewise.
6728 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6730 PR rtl-optimization/85354
6731 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6732 * sel-sched.c (sel_global_init): ... here.
6734 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6737 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6738 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6739 mode for PE-COFF targets.
6740 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6741 (i386_pe_asm_lto_end): Likewise.
6742 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6743 (TARGET_ASM_LTO_END): Likewise.
6744 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6745 (i386_pe_asm_lto_start): New function.
6746 (i386_pe_asm_lto_end): Likewise.
6748 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6749 Richard Biener <rguenther@suse.de>
6752 * lto-streamer-out.c (output_function): Fix CFG loop state before
6754 * omp-expand.c (expand_oacc_for): Handle calls to internal
6755 functions like regular functions.
6757 2018-04-12 Richard Biener <rguenther@suse.de>
6760 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6761 for the early LTO debug to properly generate references to it
6762 during DIE emission. Do not re-use that for the skeleton for
6764 (dwarf2out_early_finish): Likewise.
6766 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6769 * config/i386/sse.md
6770 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6771 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6772 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6773 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6774 and output is a reg, avoid creating invalid lowpart subreg, but
6775 instead split into a 512-bit move. Don't split if not AVX512VL,
6776 input is xmm16+ reg and output is a mem.
6777 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6778 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6779 xmm16+ reg and output is a mem.
6781 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6783 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6784 also for flag_dwarf2_cfi_asm.
6786 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6788 PR rtl-optimization/85342
6789 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6790 a bool scalar var inside of the loop instead. Don't try to update
6791 recog_data.operand after failed apply_change_group.
6793 2018-04-12 Tom de Vries <tom@codesourcery.com>
6796 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6797 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6798 array with flexible array member as array without given dimension.
6799 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6800 argument for undefined param to true.
6802 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6805 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6806 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6807 from PowerPC section.
6808 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6809 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6810 help text that is too long.
6811 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6812 help text that is too long.
6813 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6814 help text that is too long.
6816 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6818 * config/alpha/alpha.md (stack_probe_internal): Rename
6819 from "probe_stack". Update all callers.
6821 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6823 PR rtl-optimization/84566
6824 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6825 sched_macro_fuse_insns.
6827 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6830 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6831 (compute_block_dependences): ... from here.
6833 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6835 PR tree-optimization/85331
6836 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6837 from int to HOST_WIDE_INT.
6839 2018-04-11 Martin Jambor <mjambor@suse.cz>
6842 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6843 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6844 not the same as the source val.
6845 (cgraph_edge_brings_value_p): New parameter.
6846 (gather_edges_for_value): Pass destination value to
6847 cgraph_edge_brings_value_p.
6848 (perhaps_add_new_callers): Likewise.
6849 (get_info_about_necessary_edges): Likewise and exclude values brought
6850 only by self-recursive edges.
6851 (create_specialized_node): Redirect only clones of self-calling edges.
6852 (+self_recursive_pass_through_p): New function.
6853 (find_more_scalar_values_for_callers_subset): Use it.
6854 (find_aggregate_values_for_callers_subset): Likewise.
6855 (known_aggs_to_agg_replacement_list): Removed.
6856 (decide_whether_version_node): Re-calculate known constants for all
6857 remaining context clones.
6859 2018-04-11 Richard Biener <rguenther@suse.de>
6862 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6863 from early DWARF output.
6864 (dwarf2out_early_finish): Output line info unconditionally into
6865 early DWARF and add reference to it.
6867 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6870 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6871 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6872 other than V2DFmode using iptr mode attribute.
6873 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6875 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6877 PR rtl-optimization/84659
6878 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6880 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6883 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6885 (output_loc_list): Pass address of a dummy size variable even in the
6886 locview handling loop.
6887 (index_location_lists): Add comment on why skip_loc_list_entry can't
6890 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6893 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6896 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6899 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6900 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6901 and -mstring-compare-inline-limit.
6903 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6906 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6907 for stack clash protection in a register whenever we need it to be in
6910 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6912 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6913 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6915 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6918 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6920 (mlong-double-): Ditto.
6921 * config/rs6000/sysv4.opt (msdata=): Ditto.
6922 (mtls-size=): Ditto.
6924 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6926 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6927 erroneous entries for
6928 "vector int vec_ldl (int, long int *)", and
6929 "vector unsigned int vec_ldl (int, unsigned long int *)".
6930 Add comments and entries for
6931 "vector bool char vec_ldl (int, bool char *)",
6932 "vector bool short vec_ldl (int, bool short *)",
6933 "vector bool int vec_ldl (int, bool int *)",
6934 "vector bool long long vec_ldl (int, bool long long *)",
6935 "vector pixel vec_ldl (int, pixel *)",
6936 "vector long long vec_ldl (int, long long *)",
6937 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6938 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6939 type tree bool_long_long_type_node and correct definition of
6940 bool_V2DI_type_node to make reference to this new type tree.
6941 (rs6000_mangle_type): Replace erroneous reference to
6942 bool_long_type_node with bool_long_long_type_node.
6943 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6944 comments to emphasize sign distinctions for char and int types and
6945 replace RS6000_BTI_bool_long constant with
6946 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6947 use of RS6000_BTI_pixel.
6948 (bool_long_type_node): Remove this macro definition.
6949 (bool_long_long_type_node): New macro definition
6951 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6953 PR rtl-optimization/85300
6954 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6955 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6956 simplify_unary_operation fails.
6958 2018-04-10 Martin Liska <mliska@suse.cz>
6960 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6961 cgraph_edge and ipa_ref.
6963 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6967 * config/i386/sse.md
6968 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6969 computation of the VEC_MERGE selector from mask.
6970 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6971 Fix decoding of the VEC_MERGE selector into mask.
6973 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6975 PR tree-optimization/85286
6976 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6978 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6980 * final.c (final_1): Set insn_last_address as well as
6981 insn_current_address.
6983 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6986 * explow.c (emit_stack_probe): Call validize_mem on memory location
6987 before passing it to gen_probe_stack. Create address operand and
6988 legitimize it for the probe_stack_address case.
6990 2018-04-09 Jan Hubicka <jh@suse.cz>
6993 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6994 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6995 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6997 (free_lang_data): Rebuild type inheritance graph.
6999 2018-04-09 Martin Sebor <msebor@redhat.com>
7001 * invoke.texi (-finline-small-functions): Mention other optimization
7003 (-findirect-inlining, -fpartial-inlining): Same.
7004 (-finline-functions-called-once): Same.
7005 (-freorder-blocks-and-partition): Same.
7007 2018-04-09 Jan Hubicka <jh@suse.cz>
7010 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7011 jumps; choose last target that matches the criteria (i.e.
7012 no partition changes for non-crossing jumps).
7013 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7014 support for redirecting crossing jumps to non-crossing.
7016 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7018 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7019 also for naked functions.
7021 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7023 * config/arc/arc.md (add_shift): New pattern.
7024 (add_shift2): Likewise.
7025 (sub_shift): Likewise.
7026 (sub_shift_cmp0_noout): Likewise.
7027 (compare_si_ashiftsi): Likewise.
7028 (xbfu_cmp0_noout): New combine pattern.
7029 (xbfu_cmp0"): Likewise.
7030 (movsi_set_cc_insn): Place the predicable variant first.
7031 (commutative_binary_cmp0_noout): Remove clobber.
7032 (commutative_binary_cmp0): New pattern.
7033 (noncommutative_binary_cmp0): Likewise.
7034 (noncommutative_binary_cmp0_noout): Likewise.
7035 (noncommutative_binary_comparison_result_used): Removed.
7036 (rsub_cmp0): New pattern.
7037 (rsub_cmp0_noout): Likewise.
7038 (extzvsi): Changed, keep only meaningful variants.
7039 (SQH, SEZ): New iterators.
7040 (SQH_postfix): New mode attribute.
7041 (SEZ_prefix): New code attribute.
7042 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7043 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7044 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7046 (noncommutative_operator): Check the availability of barrel
7049 2018-04-09 Richard Biener <rguenther@suse.de>
7051 PR tree-optimization/85284
7052 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7053 Only use the niter constraining form of simple_iv when the exit
7056 2018-04-09 Tom de Vries <tom@codesourcery.com>
7059 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7060 (define_expand "*memory_barrier"): New define_expand.
7061 (define_insn "memory_barrier"): New insn.
7063 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7065 PR rtl-optimization/80463
7066 PR rtl-optimization/83972
7067 PR rtl-optimization/83480
7069 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7070 correct producer for the insn.
7071 (tidy_control_flow): Fixup seqnos in case of debug insns.
7073 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7075 PR rtl-optimization/83913
7077 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7078 different sched-times when merging exprs.
7080 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7082 PR rtl-optimization/83962
7084 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7085 tidy_fallthru_edge and tidy_control_flow.
7087 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7089 PR rtl-optimization/83530
7091 * sel-sched.c (force_next_insn): New global variable.
7092 (remove_insn_for_debug): When force_next_insn is true, also leave only
7093 next insn in the ready list.
7094 (sel_sched_region): When the region wasn't scheduled, make another pass
7095 over it with force_next_insn set to 1.
7097 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7099 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7101 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7102 for interrupt control.
7103 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7104 functions for interrupt control.
7105 * config/nds32/nds32-intrinsic.md: Likewise.
7106 * config/nds32/nds32_intrinsic.h: Likewise.
7107 * config/nds32/nds32.h (nds32_builtins): Likewise.
7109 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7111 * config/nds32/nds32.c (nds32_init_machine_status,
7112 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7113 strict_aligned_p field.
7114 (nds32_expand_to_rtl_hook): New function.
7115 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7116 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7118 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7119 Chung-Ju Wu <jasonwucj@gmail.com>
7121 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7122 * config/nds32/nds32-n7.md: New file.
7123 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7124 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7126 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7127 * config/nds32/nds32.md (pipeline_model): Add n7.
7128 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7129 * config/nds32/pipelines.md: Include n7 settings.
7131 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7132 Chung-Ju Wu <jasonwucj@gmail.com>
7134 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7135 * config/nds32/nds32-e8.md: New file.
7136 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7137 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7139 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7140 * config/nds32/nds32.md (pipeline_model): Add e8.
7141 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7142 * config/nds32/pipelines.md: Include e8 settings.
7144 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7145 Chung-Ju Wu <jasonwucj@gmail.com>
7147 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7148 * config/nds32/nds32-n8.md: New file.
7149 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7150 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7152 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7153 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7154 * config/nds32/nds32.md (pipeline_model): Add n8.
7155 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7156 * config/nds32/pipelines.md: Include n8 settings.
7158 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7159 Chung-Ju Wu <jasonwucj@gmail.com>
7161 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7162 * config/nds32/nds32-n9-2r1w.md: New file.
7163 * config/nds32/nds32-n9-3r2w.md: New file.
7164 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7165 nds32_register_ports): New or modify for cpu n9.
7166 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7168 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7169 * config/nds32/nds32-utils.c: New file.
7170 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7171 TARGET_MUL_SLOW): Define.
7172 * config/nds32/nds32.md (pipeline_model): New attribute.
7173 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7174 New options that support cpu n9.
7175 * config/nds32/pipelines.md: Include n9 settings.
7176 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7178 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7180 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7181 information if necessary.
7182 (output_cond_branch_compare_zero): Likewise.
7183 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7184 (nds32_target_alignment): Refine for alignment.
7185 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7186 (FUNCTION_BOUNDARY): Modify.
7187 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7189 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7191 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7193 * config/nds32/constants.md (unspec_volatile_element): Add values for
7194 TLB operation and data prefetch.
7195 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7196 functions for TLB operation and data prefetch.
7197 * config/nds32/nds32-intrinsic.md: Likewise.
7198 * config/nds32/nds32_intrinsic.h: Likewise.
7199 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7200 (nds32_print_operand): Likewise.
7201 * config/nds32/nds32.h (nds32_builtins): Likewise.
7203 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7204 Andrew Pinski <pinsika@gcc.gnu.org>
7207 * match.pd: Use constant_boolean_node of correct type instead of
7208 boolean_true_node or boolean_false_node for simplifying
7209 pointer comparisons to zero.
7211 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7213 PR tree-optimization/80021
7214 * tree.c (verify_type_variant): Make error call in verify_variant_match
7215 translatable and remove final full stop.
7217 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config/nds32/constants.md (unspec_volatile_element): Add
7220 UNSPEC_VOLATILE_EH_RETURN.
7221 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7222 nds32_output_stack_pop): Support dwarf exception handling process.
7223 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7224 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7225 exception handling process.
7226 (nds32_compute_stack_frame): Likewise.
7227 (nds32_return_addr_rtx): Likewise.
7228 (nds32_initial_elimination_offset): Likewise.
7229 (nds32_expand_prologue): Likewise.
7230 (nds32_expand_epilogue): Likewise.
7231 (nds32_dynamic_chain_address): New function.
7232 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7234 (DYNAMIC_CHAIN_ADDRESS): Define.
7235 (EH_RETURN_DATA_REGNO): Define.
7236 (EH_RETURN_STACKADJ_RTX): Define.
7237 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7238 patterns for dwarf exception handling.
7240 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7242 * config/nds32/nds32.h: Clean up obsolete macros.
7244 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7246 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7247 Add enum values for particular instructions.
7248 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7249 particular intrinsic functions.
7250 * config/nds32/nds32-intrinsic.md: Likewise.
7251 * config/nds32/nds32_intrinsic.h: Likewise.
7252 * config/nds32/nds32.h (nds32_builtins): Likewise.
7253 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7254 (btst, ave): New patterns for particular instructions.
7256 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7258 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7259 Add enum values for atomic load/store and memory sync.
7260 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7262 * config/nds32/nds32-intrinsic.md: Likewise.
7263 * config/nds32/nds32_intrinsic.h: Likewise.
7264 * config/nds32/nds32.h (nds32_builtins): Likewise.
7266 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7268 PR tree-optimization/85257
7269 * fold-const.c (native_encode_vector): If not all elts could fit
7270 and off is -1, return 0 rather than offset.
7271 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7272 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7273 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7274 adjust buffer in native_interpret_expr call.
7276 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7278 * config/nds32/constants.md (unspec_volatile_element): Add cache
7279 control enum values.
7280 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7281 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7282 * config/nds32/nds32.c (nds32_cctl_names): New.
7283 (nds32_print_operand): Handle cache control register names.
7284 * config/nds32/nds32.h (nds32_builtins): New enum values.
7285 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7287 * config/nds32/nds32.md (type): Add mmu.
7288 * config/nds32/pipelines.md (simple_insn): Add mmu.
7290 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7292 * config/nds32/nds32.md (type): Remove call.
7293 * config/nds32/pipelines.md (simple_insn): Likewise.
7295 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7297 * config/nds32/constants.md (unspec_volatile_element): Add
7298 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7299 UNSPEC_VOLATILE_FMFCFG.
7300 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7301 description for fmfcfg and fmfcsr.
7302 (bdesc_1arg): Add fmtcsr.
7303 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7304 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7305 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7306 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7307 unspec_fmfcfg): New patterns.
7308 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7309 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7310 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7311 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7312 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7313 __nds32__fmfcfg): Define.
7315 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7317 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7318 intrinsic register names.
7319 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7320 intrinsic register enum values and macros.
7322 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7324 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7325 for load/store addressing form.
7326 (nds32_print_operand_address): Likewise.
7328 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7331 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7332 based on LABEL_REF. Remove useless assertion.
7333 (pic_address_needs_scratch): Fix formatting.
7334 (sparc_legitimize_pic_address): Minor tweaks.
7335 (sparc_delegitimize_address): Adjust assertion accordingly.
7336 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7337 into symbolic_operand.
7338 (movsi_high_pic_label_ref): Likewise.
7339 (movsi_lo_sum_pic_label_ref): Likewise.
7340 (movdi_pic_label_ref): Likewise.
7341 (movdi_high_pic_label_ref): Likewise.
7342 (movdi_lo_sum_pic_label_ref): Likewise.
7344 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7346 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7347 custom LIB_SPEC setup.
7349 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7350 Kito Cheng <kito.cheng@gmail.com>
7352 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7353 * config/riscv/freebsd.h: New.
7355 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7357 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7358 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7361 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7362 Kito Cheng <kito.cheng@gmail.com>
7364 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7365 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7366 * config/nds32/nds32-protos.h (nds32_output_call,
7367 nds32_output_return): Declare.
7368 * config/nds32/nds32.md: Refine all the call and return patterns.
7370 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7373 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7374 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7376 PR rtl-optimization/84872
7377 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7378 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7381 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7383 * expr.c (copy_blkmode_to_reg): Revert 254862.
7384 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7386 2018-04-06 Richard Biener <rguenther@suse.de>
7389 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7390 after seeing a component reference with an adjacent field. Treat
7391 refs to arrays at struct end of external decls similar to
7392 refs to unconstrained commons.
7394 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7397 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7398 look through SAVE_EXPRs with non-side-effects argument. Adjust
7400 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7403 2018-04-06 Richard Biener <rguenther@suse.de>
7406 * alias.c (find_base_term): New wrapper around find_base_term
7407 unwinding CSELIB_VAL_PTR changes.
7408 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7411 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7413 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7415 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7416 constant definitions.
7417 ("nop"): lr 0,0 -> nopr r0
7418 ("nop_lr0", "nop_lr1"): New insn definitions.
7420 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7422 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7423 NDS32_V3PUSH_AVAILABLE_P macro.
7425 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7426 Chung-Ju Wu <jasonwucj@gmail.com>
7428 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7429 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7430 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7431 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7432 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7433 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7434 * config/nds32/constraints.md: New constraints and checking for hard
7435 float configuration.
7436 * config/nds32/iterators.md: New mode iterator and attribute for hard
7437 float configuration.
7438 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7440 * config/nds32/nds32-fpu.md: New file.
7441 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7442 deal with hard float code generation.
7443 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7445 (abi_type, float_reg_number): New enum type.
7446 * config/nds32/nds32-predicates.c: New predicates for hard float.
7447 * config/nds32/nds32-protos.h: Declare functions for hard float.
7448 * config/nds32/nds32.c: Implementation for hard float configuration.
7449 * config/nds32/nds32.h: Definitions for hard float configuration.
7450 * config/nds32/nds32.md: Include hard float machine description and
7451 modify patterns for hard float configuration.
7452 * config/nds32/nds32.opt: New options for hard float configuration.
7453 * config/nds32/predicates.md: New predicates for hard float
7456 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7458 * common/config/nds32/nds32-common.c
7459 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7461 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7464 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7465 CONSTRUCTOR_ELT (ctor, ...)->value.
7467 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7470 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7472 2018-04-05 Tom de Vries <tom@codesourcery.com>
7475 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7478 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7479 Kito Cheng <kito.cheng@gmail.com>
7481 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7482 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7483 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7484 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7486 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7487 Kito Cheng <kito.cheng@gmail.com>
7489 * config/nds32/constraints.md (Ufe): New memory constraint.
7490 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7491 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7492 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7494 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7495 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7497 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7499 * config/nds32/nds32.md: Use optimize_size in the condition for
7500 alu-shift instructions.
7502 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7504 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7506 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7508 * config/nds32/nds32.md (negsi2): Refine pattern.
7510 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7511 Chung-Ju Wu <jasonwucj@gmail.com>
7513 * config/nds32/iterators.md (shift_rotate): New code iterator.
7514 (shift): New code attribute.
7515 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7516 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7517 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7518 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7519 bit-wise operations.
7520 (andsi3, *andsi3): Ditto.
7521 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7522 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7523 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7524 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7525 nds32_ior_operand, nds32_xor_operand): New predicates.
7527 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7529 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7530 (addsi3, subsi3): ... this.
7532 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7534 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7536 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7538 * config/nds32/nds32.md: Adjust indention.
7540 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7542 * config/nds32/nds32.md (feature): New attribute.
7544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7546 * config/nds32/nds32.md (subtype): New attribute.
7548 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7551 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7552 expansion to perform a bitwise AND of the argument followed by a
7553 boolean negation of the result.
7555 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7557 PR rtl-optimization/84878
7558 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7559 the basic block. Assert the use reference is not artificial and that
7560 it has an associated insn.
7562 2018-04-04 Michael Matz <matz@suse.de>
7564 * builtins.c (compute_objsize): Pass correct operand
7565 to array_at_struct_end_p.
7567 2018-04-04 Richard Biener <rguenther@suse.de>
7570 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7571 from contexts for DINFO_LEVEL_TERSE and below.
7573 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7575 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7576 resiter_operand condition.
7577 * config/nds32/nds32.md (*move<mode>): Ditto.
7579 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7580 Monk Chiang <sh.chiang04@gmail.com>
7582 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7584 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7586 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7588 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7589 Kito Cheng <kito.cheng@gmail.com>
7591 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7592 nds32_cond_code_str, output_cond_branch,
7593 output_cond_branch_compare_zero, nds32_expand_cbranch,
7594 nds32_expand_cstore, nds32_expand_movcc,
7595 nds32_output_cbranchsi4_equality_zero,
7596 nds32_output_cbranchsi4_equality_reg,
7597 nds32_output_cbranchsi4_equality_reg_or_const_int,
7598 nds32_output_cbranchsi4_greater_less_zero: New functions.
7599 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7600 nds32_expand_cstore, nds32_expand_movcc,
7601 nds32_output_cbranchsi4_equality_zero,
7602 nds32_output_cbranchsi4_equality_reg,
7603 nds32_output_cbranchsi4_equality_reg_or_const_int,
7604 nds32_output_cbranchsi4_greater_less_zero): Declare.
7605 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7606 nds32_rimm11s_operand): New predicates.
7607 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7608 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7611 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7613 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7614 * config/nds32/nds32.md: Ditto.
7615 * config/nds32/pipelines.md: Ditto.
7617 2018-04-04 Richard Biener <rguenther@suse.de>
7619 PR tree-optimization/85168
7620 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7621 propagating abnormals.
7623 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7625 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7627 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7628 Kito Cheng <kito.cheng@gmail.com>
7630 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7631 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7632 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7633 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7634 * config/nds32/nds32.md (sibcall_internal): New.
7635 (sibcall_register): Remove.
7636 (sibcall_immediate): Remove.
7637 (sibcall_value_internal): New.
7638 (sibcall_value_register): Remove.
7639 (sibcall_value_immediate): Remove.
7640 * config/nds32/predicates.md (nds32_general_register_operand): New.
7641 (nds32_call_address_operand): New.
7643 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7645 PR rtl-optimization/85167
7646 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7647 bb_defs if *split_p, instead preinitialize it to NULL.
7649 PR tree-optimization/85156
7650 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7651 evaluating the argument multiple times.
7653 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7655 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7657 (_mm_cvtpd_ps): Likewise.
7658 (_mm_cvttpd_epi32): Likewise.
7659 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7660 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7661 vector, pixel, and bool following altivec.h include.
7663 2018-04-03 Martin Sebor <msebor@redhat.com>
7665 * doc/extend.texi (Common Function Attributes): Clarify.
7666 (const attribute): Likewise.
7667 (pure attribute): Likewise.
7669 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7672 * config/i386/i386.c (ix86_expand_vector_set): Use
7673 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7675 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7677 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7678 instructions when changing rounding bits to preserve precision bits
7679 in the x87 control word.
7681 2018-04-03 Martin Liska <mliska@suse.cz>
7683 PR tree-optimization/82491
7684 * rtl.h (strip_offset_and_add): Replace += suboffset with
7685 poly_uint64 () + suboffset.
7687 2018-03-29 Martin Liska <mliska@suse.cz>
7688 Martin Jambor <mjambor@suse.cz>
7691 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7692 param_type is not an integral or pointer type.
7694 2018-04-03 Richard Biener <rguenther@suse.de>
7696 * sese.h (recompute_all_dominators): Remove.
7698 2018-04-02 Martin Sebor <msebor@redhat.com>
7700 * doc/invoke.texi (-Wrestrict): Fix typos.
7702 2018-04-02 Jim Wilson <jimw@sifive.com>
7704 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7705 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7706 (<optab>di3, <optab>si3_extend): Likewise.
7707 (<optab>si3_mask, <optab>si3_mask_1): New.
7708 (<optab>di3_mask, <optab>di3_mask_1): New.
7709 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7710 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7711 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7713 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7715 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7718 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7720 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7721 (nds32_canonicalize_comparison): New function.
7723 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7724 Kito Cheng <kito.cheng@gmail.com>
7725 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7727 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7728 * config/nds32/constants.md (unspec_volatile_element): Add
7729 UNSPEC_VOLATILE_RELAX_GROUP.
7730 * config/nds32/nds32-relax-opt.c: New file.
7731 * config/nds32/nds32-predicates.c
7732 (nds32_symbol_load_store_p): New function.
7733 * config/nds32/nds32-protos.h
7734 (nds32_symbol_load_store_p): Declare function.
7735 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7736 * config/nds32/nds32.c
7737 (nds32_register_pass): New function to register pass.
7738 (nds32_register_passes): New function to register passes.
7739 * config/nds32/nds32.md (relax_group): New pattern.
7740 * config/nds32/nds32.opt (mrelax-hint): New option.
7741 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7743 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7745 * config/nds32/t-nds32: Modify files dependency.
7747 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7749 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7750 (PROFILE_HOOK): Define its implementation.
7752 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7754 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7755 type and 32-bit size.
7757 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7760 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7761 (V_128_256): New mode iterator.
7762 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7763 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7764 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7766 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7767 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7769 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7772 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7773 NaN inputs correctly.
7775 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7778 * config/rs6000/vsx.md (??r): New mode attribute.
7779 (*vsx_mov<mode>_64bit): Use it.
7780 (*vsx_mov<mode>_32bit): Likewise.
7782 2018-03-30 Martin Sebor <msebor@redhat.com>
7784 PR tree-optimization/84818
7785 * builtins.c (check_access): Use warning_n.
7787 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7790 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7792 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7795 2018-03-30 Julia Koval <julia.koval@intel.com>
7798 * x86-tune.def (movx, partial_reg_dependency): Enable for
7801 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7804 * lra-constraints.c (process_alt_operands): Move setting
7805 this_alternative_matches below.
7807 2018-03-29 Martin Liska <mliska@suse.cz>
7810 * doc/invoke.texi: Document how LTO works with debug info.
7811 Describe auto-load support of binutils. Mention 'x86-64'
7812 as valid option value of -march option.
7814 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7816 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7819 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7820 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7821 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7824 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7827 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7828 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7829 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7830 for RS6000_BTM_POWERPC64.
7831 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7832 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7833 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7838 2018-03-28 Carl Love <cel@us.ibm.com>
7841 2017-09-27 Carl Love <cel@us.ibm.com>
7843 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7844 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7845 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7848 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7850 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7851 instead of __vector bool.
7852 (_mm_max_pu8): Likewise.
7853 (_mm_min_pi16): Likewise.
7855 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7858 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7859 (DIVWEUO): Likewise.
7861 (DIVDEUO): Likewise.
7862 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7863 DIVWEUO and DIVDEUO.
7864 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7865 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7866 (div_extend): Likewise.
7867 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7869 (__builtin_divweuo): Likewise.
7870 (__builtin_divdeo): Likewise.
7871 (__builtin_divdeuo): Likewise.
7873 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7876 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7877 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7879 PR tree-optimization/82004
7880 * gimple-match-head.c (optimize_pow_to_exp): New function.
7881 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7882 Don't fold to exp if optimize_pow_to_exp is false.
7884 2018-03-28 Martin Liska <mliska@suse.cz>
7887 * calls.c (initialize_argument_information): Fix trailing space.
7888 * common.opt: Fix typo and provide better explanation for
7889 -fsanitize-coverage option.
7890 * config/i386/i386.opt: Fix typo.
7892 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7893 Martin Liska <mliska@suse.cz>
7896 * gimplify.c (asan_poison_variable): Don't do the check for
7897 gimplify_omp_ctxp here.
7898 (gimplify_decl_expr): Do it here.
7899 (gimplify_target_expr): Likewise.
7901 2018-03-28 Martin Liska <mliska@suse.cz>
7904 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7905 chkp_type_bounds_count if MPX is not enabled.
7907 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7909 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7911 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7914 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7915 function to create the function decl for complex long double
7916 multiply and divide for -mabi=ieeelongdouble.
7917 (init_float128_ieee): Call it.
7919 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7922 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7923 -fcf-protection=branch -mibt.
7924 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7926 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7929 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7930 arm_word_relocations.
7932 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7935 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7936 extern array declarations.
7938 2018-03-27 Richard Biener <rguenther@suse.de>
7941 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7942 explicit single_use checks.
7944 2018-03-27 Richard Biener <rguenther@suse.de>
7946 PR tree-optimization/85082
7947 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7950 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7952 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7953 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7954 Turn on fasynchronous-unwind-tables and funwind-tables.
7956 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7959 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7960 (*bmi_blsr_<mode>_ccz): Ditto.
7962 2018-03-26 Tom de Vries <tom@codesourcery.com>
7964 PR tree-optimization/85063
7965 * omp-general.c (offloading_function_p): New function. Factor out
7967 * omp-offload.c (pass_omp_target_link::gate): ... here.
7968 * omp-general.h (offloading_function_p): Declare.
7969 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7970 with attribute omp declare target for offloading functions.
7972 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7974 PR tree-optimization/84005
7975 * tree-data-ref.h (get_base_for_alignment): Declare.
7976 * tree-data-ref.c (get_base_for_alignment_1): New function.
7977 (get_base_for_alignment): Likewise.
7978 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7979 get_base_for_alignment to find a suitable base object, instead
7980 of always using drb->base_address.
7982 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7985 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7986 known size by default.
7988 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7991 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7992 and non BLKmode operands.
7994 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7997 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7998 Clean up attributes.
8000 2018-03-23 Richard Biener <rguenther@suse.de>
8003 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8004 we are going to emit early debug for LTO.
8006 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8009 * function.c (match_asm_constraints_1): Don't optimize if input
8010 doesn't satisfy general_operand predicate for output's mode.
8013 * alias.c (write_dependence_p): Don't require for x_canonicalized
8014 non-VOIDmode if x has VOIDmode.
8017 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8018 just don't try to optimize it rather than assert it never happens.
8020 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8022 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8023 macro expansions for definition of ST_INTERNAL_<mode> and
8024 LD_INTERNAL_<mode> builtins.
8025 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8027 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8029 (altivec_expand_st_builtin): Likewise.
8030 (altivec_expand_builtin): Remove calls to deleted functions.
8031 (rs6000_address_for_altivec): Delete this function.
8032 * config/rs6000/vector.md: Remove expands for
8033 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8035 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8038 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8039 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8040 re-computing once computed.
8041 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8042 (arm_init_machine_status): Initialize
8043 machine->static_chain_stack_bytes.
8045 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8048 * doc/extend.texi: Add four new prototypes for vec_ld.
8049 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8050 definitions for more logical presentation.
8051 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8052 entries for V1TI variants of __builtin_altivec_ld builtin.
8053 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8054 handling of V1TI variant of LVX icode pattern.
8055 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8056 (rs6000_gimple_fold_builtin): Likewise.
8057 (altivec_init_builtins): Add code to define
8058 __builtin_altivec_lvx_v1ti function.
8060 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8063 * function.c (match_asm_constraints_1): Don't do the optimization
8064 if input isn't a REG, SUBREG, MEM or constant.
8066 2018-03-22 Tom de Vries <tom@codesourcery.com>
8068 PR tree-optimization/84956
8069 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8070 bb_has_abnormal_pred.
8072 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8075 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8076 DECL_INITIAL (decl) to decl at the end.
8077 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8080 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8082 * doc/extend.texi (__builtin_tgmath): Document when complex
8083 integer types are treated as _Complex _Float64.
8085 2018-03-21 Tom de Vries <tom@codesourcery.com>
8087 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8089 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8091 PR tree-optimization/84960
8092 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8093 if it is ENTRY block, move them into single succ of ENTRY in that case.
8095 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8097 PR tree-optimization/84811
8098 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8099 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8100 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8101 (known_subrange_p): Update accordingly. Cast each value involved
8102 in the size comparison, rather than casting the result of the
8105 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8107 PR tree-optimization/84982
8108 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8109 by flipping the least significant bit rather than all bits from
8110 bitpos to bitpos + bitsize - 1.
8112 2018-03-21 Nathan Sidwell <nathan@acm.org>
8114 * doc/extend.texi (Deprecated Features): Remove mention of
8115 long-deleted deprecations.
8117 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8120 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8121 * configure: Regenerate.
8123 2018-03-21 Tom de Vries <tom@codesourcery.com>
8125 PR tree-optimization/83126
8126 * tree-parloops.c (num_phis): New function.
8127 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8129 2018-03-21 Nathan Sidwell <nathan@acm.org>
8131 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8132 mention anon-struct/union members and trailing attributes.
8134 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8136 PR tree-optimization/84969
8137 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8138 builtin memset partitions if they set different rhs values.
8140 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8142 PR rtl-optimization/84989
8143 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8144 VEC_DUPLICATE with scalar result mode.
8146 2018-03-21 Martin Liska <mliska@suse.cz>
8149 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8150 not intended return statement.
8152 2018-03-21 Martin Liska <mliska@suse.cz>
8155 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8156 (chkp_find_bound_slots_1): Limit number of iterations.
8158 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8161 * Minor grammar fixes for x86 options.
8163 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8166 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8167 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8169 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8172 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8173 (altivec_lvx_<mode>_1op): Likewise.
8174 (altivec_stvx_<mode>_2op): Likewise.
8175 (altivec_stvx_<mode>_1op): Likewise.
8176 (altivec_lvx_<VM2:mode>): New define_expand.
8177 (altivec_stvx_<VM2:mode>): Likewise.
8178 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8179 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8180 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8181 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8182 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8183 (rs6000_gen_lvx): Likewise.
8184 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8185 (altivec_expand_stv_builtin): Likewise.
8186 (altivec_expand_builtin): Likewise.
8187 * config/rs6000/vector.md: Likewise.
8189 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8192 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8195 2018-03-20 Richard Biener <rguenther@suse.de>
8198 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8199 sign-conversions as zero, fall back to standard scalar_stmt
8202 2018-03-20 Martin Liska <mliska@suse.cz>
8205 * predict.c (rebuild_frequencies): Handle case when we have
8206 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8208 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8211 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8212 flag_section_anchors.
8213 * varasm.c (use_blocks_for_decl_p): Remove hack for
8214 dw2_force_const_mem.
8217 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8219 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8220 be created, use lowpart_subreg of operands[0] rather than operands[0]
8222 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8223 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8224 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8225 and n constraint instead of aarch64_shift_imm_di and Usd.
8226 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8227 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8229 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8232 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8233 to favor GPR over NEON registers.
8234 (<shift>di3_neon): Likewise.
8236 2018-03-20 Tom de Vries <tom@codesourcery.com>
8239 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8240 (nvptx_process_pars): Emit bar.sync asap and alap.
8242 2018-03-20 Tom de Vries <tom@codesourcery.com>
8245 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8246 seen_label if seen_label is already set.
8248 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8251 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8252 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8253 Use 1U instead of 1. Formatting fixes.
8256 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8257 instead of TREE_TYPE (s1) for the return value.
8259 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8261 PR tree-optimization/84946
8262 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8263 bitsize + bitsize in poly_uint64 rather than poly_int64.
8266 * dwarf2asm.c: Include fold-const.c.
8267 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8268 of decl rather than decl itself.
8270 PR rtl-optimization/84643
8271 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8273 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8276 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8277 calling assemble_variable.
8279 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8282 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8283 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8285 2018-03-19 Jim Wilson <jimw@sifive.com>
8288 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8289 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8290 (riscv_first_stack_step): Likewise.
8291 (riscv_option_override): Use STACK_BOUNDARY instead of
8293 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8295 (BIGGEST_ALIGNMENT): Set to 128.
8296 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8297 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8300 2018-03-19 Richard Biener <rguenther@suse.de>
8302 PR tree-optimization/84933
8303 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8304 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8306 2018-03-19 Richard Biener <rguenther@suse.de>
8308 PR tree-optimization/84859
8309 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8310 (cond_if_else_store_replacement): Perform sinking operation on
8311 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8312 Generalize what a BB with a single eligible store is.
8314 2018-03-19 Richard Biener <rguenther@suse.de>
8316 PR tree-optimization/84929
8317 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8318 chrec_is_positive against non-chrec arg.
8320 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8323 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8325 2018-03-18 Martin Liska <mliska@suse.cz>
8327 PR rtl-optimization/84635
8328 * regrename.c (build_def_use): Use matches_mode only when
8331 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8333 PR tree-optimization/84913
8334 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8335 vectorize chains of COND_EXPRs.
8337 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8339 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8341 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8343 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8345 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8347 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8349 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8350 Kito Cheng <kito.cheng@gmail.com>
8352 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8353 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8354 (nds32_adjust_reg_alloc_order): New function.
8355 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8357 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8359 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8360 nds32_print_operand, nds32_print_operand_address): Use
8361 HOST_WIDE_INT_PRINT_DEC instead.
8363 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8365 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8367 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8370 * config/i386/i386.c (initial_ix86_tune_features,
8371 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8373 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8374 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8375 rather than 1u << ix86_tune. Formatting fix.
8376 (ix86_option_override_internal): Change ix86_arch_mask from
8377 unsigned int to unsigned HOST_WIDE_INT, initialize to
8378 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8379 (ix86_function_specific_restore): Likewise.
8381 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8384 * postreload.c (reload_combine_recognize_pattern): Perform
8385 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8386 truncate_int_for_mode the result for the destination's mode.
8389 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8390 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8392 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8395 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8398 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8399 regno_allocno_class_array and sorted_pseudos.
8400 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8401 insns where regno is used.
8403 2018-03-16 Martin Liska <mliska@suse.cz>
8406 * multiple_target.c (create_dispatcher_calls): Redirect
8407 reference in the symbol table.
8409 2018-03-16 Martin Liska <mliska@suse.cz>
8412 * multiple_target.c (create_dispatcher_calls): Redirect also
8415 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8419 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8420 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8421 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8423 2018-03-16 Julia Koval <julia.koval@intel.com>
8425 * doc/invoke.texi (Skylake Server): Add CLWB.
8426 Cannonlake): Remove CLWB.
8428 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8430 PR tree-optimization/84841
8431 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8433 (FLOAT_ONE_CONST_TYPE): Define.
8434 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8435 (sort_by_operand_rank): Put entries with higher constant_type last
8436 rather than first to match comments.
8438 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8440 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8443 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8446 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8447 check and use address of target always.
8449 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8452 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8453 (indirect_thunk_bnd_needed): Likewise.
8454 (indirect_thunks_used): Likewise.
8455 (indirect_thunks_bnd_used): Likewise.
8456 (indirect_return_needed): New.
8457 (indirect_return_bnd_needed): Likewise.
8458 (output_indirect_thunk_function): Add a bool argument for
8460 (output_indirect_thunk_function): Don't generate alias for
8461 function return thunk.
8462 (ix86_code_end): Call output_indirect_thunk_function to generate
8463 function return thunks.
8464 (ix86_output_function_return): Set indirect_return_bnd_needed
8465 and indirect_return_needed instead of indirect_thunk_bnd_needed
8466 and indirect_thunk_needed.
8468 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8470 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8471 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8472 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8474 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8475 Paul Hua <paul.hua.gm@gmail.com>
8478 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8480 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8482 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8483 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8486 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8489 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8490 instead of GET_MODE_SIZE when comparing Units.
8492 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8495 * varasm.c (hash_section): Return an unchangeble hash value
8496 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8497 Return !aarch64_can_use_per_function_literal_pools_p ().
8499 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8502 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8503 rather than address of cmode as last argument to prepare_cmp_insn.
8505 2018-03-15 Julia Koval <julia.koval@intel.com>
8507 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8508 F_AVX512VNNI, F_AVX512BITALG): New.
8510 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8513 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8514 insn for floating-point loads and stores.
8516 2018-03-14 Carl Love <cel@us.ibm.com>
8518 * config/rs6000/rs6000-c.c: Add macro definitions for
8519 ALTIVEC_BUILTIN_VEC_PERMXOR.
8520 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8521 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8522 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8523 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8525 * config/doc/extend.texi: Add prototypes for vec_permxor.
8527 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8530 * diagnostic-show-locus.c (class layout_point): Convert m_line
8531 from int to linenum_type.
8532 (line_span::comparator): Use linenum "compare" function when
8533 comparing line numbers.
8534 (test_line_span): New function.
8535 (layout_range::contains_point): Convert param "row" from int to
8537 (layout_range::intersects_line_p): Likewise.
8538 (layout::will_show_line_p): Likewise.
8539 (layout::print_source_line): Likewise.
8540 (layout::should_print_annotation_line_p): Likewise.
8541 (layout::print_annotation_line): Likewise.
8542 (layout::print_leading_fixits): Likewise.
8543 (layout::annotation_line_showed_range_p): Likewise.
8544 (struct line_corrections): Likewise for field m_row.
8545 (line_corrections::line_corrections): Likewise for param "row".
8546 (layout::print_trailing_fixits): Likewise.
8547 (layout::get_state_at_point): Likewise.
8548 (layout::get_x_bound_for_row): Likewise.
8549 (layout::print_line): Likewise.
8550 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8551 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8552 * input.c (selftest::test_linenum_comparisons): New function.
8553 (selftest::input_c_tests): Call it.
8554 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8555 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8556 * selftest.h (ASSERT_GT): New macro.
8557 (ASSERT_GT_AT): New macro.
8558 (ASSERT_LT): New macro.
8559 (ASSERT_LT_AT): New macro.
8561 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8563 PR rtl-optimization/84780
8564 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8566 2018-03-14 Martin Liska <mliska@suse.cz>
8568 * tree.c (record_node_allocation_statistics): Use
8569 get_stats_node_kind.
8570 (get_stats_node_kind): New function extracted from
8571 record_node_allocation_statistics.
8572 (free_node): Use get_stats_node_kind.
8574 2018-03-14 Richard Biener <rguenther@suse.de>
8576 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8577 that the value-set of ANTIC_IN doesn't grow.
8580 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8582 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8583 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8585 2018-03-14 Julia Koval <julia.koval@intel.com>
8587 * config.gcc (icelake-client, icelake-server): New.
8589 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8590 (initial_ix86_arch_features): Ditto.
8591 (PTA_SKYLAKE): Add SGX.
8592 (PTA_ICELAKE): Remove.
8593 (PTA_ICELAKE_CLIENT): New.
8594 (PTA_ICELAKE_SERVER): New.
8595 (ix86_option_override_internal): Split up icelake on icelake client and
8597 (get_builtin_code_for_version): Ditto.
8598 (fold_builtin_cpu): Ditto.
8599 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8600 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8601 * config/i386/i386.h (processor_type): Ditto.
8602 * doc/invoke.texi: Ditto.
8604 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8607 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8608 INTEGER_CST offset, add it together with bitpos / 8 and
8609 sign extend based on POINTER_SIZE.
8613 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8616 * config/i386/constraints.md (Yc): New register constraint.
8617 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8618 Use Yc constraint for alternative 2 of operand 0. Remove
8619 preferred_for_speed attribute.
8621 2018-03-14 Richard Biener <rguenther@suse.de>
8623 PR tree-optimization/84830
8624 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8625 with the old one to avoid oscillations.
8627 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8630 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8632 (assign_by_spills): Return a flag of reload assignment failure.
8633 Do not process the reload assignment failures. Do not spill other
8634 reload pseudos if they has the same reg class. Update n if
8636 (lra_assign): Add a return arg. Set up from the result of
8637 assign_by_spills call.
8638 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8639 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8640 usage_insns if it is not NULL.
8641 (spill_hard_reg_in_range): New function.
8642 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8643 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8644 function prototypes.
8645 (lra_assign): Change prototype.
8646 * lra.c (lra): Add code to deal with fails by splitting hard reg
8649 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8651 * config/riscv/riscv.opt (mrelax): New option.
8652 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8653 "norelax" when riscv_mrelax is disabled.
8654 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8656 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8659 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8660 reassociation for int modes.
8662 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8664 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8665 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8667 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8668 * config/aarch64/aarch64-sve.md
8669 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8670 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8671 (*extend<mode><Vwide>2): Rename to...
8672 (aarch64_sve_extend<mode><Vwide>2): ...this.
8673 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8674 renaming the old pattern to...
8675 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8677 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8678 define_expand, renaming the old pattern to...
8679 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8680 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8681 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8682 account when deciding which SVE instruction the optab should use.
8683 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8685 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8687 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8688 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8689 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8690 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8691 (tlsdesc_small_<mode>): Turn a define_expand and use
8692 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8693 (tlsdesc_small_advsimd_<mode>): ...this.
8694 (tlsdesc_small_sve_<mode>): New pattern.
8696 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8698 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8699 (UNSPEC_UMUL_HIGHPART): New constants.
8700 (MUL_HIGHPART): New int iteraor.
8701 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8702 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8704 (*<su>mul<mode>3_highpart): New define_insn.
8706 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8709 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8712 2018-03-13 Martin Liska <mliska@suse.cz>
8715 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8717 (sem_item_optimizer::~sem_item_optimizer): Release it.
8718 (sem_item_optimizer::merge_classes): Register variable aliases.
8719 (sem_item_optimizer::fixup_pt_set): New function.
8720 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8721 * ipa-icf.h: Declare new variables and functions.
8723 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8726 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8727 integer_pow2p@2 and test integer_pow2p in condition.
8728 (A < 0 ? C : 0): Similarly for @1.
8731 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8732 characters starting at p contain '\0' character, don't look beyond
8736 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8737 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8740 * reg-stack.c (change_stack): Change update_end var from int to
8741 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8742 also call set_block_for_insn on the newly added insns and rescan.
8745 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8746 on the last operand.
8749 * tree.c (stabilize_reference_1): Return save_expr (e) for
8750 STATEMENT_LIST even if it doesn't have side-effects.
8752 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8754 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8756 2018-03-12 Renlin Li <renlin.li@arm.com>
8758 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8759 aarch64_output_scalar_simd_mov_immediate.
8761 2018-03-12 Martin Sebor <msebor@redhat.com>
8763 PR tree-optimization/83456
8764 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8765 for perfectly overlapping calls to memcpy.
8766 (gimple_fold_builtin_memory_chk): Same.
8767 (gimple_fold_builtin_strcpy): Handle no-warning.
8768 (gimple_fold_builtin_stxcpy_chk): Same.
8769 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8771 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8773 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8774 parameter. Use it for SFmode.
8775 (rs6000_function_arg_advance_1): Adjust.
8776 (rs6000_function_arg): Adjust.
8777 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8779 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8781 PR rtl-optimization/84169
8782 PR rtl-optimization/84780
8783 * combine.c (can_combine_p): Check for a 2-insn combination whether
8784 the destination register is used between the two insns, too.
8786 2018-03-12 Richard Biener <rguenther@suse.de>
8788 PR tree-optimization/84803
8789 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8790 for refs DR analysis didn't process.
8792 2018-03-12 Richard Biener <rguenther@suse.de>
8794 PR tree-optimization/84777
8795 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8796 force-vectorize loops ignore whether we are optimizing for size.
8798 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8800 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8801 (TARGET_MD_ASM_ADJUST): Define.
8803 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8804 Kito Cheng <kito.cheng@gmail.com>
8805 Chung-Ju Wu <jasonwucj@gmail.com>
8807 * config/nds32/nds32.c (nds32_compute_stack_frame,
8808 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8809 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8810 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8811 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8812 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8813 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8814 * config/nds32/nds32.md (prologue, epilogue): Use macro
8815 NDS32_V3PUSH_AVAILABLE_P to do checking.
8817 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8820 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8821 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8822 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8823 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8824 addition of most attributes on !orig_type_die or the attribute not
8825 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8827 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8828 Chung-Ju Wu <jasonwucj@gmail.com>
8830 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8832 * config/nds32/nds32.opt (mvh): New option.
8834 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8835 Chung-Ju Wu <jasonwucj@gmail.com>
8837 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8839 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8840 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8843 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8844 Chung-Ju Wu <jasonwucj@gmail.com>
8846 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8848 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8849 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8851 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8852 Kito Cheng <kito.cheng@gmail.com>
8853 Chung-Ju Wu <jasonwucj@gmail.com>
8855 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8856 UNSPEC_FFMISM and UNSPEC_FLMISM.
8857 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8858 for ffb, ffmism and flmism.
8859 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8860 (unspec_ffmism): Ditto.
8861 (unspec_flmism): Ditto.
8862 (nds32_expand_builtin_impl): Check if string extension is available.
8863 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8864 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8866 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8869 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8872 * lra-assigns.c (assign_by_spills): Return a flag of reload
8873 assignment failure. Do not process the reload assignment
8874 failures. Do not spill other reload pseudos if they has the same
8876 (lra_assign): Add a return arg. Set up from the result of
8877 assign_by_spills call.
8878 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8879 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8880 usage_insns if it is not NULL.
8881 (spill_hard_reg_in_range): New function.
8882 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8883 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8884 function prototypes.
8885 (lra_assign): Change prototype.
8886 * lra.c (lra): Add code to deal with fails by splitting hard reg
8889 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8892 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8894 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8897 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8898 (dw_val_node): Add val_symbolic_view.
8899 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8900 (symview_upper_bound): New.
8901 (new_line_info_table): Initialize symviews_since_reset.
8902 (dwarf2out_source_line): Count symviews_since_reset and set
8903 symview_upper_bound.
8904 (dw_val_equal_p): Handle symview.
8905 (add_AT_symview): New.
8906 (print_dw_val): Handle symview.
8907 (attr_checksum, attr_checksum_ordered): Likewise.
8908 (same_dw_val_p, size_of_die): Likewise.
8909 (value_format, output_die): Likewise.
8910 (add_high_low_attributes): Use add_AT_symview for entry_view.
8911 (dwarf2out_finish): Reset symview_upper_bound, clear
8914 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8917 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8918 Add strict argument and use it.
8919 (rs6000_split_multireg_move): Update for new strict argument.
8920 (mem_operand_gpr): Disallow all non-offsettable addresses.
8921 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8923 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8926 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8927 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8928 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8931 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8932 decl, use remap_type if we want to use the type.
8934 2018-03-09 Martin Sebor <msebor@redhat.com>
8936 PR tree-optimization/84526
8937 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8939 (builtin_access::generic_overlap): Be prepared to handle non-array
8942 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8944 PR rtl-optimization/84682
8945 * lra-constraints.c (process_address_1): Check is_address flag
8946 for address constraints.
8947 (process_alt_operands): Likewise.
8948 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8949 preprocess_constraints.
8950 * recog.h (preprocess_constraints): Add oploc parameter.
8952 * recog.c (preprocess_constraints): Test address_operand for
8953 CT_ADDRESS constraints.
8955 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8958 * lra-assigns.c (assign_by_spills): Return a flag of reload
8959 assignment failure. Do not process the reload assignment
8960 failures. Do not spill other reload pseudos if they has the same
8962 (lra_assign): Add a return arg. Set up from the result of
8963 assign_by_spills call.
8964 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8965 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8966 usage_insns if it is not NULL.
8967 (spill_hard_reg_in_range): New function.
8968 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8969 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8970 function prototypes.
8971 (lra_assign): Change prototype.
8972 * lra.c (lra): Add code to deal with fails by splitting hard reg
8975 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8978 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8979 Accept complain bool parameter. Only emit errors if it is true.
8980 (arm_parse_cpu_option_name): Likewise.
8981 (arm_target_thumb_only): Adjust callers of the above.
8982 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8983 prototype to take a default true bool parameter.
8984 (arm_parse_arch_option_name): Likewise.
8986 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8987 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8991 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8992 * configure: Regenerate.
8993 * configure.ac ("linker --version-script option"): New.
8994 ("linker soname option"): New.
8996 2018-03-09 Richard Biener <rguenther@suse.de>
8998 PR tree-optimization/84775
8999 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9000 immediate uses of predicate stmts and mark them modified.
9003 PR tree-optimization/84178
9004 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9006 (version_loop_for_if_conversion): Delay update_ssa call.
9007 (tree_if_conversion): Delay update_ssa until after predicate
9010 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9013 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9014 when the function accesses prior frames.
9016 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9019 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9020 gen_llsym, otherwise call maybe_gen_llsym.
9023 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9024 has ',' character inside of it.
9026 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9029 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9030 as clobbering CC_REGNUM.
9032 2018-03-08 Richard Biener <rguenther@suse.de>
9035 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9036 (follow_copies_to_constant): Do not follow SSA names registered
9039 2018-03-08 Richard Biener <rguenther@suse.de>
9041 PR tree-optimization/84178
9042 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9044 (version_loop_for_if_conversion): Delay update_ssa call.
9045 (tree_if_conversion): Delay update_ssa until after predicate
9048 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9050 PR tree-optimization/84178
9051 * tree-if-conv.c (release_bb_predicate): Remove the
9052 the assertion that the stmts have NULL use_ops.
9053 Discard the statements, asserting that they haven't
9054 yet been added to a BB.
9056 2018-03-08 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/84746
9059 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9060 (phi_translate): Pass in destination ANTIC_OUT set.
9061 (phi_translate_1): Likewise. For a simplified result lookup
9062 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9063 (phi_translate_set): Adjust.
9064 (do_pre_regular_insertion): Likewise.
9065 (do_pre_partial_partial_insertion): Likewise.
9067 2018-03-08 Martin Liska <mliska@suse.cz>
9069 PR gcov-profile/84735
9070 * doc/gcov.texi: Document usage of profile files.
9071 * gcov-io.h: Document changes in the format.
9073 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9077 * dwarf2out.c (struct dw_line_info_table): Update comments for
9079 (FORCE_RESET_NEXT_VIEW): New.
9080 (FORCE_RESETTING_VIEW_P): New.
9081 (RESETTING_VIEW_P): Check for -1 too.
9082 (ZERO_VIEW_P): Likewise.
9083 (new_line_info_table): Force-reset next view.
9084 (dwarf2out_begin_function): Likewise.
9085 (dwarf2out_source_line): Simplify zero_view_p initialization.
9086 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9087 view directly. Omit view when omitting .loc at line 0.
9089 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9091 PR tree-optimization/84740
9092 * tree-switch-conversion.c (process_switch): Call build_constructors
9093 only if info.phi_count is non-zero.
9095 PR tree-optimization/84739
9096 * tree-tailcall.c (find_tail_calls): Check call arguments against
9097 DECL_ARGUMENTS (current_function_decl) rather than
9098 DECL_ARGUMENTS (func) when checking for tail recursion.
9100 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9102 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9103 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9104 Volker Reichelt's entry and add entries for people that perform
9105 GCC fuzzy testing and report numerous bugs.
9107 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9110 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9111 readonly data in sdata, if that is disabled.
9112 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9113 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9114 -mreadonly-in-sdata option.
9116 2018-03-07 Martin Sebor <msebor@redhat.com>
9118 PR tree-optimization/84468
9119 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9120 basic block when looking for nul assignment.
9122 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9125 * except.h (output_function_exception_table): Adjust prototype.
9126 * except.c (output_function_exception_table): Remove FNNAME parameter
9127 and add SECTION parameter. Ouput one part of the table at a time.
9128 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9129 the first part of the exception table and emit unwind directives.
9130 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9131 (i386_pe_seh_cold_init): Likewise.
9132 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9133 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9134 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9135 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9136 * config/i386/winnt.c: Include except.h.
9137 (struct seh_frame_state): Add reg_offset, after_prologue and
9138 in_cold_section fields.
9139 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9140 (i386_pe_seh_cold_init): New function.
9141 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9142 to seh->in_cold_section.
9143 (seh_emit_push): Record the offset of the push.
9144 (seh_emit_save): Record the offet of the save.
9145 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9146 Test seh->after_prologue to disregard the epilogue.
9147 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9148 (i386_pe_end_cold_function): New function.
9150 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9153 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9154 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9157 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9159 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9160 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9163 * multiple_target.c: Include tree-inline.h and intl.h.
9164 (expand_target_clones): Diagnose and fail if node->definition and
9165 !tree_versionable_function_p (node->decl).
9167 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9169 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9171 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9172 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9173 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9175 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9178 * combine.c (try_combine): Use reg_or_subregno instead of handling
9179 just paradoxical SUBREGs and REGs.
9181 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9183 * config/arc/arc.c (arc_finalize_pic): Remove function.
9184 (arc_must_save_register): We use single base PIC register, remove
9185 checks to save/restore the PIC register.
9186 (arc_expand_prologue): Likewise.
9187 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9189 (arc_verify_short): Likewise.
9190 (arc_attr_type): Likewise.
9191 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9192 (walk_stores): Likewise.
9193 (arc_address_cost): Make it static.
9194 (arc_verify_short): Likewise.
9195 (branch_dest): Likewise.
9196 (arc_attr_type): Likewise.
9197 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9198 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9199 (arc_final_prescan_insn): Remove inserting the nops due to
9200 hardware hazards. It is done in reorg step.
9201 (insn_length_variant_t): Remove.
9202 (insn_length_parameters_t): Likewise.
9203 (arc_insn_length_parameters): Likewise.
9204 (arc_get_insn_variants): Likewise.
9205 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9207 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9210 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9213 PR tree-optimization/84687
9214 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9216 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9218 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9220 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9221 Rename to ppc_speculation_barrier.
9222 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9223 __builtin_ppc_speculation_barrier.
9225 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9228 * combine.c (combine_simplify_rtx): Don't try to simplify if
9229 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9232 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9234 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9235 to 32 bytes when compiling for POWER9.
9237 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9240 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9241 regparm >= 3 with no arg reg available also for calls with
9242 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9245 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9247 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9249 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9252 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9254 2018-03-05 Richard Biener <rguenther@suse.de>
9256 PR tree-optimization/84486
9257 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9258 When inserting a __builtin_assume_aligned call set the LHS
9259 SSA name alignment info accordingly.
9261 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9263 PR tree-optimization/84114
9264 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9265 Avoid reassociation of FLOAT_MODE addition.
9267 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9269 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9270 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9271 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9272 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9273 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9274 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9275 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9277 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9278 __builtin_ia32_wbinvd): New builtins.
9279 (SPECIAL_ARGS2): New.
9280 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9281 (SPECIAL_ARGS2): New.
9282 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9283 (ix86_valid_target_attribute_inner_p): Ditto.
9284 (ix86_init_mmx_sse_builtins): Add special_args2.
9285 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9286 TARGET_WBNOINVD_P): New.
9287 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9288 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9289 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9290 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9291 * config/i386/pconfigintrin.h: New file.
9292 * config/i386/wbnoinvdintrin.h: Ditto.
9293 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9295 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9297 2018-03-05 Richard Biener <rguenther@suse.de>
9299 PR tree-optimization/84670
9300 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9302 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9303 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9304 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9305 doesn't grow if all successors (recursively) were visited at least
9308 2018-03-05 Richard Biener <rguenther@suse.de>
9310 PR tree-optimization/84650
9311 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9312 if executed in the loop pipeline.
9314 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9316 * doc/configfiles.texi (Configuration Files): Move info about
9317 conditionalizing $target-protos.h to...
9318 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9319 differs from $target-protos.h.
9321 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9322 Chung-Ju Wu <jasonwucj@gmail.com>
9324 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9325 * config/nds32/nds32-multiple.md (setmemsi): Define.
9326 * config/nds32/nds32-memory-manipulation.c
9327 (nds32_gen_dup_4_byte_to_word_value): New.
9328 (emit_setmem_word_loop): New.
9329 (emit_setmem_byte_loop): New.
9330 (nds32_expand_setmem_loop): New.
9331 (nds32_expand_setmem_loop_v3m): New.
9332 (nds32_expand_setmem_unroll): New.
9333 (nds32_expand_setmem): New.
9335 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9336 Chung-Ju Wu <jasonwucj@gmail.com>
9338 * config/nds32/nds32-memory-manipulation.c
9339 (nds32_emit_load_store): New.
9340 (nds32_emit_post_inc_load_store): New.
9341 (nds32_emit_mem_move): New.
9342 (nds32_emit_mem_move_block): New.
9343 (nds32_expand_movmemsi_loop_unknown_size): New.
9344 (nds32_expand_movmemsi_loop_known_size): New.
9345 (nds32_expand_movmemsi_loop): New.
9346 (nds32_expand_movmemsi_unroll): New.
9347 (nds32_expand_movmemqi): Rename ...
9348 (nds32_expand_movmemsi): ... to this.
9349 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9350 (movmemsi): ... to this.
9351 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9352 (nds32_expand_movmemsi): ... to this.
9354 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9355 Monk Chiang <sh.chiang04@gmail.com>
9356 Chung-Ju Wu <jasonwucj@gmail.com>
9358 * config/nds32/nds32-protos.h
9359 (nds32_expand_load_multiple): New arguments.
9360 (nds32_expand_store_multiple): Ditto.
9361 (nds32_valid_multiple_load_store): Rename ...
9362 (nds32_valid_multiple_load_store_p): ... to this.
9363 * config/nds32/nds32-memory-manipulation.c
9364 (nds32_expand_load_multiple): Refine implementation.
9365 (nds32_expand_store_multiple): Ditto.
9366 * config/nds32/nds32-multiple.md
9367 (load_multiple): Update nds32_expand_load_multiple interface.
9368 (store_multiple): Update nds32_expand_store_multiple interface.
9369 * config/nds32/nds32-predicates.c
9370 (nds32_valid_multiple_load_store): Rename ...
9371 (nds32_valid_multiple_load_store_p): ... to this and refine
9373 * config/nds32/predicates.md
9374 (nds32_load_multiple_and_update_address_operation): New predicate.
9375 (nds32_store_multiple_and_update_address_operation): New predicate.
9377 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9378 Chung-Ju Wu <jasonwucj@gmail.com>
9380 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9381 (combo): New attribute.
9382 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9384 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9386 * config/nds32/nds32.opt: Change -mcmodel= default value.
9388 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9389 Monk Chiang <sh.chiang04@gmail.com>
9390 Chung-Ju Wu <jasonwucj@gmail.com>
9392 * config/nds32/constants.md (unspec_element): New enum.
9393 * config/nds32/constraints.md (Umw): New constraint.
9394 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9395 * config/nds32/nds32-intrinsic.md: Likewise.
9396 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9397 (nds32_valid_smw_lwm_base_p): New.
9398 (nds32_output_smw_single_word): New.
9399 (nds32_output_lmw_single_word): New.
9400 (nds32_expand_unaligned_load): New.
9401 (nds32_expand_unaligned_store): New.
9402 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9403 (nds32_output_smw_single_word): Declare.
9404 (nds32_output_lmw_single_word): Declare.
9405 (nds32_expand_unaligned_load): Declare.
9406 (nds32_expand_unaligned_store): Declare.
9407 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9408 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9409 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9410 NDS32_BUILTIN_UASTORE_DW.
9411 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9414 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9415 Kito Cheng <kito.cheng@gmail.com>
9416 Chung-Ju Wu <jasonwucj@gmail.com>
9418 * config/nds32/nds32-intrinsic.c
9419 (nds32_expand_builtin_null_ftype_reg): Delete.
9420 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9421 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9422 (nds32_read_argument): New.
9423 (nds32_legitimize_target): Ditto.
9424 (nds32_legitimize_argument): Ditto.
9425 (nds32_check_constant_argument): Ditto.
9426 (nds32_expand_unop_builtin): Ditto.
9427 (nds32_expand_unopimm_builtin): Ditto.
9428 (nds32_expand_binop_builtin): Ditto.
9429 (nds32_builtin_decl_impl): Ditto.
9430 (builtin_description): Ditto.
9431 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9432 (nds32_init_builtins_impl): Ditto.
9433 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9434 (nds32_builtin_decl): New.
9435 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9436 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9438 2018-03-02 Jeff Law <law@redhat.com>
9440 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9441 (redundant_insn, fill_simple_delay_slots): Likewise.
9442 (fill_slots_from_thread): Likewise.
9443 * resource.c (mark_referenced_resources): Likewise.
9444 (mark_set_resources, find_dead_or_set_registers): Likewise.
9446 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9448 * substring-locations.h (format_warning_va): Formatting fix for
9450 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9452 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9453 * substring-locations.c: Include intl.h.
9454 (format_warning_va): Turned into small wrapper around
9455 format_warning_n_va, renamed to ...
9456 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9457 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9459 (format_warning_at_substring_n): New function.
9460 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9461 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9462 format_warning_at_substring with just a shorter name instead of
9463 const function pointer.
9464 (fmtwarn_n): New function.
9465 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9466 appropriate, get rid of all the fmtstr temporaries, move conditionals
9467 with G_() wrapped string literals directly into fmtwarn arguments,
9468 cast dir.len to (int), formatting fixes.
9470 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9472 * doc/invoke.texi: Remove "Cilk Plus" references.
9474 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9475 Richard Biener <rguenther@suse.de>
9478 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9479 for error or warning attributes if CALL_FROM_THUNK_P is set.
9482 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9485 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9486 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9489 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9490 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9492 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9493 instead of -1U in last predictors element's probability member.
9495 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9498 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9499 arguments if they are comparable.
9501 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9503 PR tree-optimization/84634
9504 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9505 masks and masked_loop_p with a single loop_masks, making sure it's
9506 null for bb vectorization.
9508 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9510 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9511 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9512 loop->force_vectorize to check whether there is no alias.
9514 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9517 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9519 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9521 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9522 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9523 instead of a loop around prev_real_insn.
9524 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9528 * config/i386/i386.c (ix86_print_operand): Use conditional
9529 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9532 2018-03-02 Richard Biener <rguenther@suse.de>
9534 PR tree-optimization/84427
9535 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9536 (bitmap_set_subtract_values): Rewrite to handle multiple
9539 (prune_clobbered_mems): Likewise.
9540 (phi_translate): Take edge instead of pred/phiblock.
9541 (phi_translate_1): Likewise.
9542 (phi_translate_set): Likewise. Insert all translated
9543 exprs for a value into the set, keeping possibly multiple
9544 expressions per value.
9545 (compute_antic_aux): Adjust for phi_translate changes.
9546 When intersecting union the expressions and prune those
9547 not in the final value set, keeping possibly multiple
9548 expressions per value. Do not use value-insertion
9549 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9550 all expressions. Add verification that the value-sets
9551 only shrink during iteration.
9552 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9553 (do_pre_regular_insertion): Likewise.
9554 (do_pre_partial_partial_insertion): Likewise.
9556 2018-03-02 Richard Biener <rguenther@suse.de>
9559 * config/darwin.c (saved_debug_info_level): New static global.
9560 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9561 (darwin_asm_lto_end): Restore debug info generation settings.
9563 2018-03-01 Martin Liska <mliska@suse.cz>
9566 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9569 2018-03-01 Richard Biener <rguenther@suse.de>
9572 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9573 type annotation with LTO when debug was disabled at compile-time.
9575 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9577 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9579 (mips_final_postscan_insn): Likewise.
9581 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9583 PR rtl-optimization/84528
9584 * alias.c (init_alias_target): Add commentary.
9585 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9586 a unique base value if the frame pointer is not eliminated
9587 to the stack pointer.
9589 2018-03-01 Tom de Vries <tom@codesourcery.com>
9591 PR rtl-optimization/83327
9592 * lra-int.h (hard_regs_spilled_into): Declare.
9593 * lra.c (hard_regs_spilled_into): Define.
9594 (init_reg_info): Init hard_regs_spilled_into.
9595 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9596 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9597 (process_bb_lives): Handle hard_regs_spilled_into.
9598 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9599 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9601 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9603 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9604 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9605 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9606 * config/rs6000/aix72.h: New file.
9608 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9610 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9611 instead of warning_at with conditional singular and plural messages
9615 * stor-layout.c (update_alignment_for_field): For
9616 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9617 && !DECL_PACKED (field), do the alignment update, just use
9618 only desired_align instead of MAX (type_align, desired_align)
9620 (place_field): Don't do known_align < desired_align handling
9621 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9622 is non-NULL, instead do it after rli->prev_field handling and
9623 only if not within a bitfield word. For DECL_PACKED (field)
9624 use type_align of BITS_PER_UNIT.
9626 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9628 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9629 superfluous parentheses and trailing spaces.
9631 2018-02-28 Richard Biener <rguenther@suse.de>
9633 PR tree-optimization/84584
9634 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9635 SCoPs with fake exit edge.
9637 2018-02-28 Martin Liska <mliska@suse.cz>
9640 * timevar.c (timer::print): Fix format to properly print 100%
9643 2018-02-28 Richard Biener <rguenther@suse.de>
9646 * genmatch.c (capture_info::walk_match): Do not mark
9647 captured expressions without operands as expr_p given
9648 they act more like predicates and should be subject to
9649 "lost tail" side-effect preserving.
9651 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9653 PR rtl-optimization/81611
9654 * auto-inc-dec.c (attempt_change): Move dead note from
9655 mem_insn if it's the next use of regno
9656 (find_address): Take address use of reg holding
9657 non-incremented value. Add parm to limit search to the named
9659 (merge_in_block): Attempt to use a mem insn that is the next
9660 use of the original regno.
9662 2018-02-27 Martin Sebor <msebor@redhat.com>
9665 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9666 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9668 2018-02-27 Martin Sebor <msebor@redhat.com>
9670 PR translation/84207
9671 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9672 n argument to unsigned HOST_WIDE_INT.
9673 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9674 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9675 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9676 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9678 2018-02-27 Richard Biener <rguenther@suse.de>
9680 PR tree-optimization/84512
9681 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9682 Do not use the estimate returned from record_stmt_cost for
9683 the scalar iteration cost but sum properly using add_stmt_cost.
9685 2018-02-27 Richard Biener <rguenther@suse.de>
9687 PR tree-optimization/84466
9688 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9689 Adjust last change to less strictly validate use operands.
9691 2018-02-27 Martin Liska <mliska@suse.cz>
9693 PR gcov-profile/84548
9694 * gcov.c (process_file): Allow partial overlap and consider it
9695 also as group functions.
9696 (output_lines): Properly calculate range of lines for a group.
9698 2018-02-27 Martin Liska <mliska@suse.cz>
9700 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9701 'ggc' suffixes. Change first column width.
9702 (timer::print): Fix formatting of the column.
9704 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9706 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9707 preserve inline entry blocks for the sake of debug inline
9708 entry point markers alone.
9709 (remove_unused_locals): Suggest in comments a better place to
9710 force the preservation of inline entry blocks that are
9711 otherwise unused, but do not preserve them.
9713 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9715 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9717 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9720 * config/i386/constraints.md (Bs): Replace
9721 ix86_indirect_branch_register with
9722 TARGET_INDIRECT_BRANCH_REGISTER.
9724 * config/i386/i386.md (indirect_jump): Likewise.
9725 (tablejump): Likewise.
9726 (*sibcall_memory): Likewise.
9727 (*sibcall_value_memory): Likewise.
9728 Peepholes of indirect call and jump via memory: Likewise.
9729 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9730 (*sibcall_value_GOT_32): Likewise.
9731 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9732 (GOT_memory_operand): Likewise.
9733 (call_insn_operand): Likewise.
9734 (sibcall_insn_operand): Likewise.
9735 (GOT32_symbol_operand): Likewise.
9736 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9738 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9740 PR rtl-optimization/83496
9741 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9742 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9743 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9744 redundant insn, if any.
9745 (relax_delay_slots): Likewise.
9746 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9748 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9750 PR tree-optimization/83965
9751 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9752 that grouped statements are part of a reduction chain. Return
9753 true if the statement is not marked as a reduction itself but
9755 (vect_recog_dot_prod_pattern): Don't check whether the statement
9756 is part of a group here.
9757 (vect_recog_sad_pattern): Likewise.
9758 (vect_recog_widen_sum_pattern): Likewise.
9760 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9763 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9765 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9768 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9770 (ix86_output_indirect_function_return): New prototype.
9771 (ix86_split_simple_return_pop_internal): Likewise.
9772 * config/i386/i386.c (indirect_return_via_cx): New.
9773 (indirect_return_via_cx_bnd): Likewise.
9774 (indirect_thunk_name): Handle return va CX_REG.
9775 (output_indirect_thunk_function): Create alias for
9776 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9777 (ix86_output_indirect_jmp): Remove the bool argument.
9778 (ix86_output_indirect_function_return): New function.
9779 (ix86_split_simple_return_pop_internal): Likewise.
9780 * config/i386/i386.md (*indirect_jump): Don't pass false
9781 to ix86_output_indirect_jmp.
9782 (*tablejump_1): Likewise.
9783 (simple_return_pop_internal): Change it to define_insn_and_split.
9784 Call ix86_split_simple_return_pop_internal to split it for
9786 (simple_return_indirect_internal): Call
9787 ix86_output_indirect_function_return instead of
9788 ix86_output_indirect_jmp.
9790 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9793 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9794 memset and value initialization afterwards.
9796 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9798 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9800 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9803 * common/config/aarch64/aarch64-common.c
9804 (aarch_option_optimization_table[]): Switch
9805 off fomit-frame-pointer
9807 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9808 Chung-Ju Wu <jasonwucj@gmail.com>
9810 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9812 (store_multiple): Ditto.
9814 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9816 * config.gcc: Add --with-cpu support for nds32 target.
9817 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9818 * config/nds32/nds32.opt: Add -mcpu= option.
9820 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9822 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9823 isel=yes): Warn for these deprecated options.
9825 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9827 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9828 ISA_2_5_MASKS_EMBEDDED.
9830 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9832 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9833 p->max as pointers rather than using iterative_hash_expr.
9835 2018-02-23 Carl Love <cel@us.ibm.com>
9837 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9838 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9840 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9841 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9842 P8V_BUILTIN_VEC_VUNSIGNED2.
9844 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9847 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9848 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9850 * lra-constraints.c (curr_insn_transform): Set up
9851 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9853 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9854 * lra-eliminations.c (spill_pseudos): Ditto.
9855 (process_insn_for_elimination): Ditto.
9856 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9857 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9858 LRA_NON_CLOBBERED_ALT.
9860 2018-02-22 Martin Sebor <msebor@redhat.com>
9862 PR tree-optimization/84480
9863 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9864 to maybe_diag_stxncpy_trunc. Call it.
9865 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9866 from gimple_fold_builtin_strcpy. Print inlining stack.
9867 (handle_builtin_stxncpy): Print inlining stack.
9868 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9870 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9873 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9874 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9875 and -fcheck-pointer-bounds are used together.
9876 (indirect_thunk_prefix): New enum.
9877 (indirect_thunk_need_prefix): New function.
9878 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9879 "_nt" instead of "_bnd" for NOTRACK prefix.
9880 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9881 (output_indirect_thunk_function): Likewise.
9883 (ix86_code_end): Update output_indirect_thunk_function calls.
9884 (ix86_output_indirect_branch_via_reg): Replace
9885 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9886 (ix86_output_indirect_branch_via_push): Likewise.
9887 (ix86_output_function_return): Likewise.
9888 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9889 incompatible with -fcf-protection=branch and
9890 -fcheck-pointer-bounds.
9892 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9895 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9896 Change gcc_assert call to output_operand_lossage.
9898 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9900 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9902 2018-02-22 DJ Delorie <dj@redhat.com>
9903 Sebastian Perta <sebastian.perta@renesas.com>
9904 Oleg Endo <olegendo@gcc.gnu.org>
9906 * config/rx/rx.c (rx_rtx_costs): New function.
9907 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9909 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9911 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9913 2018-02-22 Martin Liska <mliska@suse.cz>
9916 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9917 Add "native" as a possible value.
9919 2018-02-22 Martin Liska <mliska@suse.cz>
9922 * config/i386/i386.c (ix86_option_override_internal):
9923 Add "native" as a possible value for -march and -mtune.
9925 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9928 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9929 to all type variants.
9931 PR tree-optimization/84503
9932 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9933 width as info->bitpos + info->bitsize - start.
9934 (merged_store_group::merge_overlapping): Simplify width computation.
9935 (check_no_overlap): New function.
9936 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9937 start + width and last_order of the group, fail if check_no_overlap
9939 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9940 to group if check_no_overlap fails.
9942 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9944 * config/rs6000/altivec.md: Delete contraint arguments to
9945 define_expand, define_split, and define_peephole2, and in
9946 define_insn_and_split if always unused.
9947 * config/rs6000/darwin.md: Ditto.
9948 * config/rs6000/dfp.md: Ditto.
9949 * config/rs6000/rs6000.md: Ditto.
9950 * config/rs6000/sync.md: Ditto.
9951 * config/rs6000/vector.md: Ditto.
9952 * config/rs6000/vsx.md: Ditto.
9954 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9956 * config/rs6000/altivec.md: Write output control strings as braced
9957 blocks instead of double-quoted strings.
9958 * config/rs6000/darwin.md: Ditto.
9959 * config/rs6000/rs6000.md: Ditto.
9960 * config/rs6000/vector.md: Ditto.
9961 * config/rs6000/vsx.md: Ditto.
9963 2018-02-21 Jason Merrill <jason@redhat.com>
9965 PR c++/84314 - ICE with templates and fastcall attribute.
9966 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9968 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9970 * ipa-cp.c (determine_versionability): Fix comment typos.
9972 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9975 * ipa-cp.c (determine_versionability): Do not version functions caling
9978 2018-02-21 Martin Liska <mliska@suse.cz>
9981 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9982 Add "native" as a possible value.
9983 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9984 the macro when native cpu detection is available.
9986 2018-02-21 Martin Liska <mliska@suse.cz>
9989 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9990 Add "native" as a possible value.
9991 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9992 when native cpu detection is available.
9994 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9995 Martin Sebor <msebor@redhat.com>
9997 PR tree-optimization/84478
9998 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10000 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10001 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10002 support which is conservatively correct, for 2 only stay conservative
10003 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10004 argument to the 2 argument get_range_strlen, adjust 6 arg
10005 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10007 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10008 (gimple_fold_builtin_strlen): Pass true as last argument to
10011 2018-02-20 Martin Sebor <msebor@redhat.com>
10013 PR middle-end/84095
10014 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10015 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10016 (builtin_memref::builtin_memref): Factor out parts into
10017 set_base_and_offset and call it.
10019 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10021 PR middle-end/84406
10022 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10023 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10024 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10025 search at the associated MODE_INT.
10027 2018-02-20 Jeff Law <law@redhat.com>
10029 PR middle-end/82123
10030 PR tree-optimization/81592
10031 PR middle-end/79257
10032 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10033 for range data rather than using global data.
10034 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10035 range data rather than using global data.
10036 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10037 pass it to children as needed.
10038 (struct directive::fmtresult): Similarly.
10039 (struct directive::set_width): Similarly.
10040 (struct directive::set_precision): Similarly.
10041 (format_integer, format_directive, parse_directive): Similarly.
10042 (format_none): Accept unnamed vr_values parameter.
10043 (format_percent, format_floating, format_character): Similarly.
10044 (format_string, format_plain): Similarly.
10045 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10046 the EVRP range analyzer for range data rather than using global data.
10047 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10048 gimple-ssa-evrp-analyze.h
10049 (class sprintf_dom_walker): Add after_dom_children member function.
10050 Add evrp_range_analyzer member.
10051 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10052 range analyzer as needed.
10053 (sprintf_dom_walker::after_dom_children): New member function.
10054 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10056 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10057 (evrp_range_analyzer::pop_to_marker): Likewise.
10059 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10061 PR tree-optimization/84419
10062 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10063 with the required type if its current type is compatible but
10066 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10068 PR middle-end/82004
10069 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10070 after vectorization.
10072 2018-02-20 Martin Liska <mliska@suse.cz>
10075 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10076 possible values if we don't have a hint.
10078 2018-02-20 Martin Liska <mliska@suse.cz>
10082 * final.c (shorten_branches): Build align_tab array with one
10084 * opts.c (finish_options): Add alignment option limit check.
10085 (MAX_CODE_ALIGN): Likewise.
10086 (MAX_CODE_ALIGN_VALUE): Likewise.
10087 * doc/invoke.texi: Document maximum allowed option value for
10088 all -falign-* options.
10090 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10093 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10094 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10095 * var-tracking.c (emit_note_insn_var_location): Remove all references
10096 to NOTE_INSN_CALL_ARG_LOCATION.
10097 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10098 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10099 Use copy_rtx_if_shared.
10100 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10101 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10102 (dwarf2out_var_location): Remove handling of
10103 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10105 * final.c (final_scan_insn): Remove all references to
10106 NOTE_INSN_CALL_ARG_LOCATION.
10107 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10108 before dumping final insns.
10109 * except.c (emit_note_eh_region_end): Remove all references to
10110 NOTE_INSN_CALL_ARG_LOCATION.
10111 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10112 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10113 * config/arc/arc.c (hwloop_optimize): Likewise.
10114 * config/arm/arm.c (create_fix_barrier): Likewise.
10115 * config/s390/s390.c (s390_chunkify_start): Likewise.
10116 * config/sh/sh.c (find_barrier): Likewise.
10117 * config/i386/i386.c (rest_of_insert_endbranch,
10118 ix86_seh_fixup_eh_fallthru): Likewise.
10119 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10120 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10121 * config/frv/frv.c (frv_function_prologue): Likewise.
10122 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10124 (note_outside_basic_block_p): Remove all references to
10125 NOTE_INSN_CALL_ARG_LOCATION.
10126 * gengtype.c (adjust_field_rtx_def): Likewise.
10127 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10129 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10130 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10133 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10136 PR tree-optimization/84452
10137 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10138 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10141 2018-02-19 Martin Liska <mliska@suse.cz>
10144 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10146 2018-02-19 Martin Liska <mliska@suse.cz>
10147 Richard Sandiford <richard.sandiford@linaro.org>
10149 PR tree-optimization/82491
10150 * gimple-fold.c (get_base_constructor): Make earlier bail out
10153 2018-02-19 Carl Love <cel@us.ibm.com>
10155 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10156 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10158 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10159 P8V_BUILTIN_VEC_NEG.
10161 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10163 * config/rl78/rl78.md (movdf): New define expand.
10165 2018-02-19 Martin Liska <mliska@suse.cz>
10168 * doc/invoke.texi: Fix typo.
10169 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10171 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10173 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10174 handle rs6000_single_float and rs6000_double_float specially for
10177 2018-02-16 Jeff Law <law@redhat.com>
10179 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10180 (pop_regs): Use it.
10182 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10185 * ipa-inline.c (inline_small_functions): Fix a typo.
10187 2018-02-16 Nathan Sidwell <nathan@acm.org>
10189 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10191 2018-02-16 Carl Love <cel@us.ibm.com>
10193 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10194 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10195 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10196 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10197 expansion to P8V_BUILTIN_VEC_FLOAT2.
10199 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10201 PR rtl-optimization/70023
10202 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10203 src_regno into account.
10205 2018-02-16 Carl Love <cel@us.ibm.com>
10207 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10208 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10209 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10210 * config/rs6000/rs6000.c: Remove case statements for
10211 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10212 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10213 and P9V_BUILTIN_VEC_VINSERT4B.
10214 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10215 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10216 * config/rs6000/vsx.md:
10217 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10220 2018-02-16 Carl Love <cel@us.ibm.com>
10222 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10224 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10226 * config/rs6000/rs6000-c.c: Add the definitions for
10227 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10228 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10229 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10230 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10231 definition for insert4b and define insn *insert3b_internal.
10232 * doc/extend.texi: Add documentation for vec_extract4b.
10234 2018-02-16 Nathan Sidwell <nathan@acm.org>
10236 * doc/extend.texi (Backwards Compatibility): Mention friend
10237 injection. Note for-scope is deprecated.
10238 * doc/invoke.texi (-ffriend-injection): Deprecate.
10240 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10242 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10243 that moved to I2, also allow destinations that are a paradoxical
10244 subreg (instead of a normal reg).
10246 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10249 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10252 2018-02-16 Richard Biener <rguenther@suse.de>
10254 PR tree-optimization/84037
10255 PR tree-optimization/84016
10257 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10258 Adjust vec_construct for the fact we need additional higher latency
10259 128bit inserts for AVX256 and AVX512 vector builds.
10260 (ix86_add_stmt_cost): Scale vector construction cost for
10263 2018-02-16 Richard Biener <rguenther@suse.de>
10265 PR tree-optimization/84417
10266 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10267 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10268 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10270 2018-02-16 Martin Liska <mliska@suse.cz>
10273 * internal-fn.def (ASAN_CHECK): Set proper flags.
10274 (ASAN_MARK): Likewise.
10276 2018-02-16 Julia Koval <julia.koval@intel.com>
10278 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10279 from PTA_CANNONLAKE.
10281 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10284 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10285 Use ++iter rather than iter++ for std::list iterators.
10286 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10287 defer deleting them until all nodes in the forest are processed. Do
10288 free even leaf nodes. Change to_process into auto_vec.
10291 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10292 * vec.h (vec_default_construct): Use memset instead of placement new
10293 if BROKEN_VALUE_INITIALIZATION is defined.
10294 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10295 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10298 PR rtl-optimization/83723
10299 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10300 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10301 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10303 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10305 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10307 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10309 PR rtl-optimization/81443
10310 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10311 from inner REGs to paradoxical SUBREGs.
10313 2018-02-16 Richard Biener <rguenther@suse.de>
10315 PR tree-optimization/84399
10316 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10317 For operands we can analyze at their definition make sure we can
10318 analyze them at each use as well.
10320 2018-02-16 Richard Biener <rguenther@suse.de>
10322 PR tree-optimization/84190
10323 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10324 volatile accesses if the decl isn't volatile.
10326 2018-02-15 Jason Merrill <jason@redhat.com>
10328 PR c++/84314 - ICE with templates and fastcall attribute.
10329 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10330 TYPE_CANONICAL on an existing type.
10332 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10334 PR tree-optimization/84383
10335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10336 dstoff nor call operand_equal_p if dstbase is NULL.
10338 PR tree-optimization/84334
10339 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10340 also a CONSTANT_CLASS_P, punt.
10342 2018-02-14 Jim Wilson <jimw@sifive.com>
10344 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10345 first SMALL_OPERAND check. New local min_second_step. Move assert
10346 to where locals are set. Add TARGET_RVC support.
10347 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10349 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10351 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10353 2018-02-14 Martin Sebor <msebor@redhat.com>
10355 PR tree-optimization/83698
10356 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10357 arrays constrain the offset range to their bounds.
10358 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10359 (builtin_access::overlap): Avoid setting the size of overlap if it's
10361 (maybe_diag_overlap): Also consider arrays when deciding what values
10362 of offsets to include in diagnostics.
10364 2018-02-14 Martin Sebor <msebor@redhat.com>
10367 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10368 that correspond to the kind of a declaration.
10370 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10373 * config/pa/pa.md: Load address of PIC label using the linkage table
10374 if the label is nonlocal.
10376 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10378 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10379 warning message if user requests -maltivec=be.
10380 * doc/invoke.texi: Document deprecation of -maltivec=be.
10382 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10385 * config/rs6000/rs6000-c.c: Update definitions for
10386 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10387 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10389 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10392 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10393 add _get_ssp intrinsics. Remove argument from
10394 __builtin_ia32_rdssp[d|q].
10395 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10396 * config/i386/i386-builtin.def: Remove argument from
10397 __builtin_ia32_rdssp[d|q].
10398 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10399 ix86_expand_special_args_builtin for _rdssp[d|q].
10400 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10401 Clear register before usage.
10402 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10403 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10405 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10407 PR tree-optimization/84357
10408 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10409 operand 1 of an ARRAY_REF too.
10411 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10414 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10415 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10417 (set_of_reg): New struct.
10418 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10419 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10420 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10422 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10423 Split into bitclr, bitset, bitinvert patterns if appropriate.
10424 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10425 use rx_fuse_in_memory_bitop.
10426 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10427 to named insn, correct maximum insn length.
10429 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10432 * machmode.def: Define a complex mode for PARTIAL_INT.
10433 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10435 * doc/rtl.texi: Document CSPImode.
10436 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10438 (msp430_hard_regno_nregs_with_padding): Likewise.
10440 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10443 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10445 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10447 PR rtl-optimization/84169
10448 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10449 we generated a parallel as new i3 and we split that to new i2 and i3
10450 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10451 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10452 those to i2, not i1. Partially rewrite this scan code.
10454 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10457 * stor-layout.c (place_field): For variable length fields, adjust
10458 offset_align afterwards not just based on the field's alignment,
10459 but also on the size.
10461 PR middle-end/84309
10462 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10463 of exps and logs in the use_exp2 case.
10465 2018-02-13 Jeff Law <law@redhat.com>
10467 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10468 entry for "vector".
10470 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10473 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10477 * common.opt (gas-loc-support, gas-locview-support): New.
10478 (ginline-points, ginternal-reset-location-views): New.
10479 * doc/invoke.texi: Document them. Use @itemx where intended.
10480 (gvariable-location-views): Adjust.
10481 * target.def (reset_location_view): New.
10482 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10483 (TARGET_RESET_LOCATION_VIEW): New.
10484 * doc/tm.texi: Rebuilt.
10485 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10486 (dwarf2out_default_as_locview_support): New.
10487 (output_asm_line_debug_info): Use option variables.
10488 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10489 (output_loc_list): Likewise.
10490 (add_high_low_attributes): Check option variables.
10491 Don't output entry view attribute in strict mode.
10492 (gen_inlined_subroutine_die): Check option variables.
10493 (dwarf2out_inline_entry): Likewise.
10494 (init_sections_and_labels): Likewise.
10495 (dwarf2out_early_finish): Likewise.
10496 (maybe_reset_location_view): New, from...
10497 (dwarf2out_var_location): ... here. Call it.
10498 * debug.h (dwarf2out_default_as_loc_support): Declare.
10499 (dwarf2out_default_as_locview_support): Declare.
10500 * hooks.c (hook_int_rtx_insn_0): New.
10501 * hooks.h (hook_int_rtx_insn_0): Declare.
10502 * toplev.c (process_options): Take -gas-loc-support and
10503 -gas-locview-support from dwarf2out. Enable
10504 -gvariable-location-views by default only with locview
10505 assembler support. Enable -ginternal-reset-location-views by
10506 default only if the target defines the corresponding hook.
10507 Enable -ginline-points by default if location views are
10508 enabled; force it disabled if statement frontiers are
10510 * tree-inline.c (expand_call_inline): Check option variables.
10511 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10513 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10515 PR tree-optimization/84321
10516 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10517 handling. Also check whether the anti-range contains any values
10518 that satisfy the mask; switch to a VR_RANGE if not.
10520 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10523 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10525 2018-02-13 Martin Jambor <mjambor@suse.cz>
10528 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10529 of call statements, also set location of a load to a temporary.
10531 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10533 * config/rl78/rl78.c (add_vector_labels): New function.
10534 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10535 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10536 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10537 which checks that no arguments are passed.
10538 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10539 * doc/extend.texi: Documentation for the new attribute.
10541 2018-02-13 Andreas Schwab <schwab@suse.de>
10543 * config/riscv/linux.h (CPP_SPEC): Define.
10545 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10548 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10549 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10550 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10551 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10552 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10553 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10554 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10555 temporarily for AES and PCLMUL builtins.
10557 PR tree-optimization/84339
10558 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10559 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10562 PR middle-end/84309
10563 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10564 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10565 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10567 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10569 * omp-simd-clone.h: New file.
10570 * omp-simd-clone.c: Include omp-simd-clone.h.
10571 (expand_simd_clones): No longer static.
10572 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10573 cgraph.h and omp-simd-clone.h.
10574 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10575 (vect_recog_widen_shift_pattern): Formatting fix.
10576 (vect_pattern_recog_1): Don't check optab for calls.
10579 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10580 operands[2] into a REG before using gen_lowpart on it.
10582 2018-02-12 Jeff Law <law@redhat.com>
10585 * config/sh/sh.c (find_barrier): Consider a sibling call
10588 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10589 successfully back substituting a reg.
10591 2018-02-12 Richard Biener <rguenther@suse.de>
10593 PR tree-optimization/84037
10594 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10595 parameter, move visited init to caller.
10596 (vect_slp_analyze_operations): Separate cost from validity
10597 check, initialize visited once for all instances.
10598 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10600 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10602 (vect_model_promotion_demotion_cost): Likewise.
10603 (vectorizable_bswap): Guard cost modeling with !slp_node
10604 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10606 (vectorizable_call): Likewise.
10607 (vectorizable_conversion): Likewise.
10608 (vectorizable_assignment): Likewise.
10609 (vectorizable_shift): Likewise.
10610 (vectorizable_operation): Likewise.
10611 (vectorizable_store): Likewise.
10612 (vectorizable_load): Likewise.
10613 (vectorizable_condition): Likewise.
10614 (vectorizable_comparison): Likewise.
10616 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10619 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10620 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10621 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10623 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10625 PR middle-end/83665
10626 * params.def (inline-min-speedup): Increase from 8 to 15.
10627 (max-inline-insns-auto): Decrease from 40 to 30.
10628 * ipa-split.c (consider_split): Add some buffer for function to
10629 be considered inlining candidate.
10630 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10633 2018-02-12 Richard Biener <rguenther@suse.de>
10635 PR tree-optimization/84037
10636 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10637 matched stmts if we cannot swap the non-matched ones.
10639 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10641 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10642 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10643 _mm_maskz_scalef_round_ss): New intrinsics.
10644 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10645 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10646 __builtin_ia32_scalefss_round): Remove.
10647 (__builtin_ia32_scalefsd_mask_round,
10648 __builtin_ia32_scalefss_mask_round): New intrinsics.
10649 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10650 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10651 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10652 "<round_constraint>")): Changed to ...
10653 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10654 "<round_scalar_constraint>")): ... this.
10655 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10656 %0, %1, %2<round_op3>}"): Changed to ...
10657 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10658 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10659 %2<round_scalar_mask_op3>}"): ... this.
10660 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10662 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10664 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10665 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10666 (_mm_maskz_sqrt_round_ss): New intrinsics.
10667 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10668 (__builtin_ia32_sqrtsd_mask_round)
10669 (__builtin_ia32_sqrtss_mask_round): New builtins.
10670 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10671 (__builtin_ia32_sqrtss_round): Remove.
10672 (__builtin_ia32_sqrtsd_mask_round)
10673 (__builtin_ia32_sqrtss_mask_round): New builtins.
10674 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10675 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10676 ((match_operand:VF_128 1 "vector_operand"
10677 "xBm,<round_constraint>")): Changed to ...
10678 ((match_operand:VF_128 1 "vector_operand"
10679 "xBm,<round_scalar_constraint>")): ... this.
10680 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10681 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10682 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10683 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10684 %<iptr>1<round_scalar_mask_op3>}): ... this.
10685 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10686 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10688 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10691 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10692 Cast vec_cmpeq result to correct type.
10693 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10694 Cast vec_cmpgt result to correct type.
10696 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10698 * final.c (final_scan_insn_1): Renamed from...
10699 (final_scan_insn): ... this. New wrapper, to recover
10700 seen from the outermost call in recursive ones.
10701 * config/sparc/sparc.c (output_return): Drop seen from call.
10702 (output_sibcall): Likewise.
10703 * config/visium/visium.c (output_branch): Likewise.
10705 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10707 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10710 2018-02-10 Alan Modra <amodra@gmail.com>
10713 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10714 Specify LR as an input.
10716 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10719 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10720 remove_member_access_dummy_vars): New functions.
10721 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10722 lower_omp_1, execute_lower_omp): Use them.
10724 PR rtl-optimization/84308
10725 * shrink-wrap.c (spread_components): Release todo vector.
10727 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10729 PR rtl-optimization/57193
10730 * ira-color.c (struct allocno_color_data): Add member
10731 conflict_allocno_hard_prefs.
10732 (update_conflict_allocno_hard_prefs): New.
10733 (bucket_allocno_compare_func): Add a preference based on
10734 conflict_allocno_hard_prefs.
10735 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10736 (color_allocnos): Remove a dead code. Initiate
10737 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10739 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10742 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10743 constraint from =wa to wa. Avoid a subreg on the output operand,
10744 instead use a pseudo and subreg it in a move.
10745 (p9_xxbrd_<mode>): Changed to ...
10746 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10747 (p9_xxbrd_v2df): New expander.
10748 (p9_xxbrw_<mode>): Changed to ...
10749 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10750 (p9_xxbrw_v4sf): New expander.
10752 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10754 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10756 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10759 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10760 multiply in 32-bit mode.
10761 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10762 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10765 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10767 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10768 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10769 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10770 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10772 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10775 * dwarf2out.c (is_trivial_indirect_ref): New function.
10776 (dwarf2out_late_global_decl): Do not generate a location
10777 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10778 and that are not defined in the current unit.
10780 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10782 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10783 instead of a libcall for UNORDERED.
10785 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10788 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10789 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10791 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10794 * config/s390/s390.c (s390_set_current_function): Invoke
10795 s390_indirect_branch_settings also if fndecl didn't change.
10797 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10799 * config/rs6000/rs6000.md (blockage): Set length to zero.
10801 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10803 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10805 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10808 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10809 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10813 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10814 PARALLEL incoming that failed vt_get_decl_and_offset check.
10816 PR middle-end/84237
10817 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10818 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10820 (get_variable_section): For decls in named .bss* sections pass true as
10821 second argument to bss_initializer_p.
10823 2018-02-09 Marek Polacek <polacek@redhat.com>
10824 Jakub Jelinek <jakub@redhat.com>
10827 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10828 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10829 Sync some changes from cxx_fold_indirect_ref.
10831 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10833 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10835 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10836 (BLOCK_INLINE_ENTRY_LABEL): New.
10837 (dwarf2out_var_location): Disregard inline entry markers.
10838 (inline_entry_data): New struct.
10839 (inline_entry_data_hasher): New hashtable type.
10840 (inline_entry_data_hasher::hash): New.
10841 (inline_entry_data_hasher::equal): New.
10842 (inline_entry_data_table): New variable.
10843 (add_high_low_attributes): Add DW_AT_entry_pc and
10844 DW_AT_GNU_entry_view attributes if a pending entry is found
10845 in inline_entry_data_table. Add old entry_pc attribute only
10846 if debug nonbinding markers are disabled.
10847 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10848 markers are enabled.
10849 (block_within_block_p, dwarf2out_inline_entry): New.
10850 (dwarf2out_finish): Check that no entries remained in
10851 inline_entry_data_table.
10852 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10853 (final_scan_insn, notice_source_line): Likewise.
10854 (rest_of_clean_state): Skip inline entry markers.
10855 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10857 * gimple.c (gimple_build_debug_inline_entry): New.
10858 * gimple.h (enum gimple_debug_subcode): Add
10859 GIMPLE_DEBUG_INLINE_ENTRY.
10860 (gimple_build_debug_inline_entry): Declare.
10861 (gimple_debug_inline_entry_p): New.
10862 (gimple_debug_nonbind_marker_p): Adjust.
10863 * insn-notes.def (INLINE_ENTRY): New.
10864 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10865 inline entry marker notes.
10866 (print_insn): Likewise.
10867 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10868 (INSN_DEBUG_MARKER_KIND): Likewise.
10869 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10870 * tree-inline.c (expand_call_inline): Build and insert
10871 debug_inline_entry stmt.
10872 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10873 inline entry blocks early, if nonbind markers are enabled.
10874 (dump_scope_block): Dump fragment info.
10875 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10876 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10877 (gimple_build_debug_inline_entry): New.
10878 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10879 Enable/disable inline entry points too.
10880 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10881 (DEBUG_INSN): Describe inline entry markers.
10883 * common.opt (gvariable-location-views): New.
10884 (gvariable-location-views=incompat5): New.
10885 * config.in: Rebuilt.
10886 * configure: Rebuilt.
10887 * configure.ac: Test assembler for view support.
10888 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10889 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10890 * dwarf2out.c (var_loc_view): New typedef.
10891 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10892 (dwarf2out_locviews_in_attribute): New.
10893 (dwarf2out_locviews_in_loclist): New.
10894 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10895 (enum dw_line_info_opcode): Add LI_adv_address.
10896 (struct dw_line_info_table): Add view.
10897 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10898 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10899 (zero_view_p): New variable.
10900 (ZERO_VIEW_P): New macro.
10901 (output_asm_line_debug_info): New.
10902 (struct var_loc_node): Add view.
10903 (add_AT_view_list, AT_loc_list): New.
10904 (add_var_loc_to_decl): Add view param. Test it against last.
10905 (new_loc_list): Add view params. Record them.
10906 (AT_loc_list_ptr): Handle loc and view lists.
10907 (view_list_to_loc_list_val_node): New.
10908 (print_dw_val): Handle dw_val_class_view_list.
10909 (size_of_die): Likewise.
10910 (value_format): Likewise.
10911 (loc_list_has_views): New.
10912 (gen_llsym): Set vl_symbol too.
10913 (maybe_gen_llsym, skip_loc_list_entry): New.
10914 (dwarf2out_maybe_output_loclist_view_pair): New.
10915 (output_loc_list): Output view list or entries too.
10916 (output_view_list_offset): New.
10917 (output_die): Handle dw_val_class_view_list.
10918 (output_dwarf_version): New.
10919 (output_compilation_unit_header): Use it.
10920 (output_skeleton_debug_sections): Likewise.
10921 (output_rnglists, output_line_info): Likewise.
10922 (output_pubnames, output_aranges): Update version comments.
10923 (output_one_line_info_table): Output view numbers in asm comments.
10924 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10925 Call maybe_gen_llsym.
10926 (loc_list_from_tree_1): Adjust.
10927 (add_AT_location_description): Create view list attribute if
10928 needed, check it's absent otherwise.
10929 (convert_cfa_to_fb_loc_list): Adjust.
10930 (maybe_emit_file): Call output_asm_line_debug_info for test.
10931 (dwarf2out_var_location): Reset views as needed. Precompute
10932 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10933 attribute. Set view.
10934 (new_line_info_table): Reset next view.
10935 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10936 (dwarf2out_source_line): Likewise. Output view resets and labels to
10937 the assembler, or select appropriate line info opcodes.
10938 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10939 (optimize_string_length): Catch it. Adjust.
10940 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10941 dw_val_class_view_list, and remove it if no longer needed.
10942 (hash_loc_list): Hash view numbers.
10943 (loc_list_hasher::equal): Compare them.
10944 (optimize_location_lists): Check whether a view list symbol is
10945 needed, and whether the locview attribute is present, and
10946 whether they match. Remove the locview attribute if no longer
10948 (index_location_lists): Call skip_loc_list_entry for test.
10949 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10950 Use output_dwarf_version.
10951 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10952 (struct dw_val_node): Add val_view_list.
10953 * final.c (SEEN_NEXT_VIEW): New.
10954 (set_next_view_needed): New.
10955 (clear_next_view_needed): New.
10956 (maybe_output_next_view): New.
10957 (final_start_function): Rename to...
10958 (final_start_function_1): ... this. Take pointer to FIRST,
10959 add SEEN parameter. Emit param bindings in the initial view.
10960 (final_start_function): Reintroduce SEEN-less interface.
10961 (final): Rename to...
10962 (final_1): ... this. Take SEEN parameter. Output final pending
10963 next view at the end.
10964 (final): Reintroduce seen-less interface.
10965 (final_scan_insn): Output pending next view before switching
10966 sections or ending a block. Mark the next view as needed when
10967 outputting variable locations. Notify debug backend of section
10968 changes, and of location view changes.
10969 (rest_of_handle_final): Adjust.
10970 * toplev.c (process_options): Autodetect value for debug variable
10971 location views option. Warn on incompat5 without -gdwarf-5.
10972 * doc/invoke.texi (gvariable-location-views): New.
10973 (gvariable-location-views=incompat5): New.
10974 (gno-variable-location-views): New.
10976 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10978 PR tree-optimization/84136
10979 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10980 that the result of find_edge is non-NULL.
10982 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10985 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10986 storing integer register in SImode. Fix cost of 256 and 512
10987 byte aligned SSE register store.
10989 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10991 * config/i386/i386.c (ix86_multiplication_cost): Fix
10992 multiplication cost for TARGET_AVX512DQ.
10994 2018-02-08 Marek Polacek <polacek@redhat.com>
10996 PR tree-optimization/84238
10997 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11000 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11002 PR tree-optimization/84265
11003 * tree-vect-stmts.c (vectorizable_store): Don't treat
11004 VMAT_CONTIGUOUS accesses as grouped.
11005 (vectorizable_load): Likewise.
11007 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11009 PR tree-optimization/81635
11010 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11011 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11012 (test_round_for_mask): New functions.
11013 (wide_int_cc_tests): Call test_round_for_mask.
11014 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11015 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11016 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11017 range returned by get_range_info.
11019 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11022 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11023 * symtab.c: Include builtins.h
11024 (symtab_node::output_to_lto_symbol_table_p): Move here
11025 from lto-streamer-out.c:output_symbol_p.
11026 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11027 (output_symbol_p): Move all logic to symtab.c
11028 (produce_symtab): Update.
11030 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11032 * config/s390/s390-opts.h (enum indirect_branch): Define.
11033 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11034 (s390_indirect_branch_via_thunk)
11035 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11036 (enum s390_indirect_branch_type): Define.
11037 * config/s390/s390.c (struct s390_frame_layout, struct
11038 machine_function): Remove.
11039 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11040 (indirect_branch_table_label_no, indirect_branch_table_name):
11042 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11043 (enum s390_indirect_branch_option): Define.
11044 (s390_return_addr_from_memory): New function.
11045 (s390_handle_string_attribute): New function.
11046 (s390_attribute_table): Add new attribute handler.
11047 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11048 (s390_indirect_branch_via_thunk): New function.
11049 (s390_indirect_branch_via_inline_thunk): New function.
11050 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11051 sibling call optimization for non z10 compiles.
11052 (s390_emit_call): Force indirect branch target to be a single
11053 register. Add r1 clobber for non-z10 compiles.
11054 (s390_emit_epilogue): Emit return jump via return_use expander.
11055 (s390_reorg): Handle JUMP_INSNs as execute targets.
11056 (s390_option_override_internal): Perform validity checks for the
11057 new command line options.
11058 (s390_indirect_branch_attrvalue): New function.
11059 (s390_indirect_branch_settings): New function.
11060 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11061 (s390_output_indirect_thunk_function): New function.
11062 (s390_code_end): Implement target hook.
11063 (s390_case_values_threshold): Implement target hook.
11064 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11066 * config/s390/s390.h (struct s390_frame_layout)
11067 (struct machine_function): Move here from s390.c.
11068 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11069 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11070 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11071 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11072 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11073 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11074 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11075 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11076 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11077 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11078 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11079 (mnemonic attribute): Add values which aren't recognized
11081 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11082 pattern for branch conversion. Fix mnemonic attribute.
11083 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11084 indirect branch via thunk if requested.
11085 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11086 ("*indirect_jump"): Disable for branch conversion using out of
11088 ("indirect_jump_via_thunk<mode>_z10")
11089 ("indirect_jump_via_thunk<mode>")
11090 ("indirect_jump_via_inlinethunk<mode>_z10")
11091 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11092 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11093 ("casesi_jump_via_inlinethunk<mode>_z10")
11094 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11095 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11096 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11097 ("*indirect2_jump"): Disable for branch conversion.
11098 ("casesi_jump"): Turn into expander and expand patterns for branch
11100 ("return_use"): New expander.
11101 ("*return"): Emit return via thunk and rename it to ...
11102 ("*return<mode>"): ... this one.
11103 * config/s390/s390.opt: Add new options and and enum for the
11106 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11108 * lra-constraints.c (match_reload): Unconditionally use
11109 gen_lowpart_SUBREG, rather than selecting between that
11110 and equivalent gen_rtx_SUBREG code.
11112 2018-02-08 Richard Biener <rguenther@suse.de>
11114 PR tree-optimization/84233
11115 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11116 changed flag instead of boguously re-using phi_inserted.
11118 2018-02-08 Martin Jambor <mjambor@suse.cz>
11120 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11121 static local variables.
11123 2018-02-08 Richard Biener <rguenther@suse.de>
11125 PR tree-optimization/84278
11126 * tree-vect-stmts.c (vectorizable_store): When looking for
11127 smaller vector types to perform grouped strided loads/stores
11128 make sure the mode is supported by the target.
11129 (vectorizable_load): Likewise.
11131 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11133 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11134 Increase LDP/STP opportunities by adding adjacent callee-saves.
11136 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11138 PR rtl-optimization/84068
11139 PR rtl-optimization/83459
11140 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11142 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11144 PR tree-optimization/84224
11145 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11146 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11147 non-zero arguments.
11149 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11152 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11153 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11154 count, remove one USE.
11156 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11158 * doc/install.texi (Configuration): Document the
11159 --with-long-double-format={ibm,ieee} PowerPC configuration
11163 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11164 Convert from define_expand to be define_insn_and_split. Rework
11165 float/double/_Float128 conversions to QI/HI/SImode to work with
11166 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11167 conversions to QI/HImode types did a store and then a load to
11168 truncate the value. For conversions to VSX registers, don't split
11169 the insn, instead emit the code directly. Use the code iterator
11170 any_fix to combine signed and unsigned conversions.
11171 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11172 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11173 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11174 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11175 (fix_<mode>di2_hw): Likewise.
11176 (fixuns_<mode>di2_hw): Likewise.
11177 (fix_<mode>si2_hw): Likewise.
11178 (fixuns_<mode>si2_hw): Likewise.
11179 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11180 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11181 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11182 fix<uns>_trunc<SFDF:mode>si2_p8.
11183 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11184 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11185 (fix<uns>_<mode>_mem): Likewise.
11186 (fctiw<u>z_<mode>_mem): Likewise.
11187 (fix<uns>_<mode>_mem): Likewise.
11188 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11189 the register allocator from doing a direct move to the GPRs to do
11190 a store, and instead use the ISA 3.0 store byte/half-word from
11191 vector register instruction. For IEEE 128-bit floating point,
11192 also optimize stores of 32-bit ints.
11193 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11195 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11197 * genextract.c (push_pathstr_operand): New function to support
11199 (walk_rtx): Call push_pathstr_operand.
11200 (print_path): Support [a-zA-Z].
11202 2018-02-07 Richard Biener <rguenther@suse.de>
11204 PR tree-optimization/84037
11205 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11206 (cse_and_gimplify_to_preheader): Declare.
11207 (vect_get_place_in_interleaving_chain): Likewise.
11208 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11210 (_loop_vec_info::~_loop_vec_info): Delete it.
11211 (cse_and_gimplify_to_preheader): New function.
11212 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11213 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11214 (vectorizable_load): Likewise. For grouped stores always base
11215 the IV on the first element.
11216 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11217 condition before gimplifying.
11219 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11221 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11222 *DIV_EXPR and *MOD_EXPR.
11224 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11227 * config/i386/i386.c (ix86_option_override_internal): Mask out
11228 the CF_SET bit when checking -fcf-protection.
11230 2018-02-07 Tom de Vries <tom@codesourcery.com>
11233 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11236 2018-02-07 Richard Biener <rguenther@suse.de>
11238 PR tree-optimization/84204
11239 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11242 PR tree-optimization/84205
11243 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11244 special-case isl_ast_op_zdiv_r.
11246 PR tree-optimization/84223
11247 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11248 Only add conditions from within the region.
11249 (gather_bbs::after_dom_children): Adjust.
11251 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11254 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11255 * config/avr/avr.md: Only post-reload split REG-REG moves if
11256 either register is GENERAL_REG_P.
11258 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11260 PR tree-optimization/84235
11261 * tree-ssa-scopedtables.c
11262 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11263 if the subtraction is performed in floating point type where NaNs are
11264 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11265 build 1. Formatting fix.
11267 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11270 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11271 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11272 and skip it regardless of bb boundaries. Use CALL_P macro,
11273 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11274 unnecessarily, formatting fix.
11276 2018-02-06 Michael Collison <michael.collison@arm.com>
11278 * config/arm/thumb2.md:
11279 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11280 (*thumb_mov_notscc): Ditto.
11282 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11285 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11286 unsigned_fix, not "s".
11288 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11290 * configure.ac (gcc_fn_eh_frame_ro): New function.
11291 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11292 correct .eh_frame permissions.
11293 * configure: Regenerate.
11295 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11297 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11298 irrelevant options.
11300 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11302 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11303 Display warning message for -mno-speculate-indirect-jumps.
11305 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11307 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11309 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11311 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11313 PR tree-optimization/84225
11314 * tree-eh.c (find_trapping_overflow): Only call
11315 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11317 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11320 * config/i386/i386.c: Reimplement the check of possible options
11321 -mibt/-mshstk conbination. Change error messages.
11322 * doc/invoke.texi: Fix a typo: remove extra '='.
11324 2018-02-06 Marek Polacek <polacek@redhat.com>
11326 PR tree-optimization/84228
11327 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11329 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11332 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11333 emitted arch directives.
11334 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11335 __ARM_FEATURE_COPROC before changing architectures.
11337 2018-02-06 Richard Biener <rguenther@suse.de>
11339 * config/i386/i386.c (print_reg): Fix typo.
11340 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11342 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11344 * configure: Regenerate.
11346 2018-02-05 Martin Sebor <msebor@redhat.com>
11348 PR tree-optimization/83369
11349 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11352 2018-02-05 Martin Liska <mliska@suse.cz>
11354 * doc/invoke.texi: Cherry-pick upstream r323995.
11356 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11358 * ira.c (ira_init_register_move_cost): Adjust comment.
11360 2018-02-05 Martin Liska <mliska@suse.cz>
11362 PR gcov-profile/84137
11363 * doc/gcov.texi: Fix typo in documentation.
11365 2018-02-05 Martin Liska <mliska@suse.cz>
11367 PR gcov-profile/83879
11368 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11369 using dlopen functionality.
11371 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11373 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11374 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11375 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11376 _mm_maskz_range_round_ss): New intrinsics.
11377 (__builtin_ia32_rangesd128_round)
11378 (__builtin_ia32_rangess128_round): Remove.
11379 (__builtin_ia32_rangesd128_mask_round,
11380 __builtin_ia32_rangess128_mask_round): New builtins.
11381 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11382 __builtin_ia32_rangess128_round): Remove.
11383 (__builtin_ia32_rangesd128_mask_round,
11384 __builtin_ia32_rangess128_mask_round): New builtins.
11385 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11386 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11387 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11388 "<round_saeonly_constraint>")): Changed to ...
11389 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11390 "<round_saeonly_scalar_constraint>")): ... this.
11391 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11392 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11393 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11394 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11395 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11397 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11399 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11401 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11402 Remove all values except native, 8540 and 8548.
11404 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11406 * config/i386/i386.c (ix86_output_function_return): Pass
11407 INVALID_REGNUM, instead of -1, as invalid register number to
11408 indirect_thunk_name and output_indirect_thunk.
11410 2018-02-02 Julia Koval <julia.koval@intel.com>
11412 * config.gcc: Add -march=icelake.
11413 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11414 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11415 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11416 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11417 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11418 (processor_target_table): Add icelake.
11419 (ix86_option_override_internal): Handle new PTAs.
11420 (get_builtin_code_for_version): Handle icelake.
11421 (M_INTEL_COREI7_ICELAKE): New.
11422 (fold_builtin_cpu): Handle icelake.
11423 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11424 * doc/invoke.texi: Add -march=icelake.
11426 2018-02-02 Julia Koval <julia.koval@intel.com>
11428 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11429 to wide_int_bitmask.
11430 * wide-int-bitmask.h: New.
11432 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11435 * config/i386/i386.md: Replace Pmode with word_mode in
11436 builtin_setjmp_setup and builtin_longjmp to support x32.
11438 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11442 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11444 (rs6000_supported_cpu_names): New static variable.
11445 (linux_cpu_translation_table): Likewise.
11446 (elf_platform) <cpu>: Define new static variable and use it.
11447 Translate kernel AT_PLATFORM name to canonical name if needed.
11448 Error if platform name is unknown.
11450 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11453 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11455 2018-02-01 Jeff Law <law@redhat.com>
11458 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11459 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11460 the scratch if RELEASE_VIA_POP is false.
11461 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11462 If we have to save a temporary register, decrement SIZE appropriately.
11463 Pass new arguments to release_scratch_register_on_entry.
11464 (ix86_adjust_stack_and_probe): Likewise.
11465 (ix86_emit_probe_stack_range): Pass new arguments to
11466 release_scratch_register_on_entry.
11468 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11470 PR rtl-optimization/84157
11471 * combine.c (change_zero_ext): Use REG_P predicate in
11472 front of HARD_REGISTER_P predicate.
11474 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11476 * config/avr/avr.c (avr_option_override): Move disabling of
11477 -fdelete-null-pointer-checks to...
11478 * common/config/avr/avr-common.c (avr_option_optimization_table):
11481 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11483 PR tree-optimization/81635
11484 * tree-data-ref.c (split_constant_offset_1): For types that
11485 wrap on overflow, try to use range info to prove that wrapping
11488 2018-02-01 Renlin Li <renlin.li@arm.com>
11491 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11492 TAILCALL_ADDR_REGS.
11493 (aarch64_register_move_cost): Likewise.
11494 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11495 TAILCALL_ADDR_REGS.
11496 (REG_CLASS_NAMES): Likewise.
11497 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11498 TAILCALL_ADDR_REGS. Remove IP registers.
11499 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11501 2018-02-01 Richard Biener <rguenther@suse.de>
11503 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11504 for specifying RPO order and allow NULL for that.
11505 * domwalk.c (dom_walker::dom_walker): Likewise.
11506 (dom_walker::walk): Handle NULL RPO order.
11507 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11509 (rewrite_update_dom_walker): Likewise.
11510 (mark_def_dom_walker): Likewise.
11512 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11514 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11515 (aarch64_maybe_expand_sve_subreg_move): Declare.
11516 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11517 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11519 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11520 that are semantically a reverse operation.
11521 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11522 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11523 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11525 (aarch64_can_change_mode_class): For big-endian, forbid changes
11526 between two SVE modes if they have different element sizes.
11528 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11530 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11531 the TImode handling for big-endian targets.
11533 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11535 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11536 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11538 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11539 Remove BSWAP handing for big-endian targets and use the form of
11540 LD1RQ appropariate for the mode.
11542 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11544 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11545 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11546 duplicated element.
11548 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11551 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11552 check for operands that need to go through aarch64_sve_reload_be.
11554 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11556 PR tree-optimization/81661
11557 PR tree-optimization/84117
11558 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11559 * tree-eh.c: Include gimplify.h.
11560 (find_trapping_overflow, replace_trapping_overflow,
11561 rewrite_to_non_trapping_overflow): New functions.
11562 * tree-vect-loop.c: Include tree-eh.h.
11563 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11564 * tree-data-ref.c: Include tree-eh.h.
11565 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11567 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11569 PR rtl-optimization/84123
11570 * combine.c (change_zero_ext): Check if hard register satisfies
11571 can_change_dest_mode before calling gen_lowpart_SUBREG.
11573 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11576 * ira.c (ira_init_register_move_cost): Remove assert.
11578 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11580 PR rtl-optimization/84071
11581 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11582 * doc/tm.texi: Regenerate.
11584 2018-01-31 Richard Biener <rguenther@suse.de>
11586 PR tree-optimization/84132
11587 * tree-data-ref.c (analyze_miv_subscript): Properly
11588 check whether evolution_function_is_affine_multivariate_p
11589 before calling gcd_of_steps_may_divide_p.
11591 2018-01-31 Julia Koval <julia.koval@intel.com>
11594 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11595 * config/i386/i386.md (rdpid_rex64) New.
11596 (rdpid): Make 32bit only.
11598 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11601 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11602 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11604 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11607 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11609 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11611 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11613 PR rtl-optimization/84071
11614 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11615 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11617 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11619 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11620 (arc_attribute_table): Add 'aux' attribute.
11621 (arc_in_small_data_p): Consider aux like variables.
11622 (arc_is_aux_reg_p): New function.
11623 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11624 (arc_get_aux_arg): New function.
11625 (prepare_move_operands): Handle aux-register access.
11626 (arc_handle_aux_attribute): New function.
11627 * doc/extend.texi (ARC Variable attributes): Add subsection.
11629 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11631 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11632 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11633 (arc_attribute_table): Add 'uncached' attribute.
11634 (arc_print_operand): Print '.di' flag for uncached memory
11636 (arc_in_small_data_p): Do not consider for small data the uncached
11638 (arc_is_uncached_mem_p): New function.
11639 * config/arc/predicates.md (compact_store_memory_operand): Check
11640 for uncached memory accesses.
11641 (nonvol_nonimm_operand): Likewise.
11642 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11644 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11647 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11648 falign-loops=): Add Optimization flag.
11650 2018-01-30 Jeff Law <law@redhat.com>
11653 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11654 INT_REGISTERS_SAVED. Check it prior to calling
11655 get_scratch_register_on_entry.
11656 (ix86_adjust_stack_and_probe): Similarly.
11657 (ix86_emit_probe_stack_range): Similarly.
11658 (ix86_expand_prologue): Corresponding changes.
11660 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11663 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11664 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11666 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11669 * lra-constraints.c (curr_insn_transform): Process AND in the
11672 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11674 PR rtl-optimization/83986
11675 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11676 dependence against last_pending_memory_flush in addition to
11677 pending_jump_insns.
11679 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11681 PR tree-optimization/81611
11682 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11685 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11688 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11691 2018-01-30 Richard Biener <rguenther@suse.de>
11692 Jakub Jelinek <jakub@redhat.com>
11694 PR tree-optimization/84111
11695 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11696 inner loops added during recursion, as they don't have up-to-date
11699 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11702 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11703 (can_inline_edge_by_limits_p): ... here.
11704 (can_early_inline_edge_p, check_callers,
11705 update_caller_keys, update_callee_keys, recursive_inlining,
11706 add_new_edges_to_heap, speculation_useful_p,
11707 inline_small_functions,
11708 inline_small_functions, flatten_function,
11709 inline_to_all_callers_1): Update.
11711 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11713 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11716 2018-01-30 Richard Biener <rguenther@suse.de>
11718 PR tree-optimization/83008
11719 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11720 invariant and constant vector uses in stmts when they need
11721 more than one stmt.
11723 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11726 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11727 * configure: Regenerate.
11729 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11731 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11733 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11734 Use gen_rtx_REG rather than gen_lowpart.
11736 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11738 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11739 rather than 0 when creating partial subregs.
11741 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11743 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11746 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11749 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11750 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11751 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11752 flags. This restores the settings used before the 2017-07-24.
11753 Turning off pre increment/decrement/modify allows IVOPTS to
11754 optimize DF/SF loops where the index is an int.
11756 2018-01-29 Richard Biener <rguenther@suse.de>
11757 Kelvin Nilsen <kelvin@gcc.gnu.org>
11760 * tree-vect-stmts.c (vectorizable_call): Don't call
11761 targetm.vectorize_builtin_md_vectorized_function if callee is
11764 2018-01-22 Carl Love <cel@us.ibm.com>
11766 * doc/extend.tex: Fix typo in second arg in
11767 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11769 2018-01-29 Richard Biener <rguenther@suse.de>
11771 PR tree-optimization/84086
11772 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11773 (flush_ssaname_freelist): When SSA names were released reset
11774 the SCEV hash table.
11776 2018-01-29 Richard Biener <rguenther@suse.de>
11778 PR tree-optimization/84057
11779 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11780 removed paths when removing edges.
11782 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11784 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11785 -mfunction-return=@var{choice}.
11787 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11789 PR diagnostic/84034
11790 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11791 Handle CR like TAB.
11792 (layout::print_source_line): Likewise.
11793 (test_get_line_width_without_trailing_whitespace): Add test cases.
11795 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11797 PR middle-end/84040
11798 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11801 2018-01-26 Jim Wilson <jimw@sifive.com>
11803 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11805 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11808 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11810 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11811 and CMP + SUB-immediate -> SUBS.
11813 2018-01-26 Martin Sebor <msebor@redhat.com>
11815 PR tree-optimization/83896
11816 * tree-ssa-strlen.c (get_string_len): Rename...
11817 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11818 Avoid assuming length is constant.
11819 (handle_char_store): Use HOST_WIDE_INT for string length.
11821 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11824 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11825 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11827 2018-01-26 Richard Biener <rguenther@suse.de>
11829 PR rtl-optimization/84003
11830 * dse.c (record_store): Only record redundant stores when
11831 the earlier store aliases at least all accesses the later one does.
11833 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11835 PR rtl-optimization/83985
11836 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11837 REG_CFA_RESTORE insns.
11838 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11839 REG_CFA_RESTORE insns here.
11842 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11843 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11845 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11847 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11848 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11849 (arc_init): Likewise.
11850 (arc_override_options): Likewise.
11851 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11853 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11855 * config/arc/arc.h (TARGET_DBNZ): Define.
11856 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11857 properly set the tune attribute.
11858 (dbnz): Use TARGET_DBNZ guard.
11859 * config/arc/arc.opt (mtune): Add core3 option.
11861 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11863 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11864 recognize new pic like addresses.
11865 (arc_delegitimize_address): Clean up.
11867 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11869 * config/arc/arc-arches.def: Option mrf16 valid for all
11871 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11872 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11873 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11874 * config/arc/arc-tables.opt: Regenerate.
11875 * config/arc/arc.c (arc_conditional_register_usage): Handle
11876 reduced register file case.
11877 (arc_file_start): Set must have build attributes.
11878 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11879 mrf16 option value.
11880 * config/arc/arc.opt (mrf16): Add new option.
11881 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11882 * config/arc/genmultilib.awk: Handle new mrf16 option.
11883 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11884 * config/arc/t-multilib: Regenerate.
11885 * doc/invoke.texi (ARC Options): Document mrf16 option.
11887 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11889 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11890 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11891 (arc_attribute_table): Add 'secure_call' attribute.
11892 (arc_print_operand): Print secure call operand.
11893 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11895 (arc_is_secure_call_p): New function. * config/arc/arc.md
11896 (call_i): Add support for sjli instruction.
11897 (call_value_i): Likewise.
11898 * config/arc/constraints.md (Csc): New constraint.
11900 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11901 John Eric Martin <John.Martin@emmicro-us.com>
11903 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11904 * config/arc/arc.c (_arc_jli_section): New struct.
11905 (arc_jli_section): New type.
11906 (rc_jli_sections): New static variable.
11907 (arc_handle_jli_attribute): New function.
11908 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11909 (arc_file_end): New function.
11910 (TARGET_ASM_FILE_END): Define.
11911 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11912 (arc_add_jli_section): New function.
11913 (jli_call_scan): Likewise.
11914 (arc_reorg): Call jli_call_scan.
11915 (arc_output_addsi): Remove 'S' from printing asm operand.
11916 (arc_is_jli_call_p): New function.
11917 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11919 (movhi_insn): Likewise.
11920 (movsi_insn): Likewise.
11921 (movsi_set_cc_insn): Likewise.
11922 (loadqi_update): Likewise.
11923 (load_zeroextendqisi_update): Likewise.
11924 (load_signextendqisi_update): Likewise.
11925 (loadhi_update): Likewise.
11926 (load_zeroextendhisi_update): Likewise.
11927 (load_signextendhisi_update): Likewise.
11928 (loadsi_update): Likewise.
11929 (loadsf_update): Likewise.
11930 (movsicc_insn): Likewise.
11931 (bset_insn): Likewise.
11932 (bxor_insn): Likewise.
11933 (bclr_insn): Likewise.
11934 (bmsk_insn): Likewise.
11935 (bicsi3_insn): Likewise.
11936 (cmpsi_cc_c_insn): Likewise.
11937 (movsi_ne): Likewise.
11938 (movsi_cond_exec): Likewise.
11939 (clrsbsi2): Likewise.
11940 (norm_f): Likewise.
11950 (call_i): Remove 'S' asm letter, add jli instruction.
11951 (call_value_i): Likewise.
11952 * config/arc/arc.op (mjli-always): New option.
11953 * config/arc/constraints.md (Cji): New constraint.
11954 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11956 (subsf3_fpx): Likewise.
11957 (mulsf3_fpx): Likewise.
11958 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11960 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11961 function attrbutes.
11962 * doc/invoke.texi (ARC): Document mjli-always option.
11964 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11966 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11967 avoid addition with 0 and use incw and decw where possible.
11969 2018-01-26 Richard Biener <rguenther@suse.de>
11971 PR tree-optimization/81082
11972 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11973 association if it requires casting to unsigned.
11974 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11975 from fold_plusminus_mult_expr to catch important cases late when
11976 range info is available.
11978 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11980 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11981 * configure.ac (hidden_linkonce): New test.
11982 * configure: Regenerate.
11983 * config.in: Regenerate.
11985 2018-01-26 Julia Koval <julia.koval@intel.com>
11987 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11988 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11989 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11990 _mm_mask_bitshuffle_epi64_mask): Fix type.
11991 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11992 USI_FTYPE_V4DI_V4DI_USI): Remove.
11993 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11994 __builtin_ia32_vpshufbitqmb256_mask,
11995 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11996 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11997 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11999 2018-01-26 Alan Modra <amodra@gmail.com>
12002 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12003 UNSPEC_VBPERMQ. Sort other unspecs.
12005 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12007 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12009 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12011 PR middle-end/83055
12012 * predict.c (drop_profile): Do not push/pop cfun; update also
12014 (handle_missing_profiles): Fix logic looking for zero profiles.
12016 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12018 PR middle-end/83977
12019 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12020 on functions with #pragma omp declare simd or functions with simd
12022 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12023 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12024 Remove trailing \n from warning_at calls.
12026 2018-01-25 Tom de Vries <tom@codesourcery.com>
12029 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12030 for neutered workers.
12032 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12035 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12036 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12038 2018-01-24 Jeff Law <law@redhat.com>
12041 * i386.c (get_probe_interval): Move to earlier point.
12042 (ix86_compute_frame_layout): If -fstack-clash-protection and
12043 the frame is larger than the probe interval, then use pushes
12044 to save registers rather than reg->mem moves.
12045 (ix86_expand_prologue): Remove conditional for int_registers_saved
12048 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12051 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12052 min/max for never referenced object.
12054 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12056 PR middle-end/83977
12057 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12059 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12060 attributes from DECL_ATTRIBUTES (decl) without affecting
12061 DECL_ATTRIBUTES (current_function_decl).
12062 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12063 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12065 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12067 PR tree-optimization/83979
12068 * fold-const.c (fold_comparison): Use constant_boolean_node
12069 instead of boolean_{true,false}_node.
12071 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12073 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12076 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12078 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12079 Simplify the clause that sets the length attribute.
12080 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12081 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12082 clause that sets the length attribute.
12083 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12085 2018-01-24 Tom de Vries <tom@codesourcery.com>
12088 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12089 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12090 Add strict parameter.
12091 (prevent_branch_around_nothing): Insert dummy insn between branch to
12092 label and label with no ptx insn inbetween.
12093 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12095 2018-01-24 Tom de Vries <tom@codesourcery.com>
12098 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12099 for neutered threads in warp.
12100 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12102 2018-01-24 Richard Biener <rguenther@suse.de>
12104 PR tree-optimization/83176
12105 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12108 2018-01-24 Richard Biener <rguenther@suse.de>
12110 PR tree-optimization/82819
12111 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12112 code generating pluses that are no-ops in the target precision.
12114 2018-01-24 Richard Biener <rguenther@suse.de>
12116 PR middle-end/84000
12117 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12119 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12121 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12122 to merge probabilities.
12123 * predict.c (probably_never_executed): Also mark as cold functions
12124 with global 0 profile and guessed local profile.
12125 * profile-count.c (profile_probability::combine_with_count): New
12127 * profile-count.h (profile_probability::operator*,
12128 profile_probability::operator*=, profile_probability::operator/,
12129 profile_probability::operator/=): Reduce precision to adjusted
12130 and set value to guessed on contradictory divisions.
12131 (profile_probability::combine_with_freq): Remove.
12132 (profile_probability::combine_wiht_count): Declare.
12133 (profile_count::force_nonzero):: Set to adjusted.
12134 (profile_count::probability_in):: Set quality to adjusted.
12135 * tree-ssa-tail-merge.c (replace_block_by): Use
12136 combine_with_count.
12138 2018-01-23 Andrew Waterman <andrew@sifive.com>
12139 Jim Wilson <jimw@sifive.com>
12141 * config/riscv/riscv.c (riscv_stack_boundary): New.
12142 (riscv_option_override): Set riscv_stack_boundary. Handle
12143 riscv_preferred_stack_boundary_arg.
12144 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12145 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12146 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12147 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12148 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12149 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12151 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12154 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12155 of struct ix86_frame.
12156 (ix86_expand_epilogue): Likewise. Add a local variable for
12157 the reg_save_offset field in struct ix86_frame.
12159 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12161 PR tree-optimization/82604
12162 * tree-loop-distribution.c (enum partition_kind): New enum item
12163 PKIND_PARTIAL_MEMSET.
12164 (partition_builtin_p): Support above new enum item.
12165 (generate_code_for_partition): Ditto.
12166 (compute_access_range): Differentiate cases that equality can be
12167 proven at all loops, the innermost loops or no loops.
12168 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12169 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12170 (finalize_partitions, distribute_loop): Don't fuse partition of
12171 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12172 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12173 parloop is enabled.
12175 2018-01-23 Martin Liska <mliska@suse.cz>
12177 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12178 order to ignore the predictor.
12179 (PRED_POLYMORPHIC_CALL): Likewise.
12180 (PRED_RECURSIVE_CALL): Likewise.
12182 2018-01-23 Martin Liska <mliska@suse.cz>
12184 * tree-profile.c (tree_profiling): Print function header to
12185 aware reader which function we are working on.
12186 * value-prof.c (gimple_find_values_to_profile): Do not print
12187 not interesting value histograms.
12189 2018-01-23 Martin Liska <mliska@suse.cz>
12191 * profile-count.h (enum profile_quality): Add
12192 profile_uninitialized as the first value. Do not number values
12193 as they are zero based.
12194 (profile_count::verify): Update sanity check.
12195 (profile_probability::verify): Likewise.
12197 2018-01-23 Nathan Sidwell <nathan@acm.org>
12199 * doc/invoke.texi (ffor-scope): Deprecate.
12201 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12203 PR tree-optimization/83510
12204 * domwalk.c (set_all_edges_as_executable): New function.
12205 (dom_walker::dom_walker): Convert bool param
12206 "skip_unreachable_blocks" to enum reachability. Move setup of
12207 edge flags to set_all_edges_as_executable and only do it when
12208 reachability is REACHABLE_BLOCKS.
12209 * domwalk.h (enum dom_walker::reachability): New enum.
12210 (dom_walker::dom_walker): Convert bool param
12211 "skip_unreachable_blocks" to enum reachability.
12212 (set_all_edges_as_executable): New decl.
12213 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12214 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12216 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12217 but converting true to REACHABLE_BLOCKS.
12218 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12220 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12221 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12222 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12224 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12225 if check_all_array_refs will be called.
12227 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12229 * tree.c (selftest::test_location_wrappers): Add more test
12232 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12234 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12235 (selftest::test_bit_in_range): Likewise.
12237 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12240 * doc/sourcebuild.texi (vect_float): Say that the selector
12241 only describes the situation when -funsafe-math-optimizations is on.
12242 (vect_float_strict): Document.
12244 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12246 PR tree-optimization/83965
12247 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12248 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12249 instead of checking only for a reduction.
12250 (vect_recog_widen_sum_pattern): Likewise.
12252 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12254 * predict.c (probably_never_executed): Only use precise profile info.
12255 (compute_function_frequency): Skip after inlining hack since we now
12256 have quality checking.
12258 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12260 * profile-count.h (profile_probability::very_unlikely,
12261 profile_probability::unlikely, profile_probability::even): Set
12262 precision to guessed.
12264 2018-01-23 Richard Biener <rguenther@suse.de>
12266 PR tree-optimization/83963
12267 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12268 Properly terminate dominator walk when crossing the exit edge not
12269 when visiting its source block.
12271 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12274 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12275 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12277 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12279 PR tree-optimization/83957
12280 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12281 semicolon after for body surrounded by braces.
12283 PR tree-optimization/83081
12284 * profile-count.h (profile_probability::split): New method.
12285 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12286 Use profile_probability::split.
12287 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12288 when splitting a single conditional jump into 2.
12290 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12292 PR tree-optimization/69452
12293 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12296 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12298 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12299 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12300 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12302 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12304 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12306 * config/rl78/rl78.md (movdi): New define_expand.
12307 * config/rl78/rl78.c (rl78_split_movdi): New function.
12309 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12312 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12314 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12315 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12316 128-bit to produce an UNSPEC move to get the double word with the
12317 signbit and then a shift directly to do signbit.
12318 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12319 implementation with a new version that just does either a direct
12320 move or a regular move. Move memory interface to separate insns.
12321 Move insns so they are next to the expander.
12322 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12323 with signbit move. Split big and little endian case.
12324 (signbit<mode>2_dm_mem_le): Likewise.
12325 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12326 (signbit<mode>2_dm2): Likewise.
12328 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12330 * config/rl78/rl78.md (anddi3): New define_expand.
12332 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12334 * config/rl78/rl78.md (umindi3): New define_expand.
12336 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12338 * config/rl78/rl78.md (smindi3): New define_expand.
12340 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12342 * config/rl78/rl78.md (smaxdi3): New define_expand.
12344 2018-01-22 Carl Love <cel@us.ibm.com>
12346 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12347 LVX_V1TI): Add macro expansion.
12348 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12349 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12350 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12351 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12352 Change check to determine if the instruction is a byte reversing
12353 entry. Fix typo in comment.
12354 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12355 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12356 Add def_builtin calls for new builtins.
12357 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12358 Add define_insn expansion.
12360 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12362 * config/rl78/rl78.md (umaxdi3): New define_expand.
12364 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12366 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12367 for non-QImode registers.
12369 2018-01-22 Richard Biener <rguenther@suse.de>
12371 PR tree-optimization/83963
12372 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12373 including the loop exit block.
12374 (scop_detection::merge_sese): Likewise.
12375 (scop_detection::add_scop): Do it here instead.
12377 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12379 * doc/sourcebuild.texi (arm_softfloat): Document.
12381 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12384 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12385 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12386 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12388 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12389 David Edelsohn <dje.gcc@gmail.com>
12392 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12393 Change "crset eq" to "crset 2".
12394 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12395 (*call_indirect_aix<mode>_nospec): Likewise.
12396 (*call_value_indirect_aix<mode>_nospec): Likewise.
12397 (*call_indirect_elfv2<mode>_nospec): Likewise.
12398 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12399 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12400 change assembly output from . to $.
12401 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12402 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12403 (*tablejump<mode>_internal1_nospec): Likewise.
12405 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12408 * config/sh/sh_optimize_sett_clrt.cc:
12409 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12411 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12413 PR tree-optimization/83940
12414 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12415 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12416 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12417 use it to pass back the definition type.
12418 (vect_check_store_rhs): Likewise rhs_dt_out.
12419 (vect_build_gather_load_calls): Add a mask_dt argument and use
12420 it instead of a call to vect_is_simple_use.
12421 (vectorizable_store): Update calls to vect_check_load_store_mask
12422 and vect_check_store_rhs. Use the dt returned by the latter instead
12423 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12424 instead of calls to vect_is_simple_use. Pass the scalar rather
12425 than the vector operand to vect_is_simple_use when handling
12426 second and subsequent copies of an rhs value.
12427 (vectorizable_load): Update calls to vect_check_load_store_mask
12428 and vect_build_gather_load_calls. Use the cached mask_dt and
12429 gs_info.offset_dt instead of calls to vect_is_simple_use.
12431 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12433 PR middle-end/83945
12434 * tree-emutls.c: Include gimplify.h.
12435 (lower_emutls_2): New function.
12436 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12437 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12438 it before further processing.
12441 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12442 UINTVAL (trueop1) instead of INTVAL (op1).
12444 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12448 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12449 INCOMING_FRAME_SP_OFFSET if not defined.
12450 (scan_trace): Add ENTRY argument. If true and
12451 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12452 emit a note to adjust the CFA offset.
12453 (create_cfi_notes): Adjust scan_trace callers.
12454 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12455 INCOMING_FRAME_SP_OFFSET in the CIE.
12456 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12457 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12459 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12460 * doc/tm.texi: Regenerated.
12462 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12464 PR rtl-optimization/83147
12465 * lra-constraints.c (remove_inheritance_pseudos): Use
12466 lra_substitute_pseudo_within_insn.
12468 2018-01-19 Tom de Vries <tom@codesourcery.com>
12469 Cesar Philippidis <cesar@codesourcery.com>
12472 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12474 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12477 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12478 spaces for function labels.
12480 2018-01-19 Martin Liska <mliska@suse.cz>
12482 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12483 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12484 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12485 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12486 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12487 (PRED_CONST_RETURN): Change from 69 to 65.
12488 (PRED_NULL_RETURN): Change from 91 to 71.
12489 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12490 (PRED_LOOP_GUARD): Change from 66 to 73.
12492 2018-01-19 Martin Liska <mliska@suse.cz>
12494 * predict.c (predict_insn_def): Add new assert.
12495 (struct branch_predictor): Change type to signed integer.
12496 (test_prediction_value_range): Amend test to cover
12497 PROB_UNINITIALIZED.
12498 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12499 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12500 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12501 (PRED_LOOP_IV_COMPARE): Likewise.
12502 * predict.h (PROB_UNINITIALIZED): Define new constant.
12504 2018-01-19 Martin Liska <mliska@suse.cz>
12506 * predict.c (dump_prediction): Add new format for
12507 analyze_brprob.py script which is enabled with -details
12509 * profile-count.h (precise_p): New function.
12511 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12513 PR tree-optimization/83922
12514 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12515 there are no statements that need masking.
12516 (vect_active_double_reduction_p): New function.
12517 (vect_analyze_loop_operations): Use it when handling phis that
12518 are not in the loop header.
12520 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12522 PR tree-optimization/83914
12523 * tree-vect-loop.c (vectorizable_induction): Don't convert
12524 init_expr or apply the peeling adjustment for inductions
12525 that are nested within the vectorized loop.
12527 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12529 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12532 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12536 * function.h (gimplify_parameters): Add gimple_seq * argument.
12537 * function.c: Include gimple.h and options.h.
12538 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12539 for the added local temporaries if needed.
12540 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12541 if there are any parameter cleanups, wrap whole body into a
12542 try/finally with the cleanups.
12544 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12547 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12548 Use GET_MODE_CLASS for scalar floating point.
12550 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12554 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12555 Fix call of call_cgraph_insertion_hooks.
12557 2018-01-18 Martin Sebor <msebor@redhat.com>
12559 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12561 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12564 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12567 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12570 * common.opt: (-ffile-prefix-map): New option.
12571 * opts.c (common_handle_option): Defer it.
12572 * opts-global.c (handle_common_deferred_options): Handle it.
12573 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12574 * file-prefix-map.h: New file.
12575 (remap_debug_filename, add_debug_prefix_map): ...here.
12576 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12577 * final.c (debug_prefix_map, add_debug_prefix_map
12578 remap_debug_filename): Move to...
12579 * file-prefix-map.c: New file.
12580 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12581 generalize, get rid of alloca(), use strrchr() instead of strchr().
12582 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12583 Implement in terms of add_prefix_map().
12584 (remap_macro_filename, remap_debug_filename): Implement in term of
12586 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12587 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12588 * dbxout.c: Include file-prefix-map.h.
12589 * varasm.c: Likewise.
12590 * vmsdbgout.c: Likewise.
12591 * xcoffout.c: Likewise.
12592 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12593 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12594 * doc/invoke.texi (-ffile-prefix-map): Document.
12595 (-fdebug-prefix-map): Update description.
12597 2018-01-18 Martin Liska <mliska@suse.cz>
12599 * config/i386/i386.c (indirect_thunk_name): Document that also
12601 (output_indirect_thunk): Document why both instructions
12602 (pause and lfence) are generated.
12604 2018-01-18 Richard Biener <rguenther@suse.de>
12606 PR tree-optimization/83887
12607 * graphite-scop-detection.c
12608 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12609 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12610 (scop_detection::merge_sese): Re-implement with a flood-fill
12611 algorithm that properly finds a SESE region if it exists.
12613 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12616 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12617 pointer_diff optimizations use view_convert instead of convert.
12619 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12621 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12622 Generate different code for -mno-speculate-indirect-jumps.
12623 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12624 (*call_indirect_aix<mode>): Disable for
12625 -mno-speculate-indirect-jumps.
12626 (*call_indirect_aix<mode>_nospec): New define_insn.
12627 (*call_value_indirect_aix<mode>): Disable for
12628 -mno-speculate-indirect-jumps.
12629 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12630 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12631 -mno-speculate-indirect-jumps.
12632 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12634 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12636 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12637 long double type, set the flags for noting the default long double
12638 type, even if we don't pass or return a long double type.
12640 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12643 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12646 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12648 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12649 support for merge[hl].
12650 (fold_mergehl_helper): New helper function.
12651 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12652 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12653 (altivec_vmrglw_direct): Add xxmrglw insn.
12655 2018-01-17 Andrew Waterman <andrew@sifive.com>
12657 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12658 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12660 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12663 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12664 call the lto_location_cache before reading the
12665 DECL_SOURCE_LOCATION of the types.
12667 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12668 Richard Sandiford <richard.sandiford@linaro.org>
12670 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12671 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12672 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12673 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12674 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12676 * config/aarch64/constraints.md (aarch64_movti_operand):
12678 * config/aarch64/predicates.md (Uti): Add new constraint.
12680 2018-01-17 Carl Love <cel@us.ibm.com>
12682 * config/rs6000/vsx.md (define_expand xl_len_r,
12683 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12684 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12686 (define_expand, define_insn): Move the shift left from the
12687 define_insn to the define_expand for lxvl and stxvl instructions.
12688 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12689 and XL_LEN_R definitions to PURE.
12691 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12693 * config/i386/i386.c (indirect_thunk_name): Declare regno
12694 as unsigned int. Compare regno with INVALID_REGNUM.
12695 (output_indirect_thunk): Ditto.
12696 (output_indirect_thunk_function): Ditto.
12697 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12698 in the call to output_indirect_thunk_function.
12700 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12702 PR middle-end/83884
12703 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12704 rather than the size of inner_type to determine the stack slot size
12705 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12707 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12710 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12713 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12715 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12716 endian Linux systems to optionally enable multilibs for selecting
12717 the long double type if the user configured an explicit type.
12718 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12719 have no long double multilibs if not defined.
12720 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12721 warn if the user used -mabi={ieee,ibm}longdouble and we built
12722 multilibs for long double.
12723 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12724 appropriate multilib option.
12725 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12727 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12728 for building long double multilibs.
12729 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12731 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12733 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12736 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12738 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12741 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12744 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12747 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12749 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12750 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12752 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12754 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12755 different rtl trees depending on TARGET_64BIT.
12756 (rs6000_gen_lvx): Likewise.
12758 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12760 * config/visium/visium.md (nop): Tweak comment.
12761 (hazard_nop): Likewise.
12763 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12765 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12766 -mspeculate-indirect-jumps.
12767 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12768 for -mno-speculate-indirect-jumps.
12769 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12770 (*call_value_indirect_elfv2<mode>): Disable for
12771 -mno-speculate-indirect-jumps.
12772 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12773 (indirect_jump): Emit different RTL for
12774 -mno-speculate-indirect-jumps.
12775 (*indirect_jump<mode>): Disable for
12776 -mno-speculate-indirect-jumps.
12777 (*indirect_jump<mode>_nospec): New define_insn.
12778 (tablejump): Emit different RTL for
12779 -mno-speculate-indirect-jumps.
12780 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12781 (tablejumpsi_nospec): New define_expand.
12782 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12783 (tablejumpdi_nospec): New define_expand.
12784 (*tablejump<mode>_internal1): Disable for
12785 -mno-speculate-indirect-jumps.
12786 (*tablejump<mode>_internal1_nospec): New define_insn.
12787 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12790 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12792 * caller-save.c (insert_save): Drop unnecessary parameter. All
12795 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12796 Richard Biener <rguenth@suse.de>
12799 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12800 return early, inline manually is_gimple_sizepos. Make sure if we
12801 call gimplify_expr we don't end up with a gimple constant.
12802 * tree.c (variably_modified_type_p): Don't return true for
12803 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12804 * gimplify.h (is_gimple_sizepos): Remove.
12806 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12808 PR tree-optimization/83857
12809 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12810 vectorizable_live_operation for pure SLP statements.
12811 (vectorizable_live_operation): Handle PHIs.
12813 2018-01-16 Richard Biener <rguenther@suse.de>
12815 PR tree-optimization/83867
12816 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12817 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12819 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12822 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12823 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12824 If off is not INTEGER_CST, issue a may not be aligned warning
12825 rather than isn't aligned. Use isn%'t rather than isn't.
12826 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12828 <case MULT_EXPR>: Improve the case when bottom and one of the
12829 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12830 operand, in that case check if the other operand is multiple of
12831 bottom divided by the INTEGER_CST operand.
12833 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12836 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12837 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12838 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12839 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12840 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12841 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12842 (pa_function_arg_size): New function.
12844 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12846 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12847 in a separate statement.
12849 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12851 PR tree-optimization/83847
12852 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12853 group gathers and scatters.
12855 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12857 PR rtl-optimization/86620
12858 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12860 PR rtl-optimization/83213
12861 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12862 to last if both are JUMP_INSNs.
12864 PR tree-optimization/83843
12865 * gimple-ssa-store-merging.c
12866 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12867 store_immediate_info for bswap/nop orig_stores.
12869 2018-01-15 Andrew Waterman <andrew@sifive.com>
12871 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12873 <UDIV>: Increase cost if !TARGET_DIV.
12875 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12877 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12878 (define_attr "cr_logical_3op"): New.
12879 (cceq_ior_compare): Adjust.
12880 (cceq_ior_compare_complement): Adjust.
12881 (*cceq_rev_compare): Adjust.
12882 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12883 (is_cracked_insn): Adjust.
12884 (insn_must_be_first_in_group): Adjust.
12885 * config/rs6000/40x.md: Adjust.
12886 * config/rs6000/440.md: Adjust.
12887 * config/rs6000/476.md: Adjust.
12888 * config/rs6000/601.md: Adjust.
12889 * config/rs6000/603.md: Adjust.
12890 * config/rs6000/6xx.md: Adjust.
12891 * config/rs6000/7450.md: Adjust.
12892 * config/rs6000/7xx.md: Adjust.
12893 * config/rs6000/8540.md: Adjust.
12894 * config/rs6000/cell.md: Adjust.
12895 * config/rs6000/e300c2c3.md: Adjust.
12896 * config/rs6000/e500mc.md: Adjust.
12897 * config/rs6000/e500mc64.md: Adjust.
12898 * config/rs6000/e5500.md: Adjust.
12899 * config/rs6000/e6500.md: Adjust.
12900 * config/rs6000/mpc.md: Adjust.
12901 * config/rs6000/power4.md: Adjust.
12902 * config/rs6000/power5.md: Adjust.
12903 * config/rs6000/power6.md: Adjust.
12904 * config/rs6000/power7.md: Adjust.
12905 * config/rs6000/power8.md: Adjust.
12906 * config/rs6000/power9.md: Adjust.
12907 * config/rs6000/rs64.md: Adjust.
12908 * config/rs6000/titan.md: Adjust.
12910 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12912 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12913 ix86_indirect_branch_register logic.
12915 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12917 * config/i386/constraints.md (Bs): Update
12918 ix86_indirect_branch_register check. Don't check
12919 ix86_indirect_branch_register with GOT_memory_operand.
12921 * config/i386/predicates.md (GOT_memory_operand): Don't check
12922 ix86_indirect_branch_register here.
12923 (GOT32_symbol_operand): Likewise.
12925 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12927 * config/i386/predicates.md (constant_call_address_operand):
12928 Rewrite ix86_indirect_branch_register logic.
12929 (sibcall_insn_operand): Likewise.
12931 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12933 * config/i386/constraints.md (Bs): Replace
12934 ix86_indirect_branch_thunk_register with
12935 ix86_indirect_branch_register.
12937 * config/i386/i386.md (indirect_jump): Likewise.
12938 (tablejump): Likewise.
12939 (*sibcall_memory): Likewise.
12940 (*sibcall_value_memory): Likewise.
12941 Peepholes of indirect call and jump via memory: Likewise.
12942 * config/i386/i386.opt: Likewise.
12943 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12944 (GOT_memory_operand): Likewise.
12945 (call_insn_operand): Likewise.
12946 (sibcall_insn_operand): Likewise.
12947 (GOT32_symbol_operand): Likewise.
12949 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12951 PR middle-end/83837
12952 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12953 type rather than type addr's type points to.
12954 (expand_omp_atomic_mutex): Likewise.
12955 (expand_omp_atomic): Likewise.
12957 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12960 * config/i386/i386.c (output_indirect_thunk_function): Use
12961 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12962 for __x86_return_thunk.
12964 2018-01-15 Richard Biener <rguenther@suse.de>
12966 PR middle-end/83850
12967 * expmed.c (extract_bit_field_1): Fix typo.
12969 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12972 * config/arm/iterators.md (VF): New mode iterator.
12973 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12974 Remove integer-related logic from pattern.
12975 (neon_vabd<mode>_3): Likewise.
12977 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12979 PR middle-end/82694
12980 * common.opt (fstrict-overflow): No longer an alias.
12981 (fwrapv-pointer): New option.
12982 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12983 also for pointer types based on flag_wrapv_pointer.
12984 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12985 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12986 opts->x_flag_wrapv got set.
12987 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12988 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12989 POINTER_TYPE_OVERFLOW_UNDEFINED.
12990 * match.pd: Likewise in address comparison pattern.
12991 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12993 2018-01-15 Richard Biener <rguenther@suse.de>
12996 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12997 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12998 Reset type names to their identifier if their TYPE_DECL doesn't
12999 have linkage (and thus is used for ODR and devirt).
13000 (save_debug_info_for_decl): Remove.
13001 (save_debug_info_for_type): Likewise.
13002 (add_tree_to_fld_list): Adjust.
13003 * tree-pretty-print.c (dump_generic_node): Make dumping of
13004 type names more robust.
13006 2018-01-15 Richard Biener <rguenther@suse.de>
13008 * BASE-VER: Bump to 8.0.1.
13010 2018-01-14 Martin Sebor <msebor@redhat.com>
13013 * builtins.c (check_access): Avoid warning when the no-warning bit
13016 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13018 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13019 * ira-color (allocno_hard_regs_compare): Likewise.
13021 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13024 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13025 Use .pushsection/.popsection.
13027 2018-01-14 Martin Sebor <msebor@redhat.com>
13030 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13032 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13034 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13035 entry from extra_headers.
13036 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13037 extra_headers, make the list bitwise identical to the i?86-*-* one.
13039 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13041 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13042 -mcmodel=large with -mindirect-branch=thunk,
13043 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13044 -mfunction-return=thunk-extern.
13045 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13046 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13047 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13049 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13051 * config/i386/i386.c (print_reg): Print the name of the full
13052 integer register without '%'.
13053 (ix86_print_operand): Handle 'V'.
13054 * doc/extend.texi: Document 'V' modifier.
13056 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13058 * config/i386/constraints.md (Bs): Disallow memory operand for
13059 -mindirect-branch-register.
13061 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13062 (GOT_memory_operand): Likewise.
13063 (call_insn_operand): Likewise.
13064 (sibcall_insn_operand): Likewise.
13065 (GOT32_symbol_operand): Likewise.
13066 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13067 for -mindirect-branch-register.
13068 (tablejump): Likewise.
13069 (*sibcall_memory): Likewise.
13070 (*sibcall_value_memory): Likewise.
13071 Disallow peepholes of indirect call and jump via memory for
13072 -mindirect-branch-register.
13073 (*call_pop): Replace m with Bw.
13074 (*call_value_pop): Likewise.
13075 (*sibcall_pop_memory): Replace m with Bs.
13076 * config/i386/i386.opt (mindirect-branch-register): New option.
13077 * doc/invoke.texi: Document -mindirect-branch-register option.
13079 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13081 * config/i386/i386-protos.h (ix86_output_function_return): New.
13082 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13083 set function_return_type.
13084 (indirect_thunk_name): Add ret_p to indicate thunk for function
13086 (output_indirect_thunk_function): Pass false to
13087 indirect_thunk_name.
13088 (ix86_output_indirect_branch_via_reg): Likewise.
13089 (ix86_output_indirect_branch_via_push): Likewise.
13090 (output_indirect_thunk_function): Create alias for function
13091 return thunk if regno < 0.
13092 (ix86_output_function_return): New function.
13093 (ix86_handle_fndecl_attribute): Handle function_return.
13094 (ix86_attribute_table): Add function_return.
13095 * config/i386/i386.h (machine_function): Add
13096 function_return_type.
13097 * config/i386/i386.md (simple_return_internal): Use
13098 ix86_output_function_return.
13099 (simple_return_internal_long): Likewise.
13100 * config/i386/i386.opt (mfunction-return=): New option.
13101 (indirect_branch): Mention -mfunction-return=.
13102 * doc/extend.texi: Document function_return function attribute.
13103 * doc/invoke.texi: Document -mfunction-return= option.
13105 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13107 * config/i386/i386-opts.h (indirect_branch): New.
13108 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13109 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13110 with local indirect jump when converting indirect call and jump.
13111 (ix86_set_indirect_branch_type): New.
13112 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13113 (indirectlabelno): New.
13114 (indirect_thunk_needed): Likewise.
13115 (indirect_thunk_bnd_needed): Likewise.
13116 (indirect_thunks_used): Likewise.
13117 (indirect_thunks_bnd_used): Likewise.
13118 (INDIRECT_LABEL): Likewise.
13119 (indirect_thunk_name): Likewise.
13120 (output_indirect_thunk): Likewise.
13121 (output_indirect_thunk_function): Likewise.
13122 (ix86_output_indirect_branch_via_reg): Likewise.
13123 (ix86_output_indirect_branch_via_push): Likewise.
13124 (ix86_output_indirect_branch): Likewise.
13125 (ix86_output_indirect_jmp): Likewise.
13126 (ix86_code_end): Call output_indirect_thunk_function if needed.
13127 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13129 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13130 (ix86_attribute_table): Add indirect_branch.
13131 * config/i386/i386.h (machine_function): Add indirect_branch_type
13132 and has_local_indirect_jump.
13133 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13135 (tablejump): Likewise.
13136 (*indirect_jump): Use ix86_output_indirect_jmp.
13137 (*tablejump_1): Likewise.
13138 (simple_return_indirect_internal): Likewise.
13139 * config/i386/i386.opt (mindirect-branch=): New option.
13140 (indirect_branch): New.
13143 (thunk-inline): Likewise.
13144 (thunk-extern): Likewise.
13145 * doc/extend.texi: Document indirect_branch function attribute.
13146 * doc/invoke.texi: Document -mindirect-branch= option.
13148 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13151 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13153 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13155 * ipa-inline.c (want_inline_small_function_p): Return false if
13156 inlining has already failed with CIF_FINAL_ERROR.
13157 (update_caller_keys): Call want_inline_small_function_p before
13159 (update_callee_keys): Likewise.
13161 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13163 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13165 (rs6000_quadword_masked_address_p): Likewise.
13166 (quad_aligned_load_p): Likewise.
13167 (quad_aligned_store_p): Likewise.
13168 (const_load_sequence_p): Add comment to describe the outer-most loop.
13169 (mimic_memory_attributes_and_flags): New function.
13170 (rs6000_gen_stvx): Likewise.
13171 (replace_swapped_aligned_store): Likewise.
13172 (rs6000_gen_lvx): Likewise.
13173 (replace_swapped_aligned_load): Likewise.
13174 (replace_swapped_load_constant): Capitalize argument name in
13175 comment describing this function.
13176 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13177 and stores that access quad-word aligned addresses and replace
13178 with stvx or lvx instructions when appropriate.
13179 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13180 New function prototype.
13181 (rs6000_quadword_masked_address_p): Likewise.
13182 (rs6000_gen_lvx): Likewise.
13183 (rs6000_gen_stvx): Likewise.
13184 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13185 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13186 when memory address is aligned.
13187 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13188 this split to select lvx instruction when memory address is aligned.
13189 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13190 instruction when memory address is aligned.
13191 (*vsx_le_perm_load_v16qi): Likewise.
13192 (four unnamed splitters): Modify to select the stvx instruction
13193 when memory is aligned.
13195 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13197 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13198 which appears in the queue multiple times.
13200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13201 Alan Hayward <alan.hayward@arm.com>
13202 David Sherwood <david.sherwood@arm.com>
13204 * tree-vectorizer.h (vec_lower_bound): New structure.
13205 (_loop_vec_info): Add check_nonzero and lower_bounds.
13206 (LOOP_VINFO_CHECK_NONZERO): New macro.
13207 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13208 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13209 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13210 fields. Make seg_len the distance travelled, not including the
13212 (dr_direction_indicator): Declare.
13213 (dr_zero_step_indicator): Likewise.
13214 (dr_known_forward_stride_p): Likewise.
13215 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13217 (runtime_alias_check_p): Allow runtime alias checks with
13219 (operator ==): Compare access_size and align.
13220 (prune_runtime_alias_test_list): Rework for new distinction between
13221 the access_size and seg_len.
13222 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13224 (get_segment_min_max): New function.
13225 (create_intersect_range_checks): Use it.
13226 (dr_step_indicator): New function.
13227 (dr_direction_indicator): Likewise.
13228 (dr_zero_step_indicator): Likewise.
13229 (dr_known_forward_stride_p): Likewise.
13230 * tree-loop-distribution.c (data_ref_segment_size): Return
13231 DR_STEP * (niters - 1).
13232 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13234 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13235 (vect_preserves_scalar_order_p): New function, split out from...
13236 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13237 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13238 (vect_vfa_access_size): New function.
13239 (vect_vfa_align): Likewise.
13240 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13241 (dump_lower_bound): New function.
13242 (vect_check_lower_bound): Likewise.
13243 (vect_small_gap_p): Likewise.
13244 (vectorizable_with_step_bound_p): Likewise.
13245 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13246 depencies if the vectorization factor is 1. Convert the checks
13247 for nonzero steps into checks on the bounds of DR_STEP. Try using
13248 a bunds check for variable steps if the minimum required step is
13249 relatively small. Update calls to the dr_with_seg_len
13250 constructor and to vect_compile_time_alias.
13251 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13253 (vect_loop_versioning): Call it.
13254 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13256 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13259 Alan Hayward <alan.hayward@arm.com>
13260 David Sherwood <david.sherwood@arm.com>
13262 * doc/sourcebuild.texi (vect_scatter_store): Document.
13263 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13265 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13267 * genopinit.c (main): Add supports_vec_scatter_store and
13268 supports_vec_scatter_store_cached to target_optabs.
13269 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13270 IFN_MASK_SCATTER_STORE.
13271 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13273 * internal-fn.h (internal_store_fn_p): Declare.
13274 (internal_fn_stored_value_index): Likewise.
13275 * internal-fn.c (scatter_store_direct): New macro.
13276 (expand_scatter_store_optab_fn): New function.
13277 (direct_scatter_store_optab_supported_p): New macro.
13278 (internal_store_fn_p): New function.
13279 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13280 IFN_MASK_SCATTER_STORE.
13281 (internal_fn_mask_index): Likewise.
13282 (internal_fn_stored_value_index): New function.
13283 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13284 for scatter stores.
13285 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13286 * optabs-query.c (supports_vec_scatter_store_p): New function.
13287 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13288 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13289 true for scatter stores.
13290 (vect_gather_scatter_fn_p): Handle scatter stores too.
13291 (vect_check_gather_scatter): Consider using scatter stores if
13292 supports_vec_scatter_store_p.
13293 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13294 scatter stores too.
13295 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13296 internal_fn_stored_value_index.
13297 (check_load_store_masking): Handle scatter stores too.
13298 (vect_get_store_rhs): Make public.
13299 (vectorizable_call): Use internal_store_fn_p.
13300 (vectorizable_store): Handle scatter store internal functions.
13301 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13302 when deciding whether the end of the group has been reached.
13303 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13304 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13305 (mask_scatter_store<mode>): New insns.
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_gather_scatter_fn_p): Declare.
13312 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13313 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13315 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13316 Use vect_truncate_gather_scatter_offset if we can't treat the
13317 operation as a normal gather load or scatter store.
13318 (get_group_load_store_type): Take the gather_scatter_info
13319 as argument. Try using a gather load or scatter store for
13320 single-element groups.
13321 (get_load_store_type): Update calls to get_group_load_store_type
13322 and vect_use_strided_gather_scatters_p.
13324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13325 Alan Hayward <alan.hayward@arm.com>
13326 David Sherwood <david.sherwood@arm.com>
13328 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13329 optional tree argument.
13330 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13332 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13333 but continue to use the current value as a fallback.
13334 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13335 to compare the updates.
13336 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13337 (get_load_store_type): Use it when handling a strided access.
13338 (vect_get_strided_load_store_ops): New function.
13339 (vect_get_data_ptr_increment): Likewise.
13340 (vectorizable_load): Handle strided gather loads. Always pass
13341 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13344 Alan Hayward <alan.hayward@arm.com>
13345 David Sherwood <david.sherwood@arm.com>
13347 * doc/md.texi (gather_load@var{m}): Document.
13348 (mask_gather_load@var{m}): Likewise.
13349 * genopinit.c (main): Add supports_vec_gather_load and
13350 supports_vec_gather_load_cached to target_optabs.
13351 * optabs-tree.c (init_tree_optimization_optabs): Use
13352 ggc_cleared_alloc to allocate target_optabs.
13353 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13354 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13356 * internal-fn.h (internal_load_fn_p): Declare.
13357 (internal_gather_scatter_fn_p): Likewise.
13358 (internal_fn_mask_index): Likewise.
13359 (internal_gather_scatter_fn_supported_p): Likewise.
13360 * internal-fn.c (gather_load_direct): New macro.
13361 (expand_gather_load_optab_fn): New function.
13362 (direct_gather_load_optab_supported_p): New macro.
13363 (direct_internal_fn_optab): New function.
13364 (internal_load_fn_p): Likewise.
13365 (internal_gather_scatter_fn_p): Likewise.
13366 (internal_fn_mask_index): Likewise.
13367 (internal_gather_scatter_fn_supported_p): Likewise.
13368 * optabs-query.c (supports_at_least_one_mode_p): New function.
13369 (supports_vec_gather_load_p): Likewise.
13370 * optabs-query.h (supports_vec_gather_load_p): Declare.
13371 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13372 and memory_type field.
13373 (NUM_PATTERNS): Bump to 15.
13374 * tree-vect-data-refs.c: Include internal-fn.h.
13375 (vect_gather_scatter_fn_p): New function.
13376 (vect_describe_gather_scatter_call): Likewise.
13377 (vect_check_gather_scatter): Try using internal functions for
13378 gather loads. Recognize existing calls to a gather load function.
13379 (vect_analyze_data_refs): Consider using gather loads if
13380 supports_vec_gather_load_p.
13381 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13382 (vect_get_gather_scatter_offset_type): Likewise.
13383 (vect_convert_mask_for_vectype): Likewise.
13384 (vect_add_conversion_to_patterm): Likewise.
13385 (vect_try_gather_scatter_pattern): Likewise.
13386 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13387 (vect_vect_recog_func_ptrs): Add it.
13388 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13389 internal_fn_mask_index and internal_gather_scatter_fn_p.
13390 (check_load_store_masking): Take the gather_scatter_info as an
13391 argument and handle gather loads.
13392 (vect_get_gather_scatter_ops): New function.
13393 (vectorizable_call): Check internal_load_fn_p.
13394 (vectorizable_load): Likewise. Handle gather load internal
13396 (vectorizable_store): Update call to check_load_store_masking.
13397 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13398 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13399 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13400 (aarch64_gather_scale_operand_d): New predicates.
13401 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13402 (mask_gather_load<mode>): New insns.
13404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13405 Alan Hayward <alan.hayward@arm.com>
13406 David Sherwood <david.sherwood@arm.com>
13408 * optabs.def (fold_left_plus_optab): New optab.
13409 * doc/md.texi (fold_left_plus_@var{m}): Document.
13410 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13411 * internal-fn.c (fold_left_direct): Define.
13412 (expand_fold_left_optab_fn): Likewise.
13413 (direct_fold_left_optab_supported_p): Likewise.
13414 * fold-const-call.c (fold_const_fold_left): New function.
13415 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13416 * tree-parloops.c (valid_reduction_p): New function.
13417 (gather_scalar_reductions): Use it.
13418 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13419 (vect_finish_replace_stmt): Declare.
13420 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13421 (needs_fold_left_reduction_p): New function, split out from...
13422 (vect_is_simple_reduction): ...here. Accept reductions that
13423 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13424 (vect_force_simple_reduction): Also store the reduction type in
13425 the assignment's STMT_VINFO_REDUC_TYPE.
13426 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13427 (merge_with_identity): New function.
13428 (vect_expand_fold_left): Likewise.
13429 (vectorize_fold_left_reduction): Likewise.
13430 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13431 scalar phi in place for it. Check for target support and reject
13432 cases that would reassociate the operation. Defer the transform
13433 phase to vectorize_fold_left_reduction.
13434 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13435 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13436 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13440 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13441 call to ifc_temp_var.
13443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13444 Alan Hayward <alan.hayward@arm.com>
13445 David Sherwood <david.sherwood@arm.com>
13447 * target.def (legitimize_address_displacement): Take the original
13448 offset as a poly_int.
13449 * targhooks.h (default_legitimize_address_displacement): Update
13451 * targhooks.c (default_legitimize_address_displacement): Likewise.
13452 * doc/tm.texi: Regenerate.
13453 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13454 as an argument, moving assert of ad->disp == ad->disp_term to...
13455 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13456 Try calling targetm.legitimize_address_displacement before expanding
13457 the address rather than afterwards, and adjust for the new interface.
13458 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13459 Match the new hook interface. Handle SVE addresses.
13460 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13461 new hook interface.
13463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13465 * Makefile.in (OBJS): Add early-remat.o.
13466 * target.def (select_early_remat_modes): New hook.
13467 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13468 * doc/tm.texi: Regenerate.
13469 * targhooks.h (default_select_early_remat_modes): Declare.
13470 * targhooks.c (default_select_early_remat_modes): New function.
13471 * timevar.def (TV_EARLY_REMAT): New timevar.
13472 * passes.def (pass_early_remat): New pass.
13473 * tree-pass.h (make_pass_early_remat): Declare.
13474 * early-remat.c: New file.
13475 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13477 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13480 Alan Hayward <alan.hayward@arm.com>
13481 David Sherwood <david.sherwood@arm.com>
13483 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13484 vfm1 with a bound_epilog parameter.
13485 (vect_do_peeling): Update calls accordingly, and move the prologue
13486 call earlier in the function. Treat the base bound_epilog as 0 for
13487 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13488 this base when peeling for gaps.
13489 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13490 with fully-masked loops.
13491 (vect_estimate_min_profitable_iters): Handle the single peeled
13492 iteration in that case.
13494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13495 Alan Hayward <alan.hayward@arm.com>
13496 David Sherwood <david.sherwood@arm.com>
13498 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13499 single-element interleaving even if the size is not a power of 2.
13500 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13501 accesses for single-element interleaving if the group size is
13504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13505 Alan Hayward <alan.hayward@arm.com>
13506 David Sherwood <david.sherwood@arm.com>
13508 * doc/md.texi (fold_extract_last_@var{m}): Document.
13509 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13510 * optabs.def (fold_extract_last_optab): New optab.
13511 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13512 * internal-fn.c (fold_extract_direct): New macro.
13513 (expand_fold_extract_optab_fn): Likewise.
13514 (direct_fold_extract_optab_supported_p): Likewise.
13515 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13516 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13517 EXTRACT_LAST_REDUCTION.
13518 (get_initial_def_for_reduction): Do not create an initial vector
13519 for EXTRACT_LAST_REDUCTION reductions.
13520 (vectorizable_reduction): Leave the scalar phi in place for
13521 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13522 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13523 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13524 transform phase to vectorizable_condition.
13525 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13527 (vect_finish_stmt_generation): ...here.
13528 (vect_finish_replace_stmt): New function.
13529 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13530 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13532 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13535 Alan Hayward <alan.hayward@arm.com>
13536 David Sherwood <david.sherwood@arm.com>
13538 * doc/md.texi (extract_last_@var{m}): Document.
13539 * optabs.def (extract_last_optab): New optab.
13540 * internal-fn.def (EXTRACT_LAST): New internal function.
13541 * internal-fn.c (cond_unary_direct): New macro.
13542 (expand_cond_unary_optab_fn): Likewise.
13543 (direct_cond_unary_optab_supported_p): Likewise.
13544 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13545 loops using EXTRACT_LAST.
13546 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13547 (extract_last_<mode>): ...this optab.
13548 (vec_extract<mode><Vel>): Update accordingly.
13550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13551 Alan Hayward <alan.hayward@arm.com>
13552 David Sherwood <david.sherwood@arm.com>
13554 * target.def (empty_mask_is_expensive): New hook.
13555 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13556 * doc/tm.texi: Regenerate.
13557 * targhooks.h (default_empty_mask_is_expensive): Declare.
13558 * targhooks.c (default_empty_mask_is_expensive): New function.
13559 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13560 if the target says that empty masks are expensive.
13561 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13563 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13566 Alan Hayward <alan.hayward@arm.com>
13567 David Sherwood <david.sherwood@arm.com>
13569 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13570 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13571 (vect_use_loop_mask_for_alignment_p): New function.
13572 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13573 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13574 niters_skip argument. Make sure that the first niters_skip elements
13575 of the first iteration are inactive.
13576 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13577 Update call to vect_set_loop_masks_directly.
13578 (get_misalign_in_elems): New function, split out from...
13579 (vect_gen_prolog_loop_niters): ...here.
13580 (vect_update_init_of_dr): Take a code argument that specifies whether
13581 the adjustment should be added or subtracted.
13582 (vect_update_init_of_drs): Likewise.
13583 (vect_prepare_for_masked_peels): New function.
13584 (vect_do_peeling): Skip prologue peeling if we're using a mask
13585 instead. Update call to vect_update_inits_of_drs.
13586 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13588 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13589 alignment. Do not include the number of peeled iterations in
13590 the minimum threshold in that case.
13591 (vectorizable_induction): Adjust the start value down by
13592 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13593 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13594 Take the number of skipped iterations into account when calculating
13596 * tree-vect-stmts.c (vect_gen_while_not): New function.
13598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13599 Alan Hayward <alan.hayward@arm.com>
13600 David Sherwood <david.sherwood@arm.com>
13602 * doc/sourcebuild.texi (vect_fully_masked): Document.
13603 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13604 default value to 0.
13605 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13607 (vect_analyze_loop_2): ...here. Don't check the vectorization
13608 factor against the number of loop iterations if the loop is
13611 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13612 Alan Hayward <alan.hayward@arm.com>
13613 David Sherwood <david.sherwood@arm.com>
13615 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13616 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13617 (dump_groups): Update accordingly.
13618 (iv_use::mem_type): New member variable.
13619 (address_p): New function.
13620 (record_use): Add a mem_type argument and initialize the new
13622 (record_group_use): Add a mem_type argument. Use address_p.
13623 Remove obsolete null checks of base_object. Update call to record_use.
13624 (find_interesting_uses_op): Update call to record_group_use.
13625 (find_interesting_uses_cond): Likewise.
13626 (find_interesting_uses_address): Likewise.
13627 (get_mem_type_for_internal_fn): New function.
13628 (find_address_like_use): Likewise.
13629 (find_interesting_uses_stmt): Try find_address_like_use before
13630 calling find_interesting_uses_op.
13631 (addr_offset_valid_p): Use the iv mem_type field as the type
13632 of the addressed memory.
13633 (add_autoinc_candidates): Likewise.
13634 (get_address_cost): Likewise.
13635 (split_small_address_groups_p): Use address_p.
13636 (split_address_groups): Likewise.
13637 (add_iv_candidate_for_use): Likewise.
13638 (autoinc_possible_for_pair): Likewise.
13639 (rewrite_groups): Likewise.
13640 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13641 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13642 (get_alias_ptr_type_for_ptr_address): New function.
13643 (rewrite_use_address): Rewrite address uses in calls that were
13644 identified by find_address_like_use.
13646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13647 Alan Hayward <alan.hayward@arm.com>
13648 David Sherwood <david.sherwood@arm.com>
13650 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13652 * gimple-expr.h (is_gimple_addressable: Likewise.
13653 * gimple-expr.c (is_gimple_address): Likewise.
13654 * internal-fn.c (expand_call_mem_ref): New function.
13655 (expand_mask_load_optab_fn): Use it.
13656 (expand_mask_store_optab_fn): Likewise.
13658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13663 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13664 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13665 (cond_umax@var{mode}): Document.
13666 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13667 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13668 (cond_umin_optab, cond_umax_optab): New optabs.
13669 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13670 (COND_IOR, COND_XOR): New internal functions.
13671 * internal-fn.h (get_conditional_internal_fn): Declare.
13672 * internal-fn.c (cond_binary_direct): New macro.
13673 (expand_cond_binary_optab_fn): Likewise.
13674 (direct_cond_binary_optab_supported_p): Likewise.
13675 (get_conditional_internal_fn): New function.
13676 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13677 Cope with reduction statements that are vectorized as calls rather
13679 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13680 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13681 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13682 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13683 (UNSPEC_COND_EOR): New unspecs.
13684 (optab): Add mappings for them.
13685 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13686 (sve_int_op, sve_fp_op): New int attributes.
13688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13689 Alan Hayward <alan.hayward@arm.com>
13690 David Sherwood <david.sherwood@arm.com>
13692 * optabs.def (while_ult_optab): New optab.
13693 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13694 * internal-fn.def (WHILE_ULT): New internal function.
13695 * internal-fn.h (direct_internal_fn_supported_p): New override
13696 that takes two types as argument.
13697 * internal-fn.c (while_direct): New macro.
13698 (expand_while_optab_fn): New function.
13699 (convert_optab_supported_p): Likewise.
13700 (direct_while_optab_supported_p): New macro.
13701 * wide-int.h (wi::udiv_ceil): New function.
13702 * tree-vectorizer.h (rgroup_masks): New structure.
13703 (vec_loop_masks): New typedef.
13704 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13705 and fully_masked_p.
13706 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13707 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13708 (vect_max_vf): New function.
13709 (slpeel_make_loop_iterate_ntimes): Delete.
13710 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13711 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13712 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13713 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13714 internal-fn.h, stor-layout.h and optabs-query.h.
13715 (vect_set_loop_mask): New function.
13716 (add_preheader_seq): Likewise.
13717 (add_header_seq): Likewise.
13718 (interleave_supported_p): Likewise.
13719 (vect_maybe_permute_loop_masks): Likewise.
13720 (vect_set_loop_masks_directly): Likewise.
13721 (vect_set_loop_condition_masked): Likewise.
13722 (vect_set_loop_condition_unmasked): New function, split out from
13723 slpeel_make_loop_iterate_ntimes.
13724 (slpeel_make_loop_iterate_ntimes): Rename to..
13725 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13726 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13727 (vect_do_peeling): Update call accordingly.
13728 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13730 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13731 mask_compare_type, can_fully_mask_p and fully_masked_p.
13732 (release_vec_loop_masks): New function.
13733 (_loop_vec_info): Use it to free the loop masks.
13734 (can_produce_all_loop_masks_p): New function.
13735 (vect_get_max_nscalars_per_iter): Likewise.
13736 (vect_verify_full_masking): Likewise.
13737 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13738 retries, and free the mask rgroups before retrying. Check loop-wide
13739 reasons for disallowing fully-masked loops. Make the final decision
13740 about whether use a fully-masked loop or not.
13741 (vect_estimate_min_profitable_iters): Do not assume that peeling
13742 for the number of iterations will be needed for fully-masked loops.
13743 (vectorizable_reduction): Disable fully-masked loops.
13744 (vectorizable_live_operation): Likewise.
13745 (vect_halve_mask_nunits): New function.
13746 (vect_double_mask_nunits): Likewise.
13747 (vect_record_loop_mask): Likewise.
13748 (vect_get_loop_mask): Likewise.
13749 (vect_transform_loop): Handle the case in which the final loop
13750 iteration might handle a partial vector. Call vect_set_loop_condition
13751 instead of slpeel_make_loop_iterate_ntimes.
13752 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13753 (check_load_store_masking): New function.
13754 (prepare_load_store_mask): Likewise.
13755 (vectorizable_store): Handle fully-masked loops.
13756 (vectorizable_load): Likewise.
13757 (supportable_widening_operation): Use vect_halve_mask_nunits for
13759 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13760 (vect_gen_while): New function.
13761 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13762 (aarch64_uqdec<mode>): New insn.
13764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13765 Alan Hayward <alan.hayward@arm.com>
13766 David Sherwood <david.sherwood@arm.com>
13768 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13769 (reduc_xor_scal_optab): New optabs.
13770 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13771 (reduc_xor_scal_@var{m}): Document.
13772 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13773 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13774 internal functions.
13775 * fold-const-call.c (fold_const_call): Handle them.
13776 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13777 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13778 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13779 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13780 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13781 (UNSPEC_XORV): New unspecs.
13782 (optab): Add entries for them.
13783 (BITWISEV): New int iterator.
13784 (bit_reduc_op): New int attributes.
13786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13787 Alan Hayward <alan.hayward@arm.com>
13788 David Sherwood <david.sherwood@arm.com>
13790 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13791 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13792 * optabs.def (vec_shl_insert_optab): New optab.
13793 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13794 (duplicate_and_interleave): Likewise.
13795 * tree-vect-loop.c: Include internal-fn.h.
13796 (neutral_op_for_slp_reduction): New function, split out from
13797 get_initial_defs_for_reduction.
13798 (get_initial_def_for_reduction): Handle option 2 for variable-length
13799 vectors by loading the neutral value into a vector and then shifting
13800 the initial value into element 0.
13801 (get_initial_defs_for_reduction): Replace the code argument with
13802 the neutral value calculated by neutral_op_for_slp_reduction.
13803 Use gimple_build_vector for constant-length vectors.
13804 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13805 but the first group_size elements have a neutral value.
13806 Use duplicate_and_interleave otherwise.
13807 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13808 Update call to get_initial_defs_for_reduction. Handle SLP
13809 reductions for variable-length vectors by creating one vector
13810 result for each scalar result, with the elements associated
13811 with other scalar results stubbed out with the neutral value.
13812 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13813 Require IFN_VEC_SHL_INSERT for double reductions on
13814 variable-length vectors, or SLP reductions that have
13815 a neutral value. Require can_duplicate_and_interleave_p
13816 support for variable-length unchained SLP reductions if there
13817 is no neutral value, such as for MIN/MAX reductions. Also require
13818 the number of vector elements to be a multiple of the number of
13819 SLP statements when doing variable-length unchained SLP reductions.
13820 Update call to vect_create_epilog_for_reduction.
13821 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13822 and remove initial values.
13823 (duplicate_and_interleave): Make public.
13824 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13825 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13828 Alan Hayward <alan.hayward@arm.com>
13829 David Sherwood <david.sherwood@arm.com>
13831 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13832 (can_duplicate_and_interleave_p): New function.
13833 (vect_get_and_check_slp_defs): Take the vector of statements
13834 rather than just the current one. Remove excess parentheses.
13835 Restriction rejectinon of vect_constant_def and vect_external_def
13836 for variable-length vectors to boolean types, or types for which
13837 can_duplicate_and_interleave_p is false.
13838 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13839 (duplicate_and_interleave): New function.
13840 (vect_get_constant_vectors): Use gimple_build_vector for
13841 constant-length vectors and suitable variable-length constant
13842 vectors. Use duplicate_and_interleave for other variable-length
13843 vectors. Don't defer the update when inserting new statements.
13845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13846 Alan Hayward <alan.hayward@arm.com>
13847 David Sherwood <david.sherwood@arm.com>
13849 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13850 min_profitable_iters doesn't go negative.
13852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13853 Alan Hayward <alan.hayward@arm.com>
13854 David Sherwood <david.sherwood@arm.com>
13856 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13857 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13858 * optabs.def (vec_mask_load_lanes_optab): New optab.
13859 (vec_mask_store_lanes_optab): Likewise.
13860 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13861 (MASK_STORE_LANES): Likewise.
13862 * internal-fn.c (mask_load_lanes_direct): New macro.
13863 (mask_store_lanes_direct): Likewise.
13864 (expand_mask_load_optab_fn): Handle masked operations.
13865 (expand_mask_load_lanes_optab_fn): New macro.
13866 (expand_mask_store_optab_fn): Handle masked operations.
13867 (expand_mask_store_lanes_optab_fn): New macro.
13868 (direct_mask_load_lanes_optab_supported_p): Likewise.
13869 (direct_mask_store_lanes_optab_supported_p): Likewise.
13870 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13872 (vect_load_lanes_supported): Likewise.
13873 * tree-vect-data-refs.c (strip_conversion): New function.
13874 (can_group_stmts_p): Likewise.
13875 (vect_analyze_data_ref_accesses): Use it instead of checking
13876 for a pair of assignments.
13877 (vect_store_lanes_supported): Take a masked_p parameter.
13878 (vect_load_lanes_supported): Likewise.
13879 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13880 vect_store_lanes_supported and vect_load_lanes_supported.
13881 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13882 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13883 parameter. Don't allow gaps for masked accesses.
13884 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13885 and vect_load_lanes_supported.
13886 (get_load_store_type): Take a masked_p parameter and update
13887 call to get_group_load_store_type.
13888 (vectorizable_store): Update call to get_load_store_type.
13889 Handle IFN_MASK_STORE_LANES.
13890 (vectorizable_load): Update call to get_load_store_type.
13891 Handle IFN_MASK_LOAD_LANES.
13893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13899 * config/aarch64/aarch64-protos.h
13900 (aarch64_sve_struct_memory_operand_p): Declare.
13901 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13902 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13903 (VPRED, vpred): Handle SVE structure modes.
13904 * config/aarch64/constraints.md (Utx): New constraint.
13905 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13906 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13907 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13908 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13909 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13910 structure modes. Split into pieces after RA.
13911 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13912 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13914 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13915 SVE structure modes.
13916 (aarch64_classify_address): Likewise.
13917 (sizetochar): Move earlier in file.
13918 (aarch64_print_operand): Handle SVE register lists.
13919 (aarch64_array_mode): New function.
13920 (aarch64_sve_struct_memory_operand_p): Likewise.
13921 (TARGET_ARRAY_MODE): Redefine.
13923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13924 Alan Hayward <alan.hayward@arm.com>
13925 David Sherwood <david.sherwood@arm.com>
13927 * target.def (array_mode): New target hook.
13928 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13929 * doc/tm.texi: Regenerate.
13930 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13931 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13932 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13933 targetm.array_mode.
13934 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13938 Alan Hayward <alan.hayward@arm.com>
13939 David Sherwood <david.sherwood@arm.com>
13941 * fold-const.c (fold_binary_loc): Check the argument types
13942 rather than the result type when testing for a vector operation.
13944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13946 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13947 * doc/tm.texi: Regenerate.
13949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13950 Alan Hayward <alan.hayward@arm.com>
13951 David Sherwood <david.sherwood@arm.com>
13953 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13954 (sve): Document new AArch64 extension.
13955 * doc/md.texi (w): Extend the description of the AArch64
13956 constraint to include SVE vectors.
13957 (Upl, Upa): Document new AArch64 predicate constraints.
13958 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13960 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13961 (msve-vector-bits=): New option.
13962 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13963 SVE when these are disabled.
13964 (sve): New extension.
13965 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13966 modes. Adjust their number of units based on aarch64_sve_vg.
13967 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13968 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13969 aarch64_addr_query_type.
13970 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13971 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13972 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13973 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13974 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13975 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13976 (aarch64_simd_imm_zero_p): Delete.
13977 (aarch64_check_zero_based_sve_index_immediate): Declare.
13978 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13979 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13980 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13981 (aarch64_sve_float_mul_immediate_p): Likewise.
13982 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13983 rather than an rtx.
13984 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13985 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13986 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13987 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13988 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13989 (aarch64_regmode_natural_size): Likewise.
13990 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13991 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13993 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13994 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13995 for VG and the SVE predicate registers.
13996 (V_ALIASES): Add a "z"-prefixed alias.
13997 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13998 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13999 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14000 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14001 (REG_CLASS_NAMES): Add entries for them.
14002 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14003 and the predicate registers.
14004 (aarch64_sve_vg): Declare.
14005 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14006 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14007 (REGMODE_NATURAL_SIZE): Define.
14008 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14010 * config/aarch64/aarch64.c: Include cfgrtl.h.
14011 (simd_immediate_info): Add a constructor for series vectors,
14012 and an associated step field.
14013 (aarch64_sve_vg): New variable.
14014 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14015 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14016 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14017 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14018 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14019 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14020 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14021 (aarch64_get_mask_mode): New functions.
14022 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14023 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14024 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14025 predicate modes and predicate registers. Explicitly restrict
14026 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14027 to store a vector mode if it is recognized by
14028 aarch64_classify_vector_mode.
14029 (aarch64_regmode_natural_size): New function.
14030 (aarch64_hard_regno_caller_save_mode): Return the original mode
14032 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14033 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14034 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14035 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14037 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14038 does not overlap dest if the function is frame-related. Handle
14040 (aarch64_split_add_offset): New function.
14041 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14042 them aarch64_add_offset.
14043 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14044 and update call to aarch64_sub_sp.
14045 (aarch64_add_cfa_expression): New function.
14046 (aarch64_expand_prologue): Pass extra temporary registers to the
14047 functions above. Handle the case in which we need to emit new
14048 DW_CFA_expressions for registers that were originally saved
14049 relative to the stack pointer, but now have to be expressed
14050 relative to the frame pointer.
14051 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14053 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14054 IP0 and IP1 values for SVE frames.
14055 (aarch64_expand_vec_series): New function.
14056 (aarch64_expand_sve_widened_duplicate): Likewise.
14057 (aarch64_expand_sve_const_vector): Likewise.
14058 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14059 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14060 into the register, rather than emitting a SET directly.
14061 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14062 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14063 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14064 (offset_9bit_signed_scaled_p): New functions.
14065 (aarch64_replicate_bitmask_imm): New function.
14066 (aarch64_bitmask_imm): Use it.
14067 (aarch64_cannot_force_const_mem): Reject expressions involving
14068 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14069 (aarch64_classify_index): Handle SVE indices, by requiring
14070 a plain register index with a scale that matches the element size.
14071 (aarch64_classify_address): Handle SVE addresses. Assert that
14072 the mode of the address is VOIDmode or an integer mode.
14073 Update call to aarch64_classify_symbol.
14074 (aarch64_classify_symbolic_expression): Update call to
14075 aarch64_classify_symbol.
14076 (aarch64_const_vec_all_in_range_p): New function.
14077 (aarch64_print_vector_float_operand): Likewise.
14078 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14079 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14080 and the FP immediates 1.0 and 0.5.
14081 (aarch64_print_address_internal): Handle SVE addresses.
14082 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14083 (aarch64_regno_regclass): Handle predicate registers.
14084 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14086 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14087 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14088 (aarch64_convert_sve_vector_bits): New function.
14089 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14090 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14091 rather than an rtx.
14092 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14093 Handle SVE vector and predicate modes. Accept VL-based constants
14094 that need only one temporary register, and VL offsets that require
14095 no temporary registers.
14096 (aarch64_conditional_register_usage): Mark the predicate registers
14097 as fixed if SVE isn't available.
14098 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14099 Return true for SVE vector and predicate modes.
14100 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14101 rather than an unsigned int. Handle SVE modes.
14102 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14104 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14106 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14107 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14108 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14109 (aarch64_sve_float_mul_immediate_p): New functions.
14110 (aarch64_sve_valid_immediate): New function.
14111 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14112 Explicitly reject structure modes. Check for INDEX constants.
14113 Handle PTRUE and PFALSE constants.
14114 (aarch64_check_zero_based_sve_index_immediate): New function.
14115 (aarch64_simd_imm_zero_p): Delete.
14116 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14117 vector modes. Accept constants in the range of CNT[BHWD].
14118 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14119 ask for an Advanced SIMD mode.
14120 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14121 (aarch64_simd_vector_alignment): Handle SVE predicates.
14122 (aarch64_vectorize_preferred_vector_alignment): New function.
14123 (aarch64_simd_vector_alignment_reachable): Use it instead of
14125 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14126 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14128 (MAX_VECT_LEN): Delete.
14129 (expand_vec_perm_d): Add a vec_flags field.
14130 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14131 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14132 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14134 (aarch64_evpc_rev): Rename to...
14135 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14136 (aarch64_evpc_rev_global): New function.
14137 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14138 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14140 (aarch64_evpc_sve_tbl): New function.
14141 (aarch64_expand_vec_perm_const_1): Update after rename of
14142 aarch64_evpc_rev. Handle SVE permutes too, trying
14143 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14144 than aarch64_evpc_tbl.
14145 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14146 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14147 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14148 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14149 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14150 (aarch64_expand_sve_vcond): New functions.
14151 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14152 of aarch64_vector_mode_p.
14153 (aarch64_dwarf_poly_indeterminate_value): New function.
14154 (aarch64_compute_pressure_classes): Likewise.
14155 (aarch64_can_change_mode_class): Likewise.
14156 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14157 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14158 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14159 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14160 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14161 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14162 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14163 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14165 (Dn, Dl, Dr): Accept const as well as const_vector.
14166 (Dz): Likewise. Compare against CONST0_RTX.
14167 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14168 of "vector" where appropriate.
14169 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14170 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14171 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14172 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14173 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14174 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14175 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14176 (v_int_equiv): Extend to SVE modes.
14177 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14179 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14180 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14181 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14182 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14183 (SVE_COND_FP_CMP): New int iterators.
14184 (perm_hilo): Handle the new unpack unspecs.
14185 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14187 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14188 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14189 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14190 (aarch64_equality_operator, aarch64_constant_vector_operand)
14191 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14192 (aarch64_sve_nonimmediate_operand): Likewise.
14193 (aarch64_sve_general_operand): Likewise.
14194 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14195 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14196 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14197 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14198 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14199 (aarch64_sve_float_arith_immediate): Likewise.
14200 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14201 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14202 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14203 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14204 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14205 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14206 (aarch64_sve_float_arith_operand): Likewise.
14207 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14208 (aarch64_sve_float_mul_operand): Likewise.
14209 (aarch64_sve_vec_perm_operand): Likewise.
14210 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14211 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14212 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14213 as well as const_vector.
14214 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14215 in file. Use CONST0_RTX and CONSTM1_RTX.
14216 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14217 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14218 Use aarch64_simd_imm_zero.
14219 * config/aarch64/aarch64-sve.md: New file.
14220 * config/aarch64/aarch64.md: Include it.
14221 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14222 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14223 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14224 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14225 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14226 (sve): New attribute.
14227 (enabled): Disable instructions with the sve attribute unless
14229 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14230 aarch64_expand_mov_immediate.
14231 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14232 CNT[BHSD] immediates.
14233 (movti): Split CONST_POLY_INT moves into two halves.
14234 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14235 Split additions that need a temporary here if the destination
14236 is the stack pointer.
14237 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14238 (*add<mode>3_poly_1): New instruction.
14239 (set_clobber_cc): New expander.
14241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14243 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14244 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14245 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14246 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14247 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14248 Change innermode from fixed_mode_size to machine_mode.
14249 (simplify_subreg): Update call accordingly. Handle a constant-sized
14250 subreg of a variable-length CONST_VECTOR.
14252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14253 Alan Hayward <alan.hayward@arm.com>
14254 David Sherwood <david.sherwood@arm.com>
14256 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14257 (add_offset_to_base): New function, split out from...
14258 (create_mem_ref): ...here. When handling a scale other than 1,
14259 check first whether the address is valid without the offset.
14260 Add it into the base if so, leaving the index and scale as-is.
14262 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14265 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14266 fold_for_warn before checking if arg2 is INTEGER_CST.
14268 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14270 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14271 (store_multiple_operation): Delete.
14272 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14273 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14274 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14275 guarded by TARGET_STRING.
14276 (rs6000_output_load_multiple): Delete.
14277 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14278 OPTION_MASK_STRING / TARGET_STRING handling.
14279 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14280 (const rs6000_opt_masks) <"string">: Change mask to 0.
14281 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14282 (MASK_STRING): Delete.
14283 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14285 (load_multiple): Delete.
14292 (store_multiple): Delete.
14299 (movmemsi_8reg): Delete.
14300 (corresponding unnamed define_insn): Delete.
14301 (movmemsi_6reg): Delete.
14302 (corresponding unnamed define_insn): Delete.
14303 (movmemsi_4reg): Delete.
14304 (corresponding unnamed define_insn): Delete.
14305 (movmemsi_2reg): Delete.
14306 (corresponding unnamed define_insn): Delete.
14307 (movmemsi_1reg): Delete.
14308 (corresponding unnamed define_insn): Delete.
14309 * config/rs6000/rs6000.opt (mno-string): New.
14310 (mstring): Replace by deprecation warning stub.
14311 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14313 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14315 * regrename.c (regrename_do_replace): If replacing the same
14316 reg multiple times, try to reuse last created gen_raw_REG.
14319 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14320 main to workaround a bug in GDB.
14322 2018-01-12 Tom de Vries <tom@codesourcery.com>
14325 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14327 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14329 PR rtl-optimization/80481
14330 * ira-color.c (get_cap_member): New function.
14331 (allocnos_conflict_by_live_ranges_p): Use it.
14332 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14333 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14335 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14338 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14339 (*saddl_se_1): Ditto.
14340 (*ssubsi_1): Ditto.
14341 (*ssubl_se_1): Ditto.
14343 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14345 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14346 rather than wi::to_widest for DR_INITs.
14347 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14348 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14349 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14351 (vect_analyze_group_access_1): Note that here.
14353 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14355 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14356 polynomial type sizes.
14358 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14360 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14361 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14362 (gimple_add_tmp_var): Likewise.
14364 2018-01-12 Martin Liska <mliska@suse.cz>
14366 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14367 (gimple_alloc_sizes): Likewise.
14368 (dump_gimple_statistics): Use PRIu64 in printf format.
14369 * gimple.h: Change uint64_t to int.
14371 2018-01-12 Martin Liska <mliska@suse.cz>
14373 * tree-core.h: Use uint64_t instead of int.
14374 * tree.c (tree_node_counts): Likewise.
14375 (tree_node_sizes): Likewise.
14376 (dump_tree_statistics): Use PRIu64 in printf format.
14378 2018-01-12 Martin Liska <mliska@suse.cz>
14380 * Makefile.in: As qsort_chk is implemented in vec.c, add
14381 vec.o to linkage of gencfn-macros.
14382 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14383 passing the info to record_node_allocation_statistics.
14384 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14386 * ggc-common.c (struct ggc_usage): Add operator== and use
14387 it in operator< and compare function.
14388 * mem-stats.h (struct mem_usage): Likewise.
14389 * vec.c (struct vec_usage): Remove operator< and compare
14390 function. Can be simply inherited.
14392 2018-01-12 Martin Jambor <mjambor@suse.cz>
14395 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14396 * tree-ssa-math-opts.c: Include domwalk.h.
14397 (convert_mult_to_fma_1): New function.
14398 (fma_transformation_info): New type.
14399 (fma_deferring_state): Likewise.
14400 (cancel_fma_deferring): New function.
14401 (result_of_phi): Likewise.
14402 (last_fma_candidate_feeds_initial_phi): Likewise.
14403 (convert_mult_to_fma): Added deferring logic, split actual
14404 transformation to convert_mult_to_fma_1.
14405 (math_opts_dom_walker): New type.
14406 (math_opts_dom_walker::after_dom_children): New method, body moved
14407 here from pass_optimize_widening_mul::execute, added deferring logic
14409 (pass_optimize_widening_mul::execute): Moved most of code to
14410 math_opts_dom_walker::after_dom_children.
14411 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14412 * config/i386/i386.c (ix86_option_override_internal): Added
14413 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14415 2018-01-12 Richard Biener <rguenther@suse.de>
14418 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14419 inline instance vars.
14421 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14424 * config/rx/rx.c (rx_is_restricted_memory_address):
14425 Handle SUBREG case.
14427 2018-01-12 Richard Biener <rguenther@suse.de>
14429 PR tree-optimization/80846
14430 * target.def (split_reduction): New target hook.
14431 * targhooks.c (default_split_reduction): New function.
14432 * targhooks.h (default_split_reduction): Declare.
14433 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14434 target requests first reduce vectors by combining low and high
14436 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14437 (get_vectype_for_scalar_type_and_size): Export.
14438 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14439 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14440 * doc/tm.texi: Regenerate.
14441 * config/i386/i386.c (ix86_split_reduction): Implement
14442 TARGET_VECTORIZE_SPLIT_REDUCTION.
14444 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14447 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14448 in PIC mode except for TARGET_VXWORKS_RTP.
14449 * config/sparc/sparc.c: Include cfgrtl.h.
14450 (TARGET_INIT_PIC_REG): Define.
14451 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14452 (sparc_pic_register_p): New predicate.
14453 (sparc_legitimate_address_p): Use it.
14454 (sparc_legitimize_pic_address): Likewise.
14455 (sparc_delegitimize_address): Likewise.
14456 (sparc_mode_dependent_address_p): Likewise.
14457 (gen_load_pcrel_sym): Remove 4th parameter.
14458 (load_got_register): Adjust call to above. Remove obsolete stuff.
14459 (sparc_expand_prologue): Do not call load_got_register here.
14460 (sparc_flat_expand_prologue): Likewise.
14461 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14462 (sparc_use_pseudo_pic_reg): New function.
14463 (sparc_init_pic_reg): Likewise.
14464 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14465 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14467 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14469 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14470 Add item for branch_cost.
14472 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14474 PR rtl-optimization/83565
14475 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14476 not extend the result to a larger mode for rotate operations.
14477 (num_sign_bit_copies1): Likewise.
14479 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14482 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14484 Use values-Xc.o for -pedantic.
14485 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14487 2018-01-12 Martin Liska <mliska@suse.cz>
14490 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14492 (possible_polymorphic_call_targets): Use it.
14493 (ipa_devirt): Likewise.
14495 2018-01-12 Martin Liska <mliska@suse.cz>
14497 * profile-count.h (enum profile_quality): Use 0 as invalid
14498 enum value of profile_quality.
14500 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14502 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14503 -mext-string options.
14505 2018-01-12 Richard Biener <rguenther@suse.de>
14507 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14508 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14509 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14511 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14513 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14515 * configure.ac (--with-long-double-format): Add support for the
14516 configuration option to change the default long double format on
14518 * config.gcc (powerpc*-linux*-*): Likewise.
14519 * configure: Regenerate.
14520 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14521 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14522 used without modification.
14524 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14526 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14527 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14528 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14529 MISC_BUILTIN_SPEC_BARRIER.
14530 (rs6000_init_builtins): Likewise.
14531 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14533 (speculation_barrier): New define_insn.
14534 * doc/extend.texi: Document __builtin_speculation_barrier.
14536 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14539 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14540 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14541 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14543 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14544 integral modes instead of "ss" and "sd".
14545 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14546 vectors with 32-bit and 64-bit elements.
14547 (vecdupssescalarmodesuffix): New mode attribute.
14548 (vec_dup<mode>): Use it.
14550 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14553 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14554 frame if argument is passed on stack.
14556 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14559 * ree.c (combine_reaching_defs): Optimize also
14560 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14561 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14563 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14565 PR middle-end/83189
14566 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14568 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14570 PR middle-end/83718
14571 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14572 after they are computed.
14574 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14576 PR tree-optimization/83695
14577 * gimple-loop-linterchange.cc
14578 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14579 reset cached scev information after interchange.
14580 (pass_linterchange::execute): Remove call to scev_reset_htab.
14582 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14584 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14585 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14586 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14587 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14588 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14589 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14590 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14591 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14592 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14593 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14594 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14595 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14596 (V_lane_reg): Likewise.
14597 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14599 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14600 (vfmal_lane_low<mode>_intrinsic,
14601 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14602 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14603 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14604 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14605 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14606 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14608 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14610 * config/arm/arm-cpus.in (fp16fml): New feature.
14611 (ALL_SIMD): Add fp16fml.
14612 (armv8.2-a): Add fp16fml as an option.
14613 (armv8.3-a): Likewise.
14614 (armv8.4-a): Add fp16fml as part of fp16.
14615 * config/arm/arm.h (TARGET_FP16FML): Define.
14616 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14618 * config/arm/arm-modes.def (V2HF): Define.
14619 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14620 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14621 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14622 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14623 vfmsl_low, vfmsl_high): New set of builtins.
14624 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14625 (vfml_op): New code attribute.
14626 (VFMLHALVES): New int iterator.
14627 (VFML, VFMLSEL): New mode attributes.
14628 (V_reg): Define mapping for V2HF.
14629 (V_hi, V_lo): New mode attributes.
14630 (VF_constraint): Likewise.
14631 (vfml_half, vfml_half_selector): New int attributes.
14632 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14634 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14635 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14637 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14638 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14639 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14640 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14642 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14643 Document new effective target and option set.
14645 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14647 * config/arm/arm-cpus.in (armv8_4): New feature.
14648 (ARMv8_4a): New fgroup.
14649 (armv8.4-a): New arch.
14650 * config/arm/arm-tables.opt: Regenerate.
14651 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14652 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14653 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14654 Add matching rules for -march=armv8.4-a and extensions.
14655 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14657 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14660 * config/rx/rx.md (BW): New mode attribute.
14661 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14663 2018-01-11 Richard Biener <rguenther@suse.de>
14665 PR tree-optimization/83435
14666 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14667 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14668 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14670 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14671 Alan Hayward <alan.hayward@arm.com>
14672 David Sherwood <david.sherwood@arm.com>
14674 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14676 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14677 (aarch64_print_address_internal): Use it to check for a zero offset.
14679 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14680 Alan Hayward <alan.hayward@arm.com>
14681 David Sherwood <david.sherwood@arm.com>
14683 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14684 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14685 Return a poly_int64 rather than a HOST_WIDE_INT.
14686 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14687 rather than a HOST_WIDE_INT.
14688 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14689 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14690 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14691 final_offset from HOST_WIDE_INT to poly_int64.
14692 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14693 to_constant when getting the number of units in an Advanced SIMD
14695 (aarch64_builtin_vectorized_function): Check for a constant number
14697 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14699 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14700 attribute instead of GET_MODE_NUNITS.
14701 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14702 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14703 GET_MODE_SIZE for fixed-size registers.
14704 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14705 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14706 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14707 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14708 (aarch64_print_operand, aarch64_print_address_internal)
14709 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14710 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14711 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14712 Handle polynomial GET_MODE_SIZE.
14713 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14714 wider than SImode without modification.
14715 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14716 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14717 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14718 passing and returning SVE modes.
14719 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14720 rather than GEN_INT.
14721 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14722 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14723 (aarch64_allocate_and_probe_stack_space): Likewise.
14724 (aarch64_layout_frame): Cope with polynomial offsets.
14725 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14726 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14727 polynomial offsets.
14728 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14729 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14730 poly_int64 rather than a HOST_WIDE_INT.
14731 (aarch64_get_separate_components, aarch64_process_components)
14732 (aarch64_expand_prologue, aarch64_expand_epilogue)
14733 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14734 (aarch64_anchor_offset): New function, split out from...
14735 (aarch64_legitimize_address): ...here.
14736 (aarch64_builtin_vectorization_cost): Handle polynomial
14737 TYPE_VECTOR_SUBPARTS.
14738 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14740 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14741 number of elements from the PARALLEL rather than the mode.
14742 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14743 rather than GET_MODE_BITSIZE.
14744 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14745 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14746 (aarch64_expand_vec_perm_const_1): Handle polynomial
14747 d->perm.length () and d->perm elements.
14748 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14749 Apply to_constant to d->perm elements.
14750 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14751 polynomial CONST_VECTOR_NUNITS.
14752 (aarch64_move_pointer): Take amount as a poly_int64 rather
14754 (aarch64_progress_pointer): Avoid temporary variable.
14755 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14756 the mode attribute instead of GET_MODE.
14758 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14759 Alan Hayward <alan.hayward@arm.com>
14760 David Sherwood <david.sherwood@arm.com>
14762 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14763 x exists before using it.
14764 (aarch64_add_constant_internal): Rename to...
14765 (aarch64_add_offset_1): ...this. Replace regnum with separate
14766 src and dest rtxes. Handle the case in which they're different,
14767 including when the offset is zero. Replace scratchreg with an rtx.
14768 Use 2 additions if there is no spare register into which we can
14769 move a 16-bit constant.
14770 (aarch64_add_constant): Delete.
14771 (aarch64_add_offset): Replace reg with separate src and dest
14772 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14773 Use aarch64_add_offset_1.
14774 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14775 an rtx rather than an int. Take the delta as a poly_int64
14776 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14777 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14778 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14779 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14780 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14781 and aarch64_add_sp.
14782 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14783 aarch64_add_constant.
14785 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14787 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14788 Use scalar_float_mode.
14790 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14792 * config/aarch64/aarch64-simd.md
14793 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14794 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14795 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14796 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14797 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14798 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14799 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14800 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14801 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14802 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14804 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14807 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14808 targ_options->x_arm_arch_string is non NULL.
14810 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14812 * config/aarch64/aarch64.h
14813 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14815 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14818 * expmed.c (emit_store_flag_force): Swap if const op0
14819 and change VOIDmode to mode of op0.
14821 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14823 PR rtl-optimization/83761
14824 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14825 than bytes to mode_for_size.
14827 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14829 PR middle-end/83189
14830 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14831 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14834 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14836 PR middle-end/83575
14837 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14838 when in layout mode.
14839 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14840 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14843 2018-01-10 Michael Collison <michael.collison@arm.com>
14845 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14846 * config/aarch64/aarch64-option-extension.def: Add
14847 AARCH64_OPT_EXTENSION of 'fp16fml'.
14848 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14849 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14850 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14851 * config/aarch64/constraints.md (Ui7): New constraint.
14852 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14853 (VFMLA_SEL_W): Ditto.
14856 (VFMLA16_LOW): New int iterator.
14857 (VFMLA16_HIGH): Ditto.
14858 (UNSPEC_FMLAL): New unspec.
14859 (UNSPEC_FMLSL): Ditto.
14860 (UNSPEC_FMLAL2): Ditto.
14861 (UNSPEC_FMLSL2): Ditto.
14862 (f16mac): New code attribute.
14863 * config/aarch64/aarch64-simd-builtins.def
14864 (aarch64_fmlal_lowv2sf): Ditto.
14865 (aarch64_fmlsl_lowv2sf): Ditto.
14866 (aarch64_fmlalq_lowv4sf): Ditto.
14867 (aarch64_fmlslq_lowv4sf): Ditto.
14868 (aarch64_fmlal_highv2sf): Ditto.
14869 (aarch64_fmlsl_highv2sf): Ditto.
14870 (aarch64_fmlalq_highv4sf): Ditto.
14871 (aarch64_fmlslq_highv4sf): Ditto.
14872 (aarch64_fmlal_lane_lowv2sf): Ditto.
14873 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14874 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14875 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14876 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14877 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14878 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14879 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14880 (aarch64_fmlal_lane_highv2sf): Ditto.
14881 (aarch64_fmlsl_lane_highv2sf): Ditto.
14882 (aarch64_fmlal_laneq_highv2sf): Ditto.
14883 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14884 (aarch64_fmlalq_lane_highv4sf): Ditto.
14885 (aarch64_fmlsl_lane_highv4sf): Ditto.
14886 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14887 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14888 * config/aarch64/aarch64-simd.md:
14889 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14890 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14891 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14892 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14893 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14894 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14895 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14896 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14897 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14898 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14899 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14900 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14901 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14902 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14903 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14904 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14905 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14906 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14907 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14908 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14909 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14910 (vfmlsl_low_u32): Ditto.
14911 (vfmlalq_low_u32): Ditto.
14912 (vfmlslq_low_u32): Ditto.
14913 (vfmlal_high_u32): Ditto.
14914 (vfmlsl_high_u32): Ditto.
14915 (vfmlalq_high_u32): Ditto.
14916 (vfmlslq_high_u32): Ditto.
14917 (vfmlal_lane_low_u32): Ditto.
14918 (vfmlsl_lane_low_u32): Ditto.
14919 (vfmlal_laneq_low_u32): Ditto.
14920 (vfmlsl_laneq_low_u32): Ditto.
14921 (vfmlalq_lane_low_u32): Ditto.
14922 (vfmlslq_lane_low_u32): Ditto.
14923 (vfmlalq_laneq_low_u32): Ditto.
14924 (vfmlslq_laneq_low_u32): Ditto.
14925 (vfmlal_lane_high_u32): Ditto.
14926 (vfmlsl_lane_high_u32): Ditto.
14927 (vfmlal_laneq_high_u32): Ditto.
14928 (vfmlsl_laneq_high_u32): Ditto.
14929 (vfmlalq_lane_high_u32): Ditto.
14930 (vfmlslq_lane_high_u32): Ditto.
14931 (vfmlalq_laneq_high_u32): Ditto.
14932 (vfmlslq_laneq_high_u32): Ditto.
14933 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14934 (AARCH64_FL_FOR_ARCH8_4): New.
14935 (AARCH64_ISA_F16FML): New ISA flag.
14936 (TARGET_F16FML): New feature flag for fp16fml.
14937 (doc/invoke.texi): Document new fp16fml option.
14939 2018-01-10 Michael Collison <michael.collison@arm.com>
14941 * config/aarch64/aarch64-builtins.c:
14942 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14943 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14944 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14945 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14946 (AARCH64_ISA_SHA3): New ISA flag.
14947 (TARGET_SHA3): New feature flag for sha3.
14948 * config/aarch64/iterators.md (sha512_op): New int attribute.
14949 (CRYPTO_SHA512): New int iterator.
14950 (UNSPEC_SHA512H): New unspec.
14951 (UNSPEC_SHA512H2): Ditto.
14952 (UNSPEC_SHA512SU0): Ditto.
14953 (UNSPEC_SHA512SU1): Ditto.
14954 * config/aarch64/aarch64-simd-builtins.def
14955 (aarch64_crypto_sha512hqv2di): New builtin.
14956 (aarch64_crypto_sha512h2qv2di): Ditto.
14957 (aarch64_crypto_sha512su0qv2di): Ditto.
14958 (aarch64_crypto_sha512su1qv2di): Ditto.
14959 (aarch64_eor3qv8hi): Ditto.
14960 (aarch64_rax1qv2di): Ditto.
14961 (aarch64_xarqv2di): Ditto.
14962 (aarch64_bcaxqv8hi): Ditto.
14963 * config/aarch64/aarch64-simd.md:
14964 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14965 (aarch64_crypto_sha512su0qv2di): Ditto.
14966 (aarch64_crypto_sha512su1qv2di): Ditto.
14967 (aarch64_eor3qv8hi): Ditto.
14968 (aarch64_rax1qv2di): Ditto.
14969 (aarch64_xarqv2di): Ditto.
14970 (aarch64_bcaxqv8hi): Ditto.
14971 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14972 (vsha512h2q_u64): Ditto.
14973 (vsha512su0q_u64): Ditto.
14974 (vsha512su1q_u64): Ditto.
14975 (veor3q_u16): Ditto.
14976 (vrax1q_u64): Ditto.
14977 (vxarq_u64): Ditto.
14978 (vbcaxq_u16): Ditto.
14979 * config/arm/types.md (crypto_sha512): New type attribute.
14980 (crypto_sha3): Ditto.
14981 (doc/invoke.texi): Document new sha3 option.
14983 2018-01-10 Michael Collison <michael.collison@arm.com>
14985 * config/aarch64/aarch64-builtins.c:
14986 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14987 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14988 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14989 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14990 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14991 (AARCH64_ISA_SM4): New ISA flag.
14992 (TARGET_SM4): New feature flag for sm4.
14993 * config/aarch64/aarch64-simd-builtins.def
14994 (aarch64_sm3ss1qv4si): Ditto.
14995 (aarch64_sm3tt1aq4si): Ditto.
14996 (aarch64_sm3tt1bq4si): Ditto.
14997 (aarch64_sm3tt2aq4si): Ditto.
14998 (aarch64_sm3tt2bq4si): Ditto.
14999 (aarch64_sm3partw1qv4si): Ditto.
15000 (aarch64_sm3partw2qv4si): Ditto.
15001 (aarch64_sm4eqv4si): Ditto.
15002 (aarch64_sm4ekeyqv4si): Ditto.
15003 * config/aarch64/aarch64-simd.md:
15004 (aarch64_sm3ss1qv4si): Ditto.
15005 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15006 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15007 (aarch64_sm4eqv4si): Ditto.
15008 (aarch64_sm4ekeyqv4si): Ditto.
15009 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15010 (sm3part_op): Ditto.
15011 (CRYPTO_SM3TT): Ditto.
15012 (CRYPTO_SM3PART): Ditto.
15013 (UNSPEC_SM3SS1): New unspec.
15014 (UNSPEC_SM3TT1A): Ditto.
15015 (UNSPEC_SM3TT1B): Ditto.
15016 (UNSPEC_SM3TT2A): Ditto.
15017 (UNSPEC_SM3TT2B): Ditto.
15018 (UNSPEC_SM3PARTW1): Ditto.
15019 (UNSPEC_SM3PARTW2): Ditto.
15020 (UNSPEC_SM4E): Ditto.
15021 (UNSPEC_SM4EKEY): Ditto.
15022 * config/aarch64/constraints.md (Ui2): New constraint.
15023 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15024 * config/arm/types.md (crypto_sm3): New type attribute.
15025 (crypto_sm4): Ditto.
15026 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15027 (vsm3tt1aq_u32): Ditto.
15028 (vsm3tt1bq_u32): Ditto.
15029 (vsm3tt2aq_u32): Ditto.
15030 (vsm3tt2bq_u32): Ditto.
15031 (vsm3partw1q_u32): Ditto.
15032 (vsm3partw2q_u32): Ditto.
15033 (vsm4eq_u32): Ditto.
15034 (vsm4ekeyq_u32): Ditto.
15035 (doc/invoke.texi): Document new sm4 option.
15037 2018-01-10 Michael Collison <michael.collison@arm.com>
15039 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15040 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15041 (AARCH64_FL_FOR_ARCH8_4): New.
15042 (AARCH64_FL_V8_4): New flag.
15043 (doc/invoke.texi): Document new armv8.4-a option.
15045 2018-01-10 Michael Collison <michael.collison@arm.com>
15047 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15048 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15049 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15050 * config/aarch64/aarch64-option-extension.def: Add
15051 AARCH64_OPT_EXTENSION of 'sha2'.
15052 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15053 (crypto): Disable sha2 and aes if crypto disabled.
15054 (crypto): Enable aes and sha2 if enabled.
15055 (simd): Disable sha2 and aes if simd disabled.
15056 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15058 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15059 (TARGET_SHA2): New feature flag for sha2.
15060 (TARGET_AES): New feature flag for aes.
15061 * config/aarch64/aarch64-simd.md:
15062 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15063 conditional on TARGET_AES.
15064 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15065 (aarch64_crypto_sha1hsi): Make pattern conditional
15067 (aarch64_crypto_sha1hv4si): Ditto.
15068 (aarch64_be_crypto_sha1hv4si): Ditto.
15069 (aarch64_crypto_sha1su1v4si): Ditto.
15070 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15071 (aarch64_crypto_sha1su0v4si): Ditto.
15072 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15073 (aarch64_crypto_sha256su0v4si): Ditto.
15074 (aarch64_crypto_sha256su1v4si): Ditto.
15075 (doc/invoke.texi): Document new aes and sha2 options.
15077 2018-01-10 Martin Sebor <msebor@redhat.com>
15079 PR tree-optimization/83781
15080 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15083 2018-01-11 Martin Sebor <msebor@gmail.com>
15084 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15086 PR tree-optimization/83501
15087 PR tree-optimization/81703
15089 * tree-ssa-strlen.c (get_string_cst): Rename...
15090 (get_string_len): ...to this. Handle global constants.
15091 (handle_char_store): Adjust.
15093 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15094 Jim Wilson <jimw@sifive.com>
15096 * config/riscv/riscv-protos.h (riscv_output_return): New.
15097 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15098 (riscv_attribute_table, riscv_output_return),
15099 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15100 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15101 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15103 (riscv_expand_prologue): Add early return for naked function.
15104 (riscv_expand_epilogue): Likewise.
15105 (riscv_function_ok_for_sibcall): Return false for naked function.
15106 (riscv_set_current_function): New.
15107 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15108 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15109 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15110 * doc/extend.texi (RISC-V Function Attributes): New.
15112 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15114 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15115 check for 128-bit long double before checking TCmode.
15116 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15117 128-bit long doubles before checking TFmode or TCmode.
15118 (FLOAT128_IBM_P): Likewise.
15120 2018-01-10 Martin Sebor <msebor@redhat.com>
15122 PR tree-optimization/83671
15123 * builtins.c (c_strlen): Unconditionally return zero for the empty
15125 Use -Warray-bounds for warnings.
15126 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15127 for non-constant array indices with COMPONENT_REF, arrays of
15128 arrays, and pointers to arrays.
15129 (gimple_fold_builtin_strlen): Determine and set length range for
15130 non-constant character arrays.
15132 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15134 PR middle-end/81897
15135 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15138 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15140 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15142 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15145 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15146 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15147 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15148 indexed_or_indirect_operand predicate.
15149 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15150 (*vsx_le_perm_load_v8hi): Likewise.
15151 (*vsx_le_perm_load_v16qi): Likewise.
15152 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15153 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15154 (*vsx_le_perm_store_v8hi): Likewise.
15155 (*vsx_le_perm_store_v16qi): Likewise.
15156 (eight unnamed splitters): Likewise.
15158 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15160 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15161 * config/rs6000/emmintrin.h: Likewise.
15162 * config/rs6000/mmintrin.h: Likewise.
15163 * config/rs6000/xmmintrin.h: Likewise.
15165 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15168 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15170 * tree.c (tree_nop_conversion): Return true for location wrapper
15172 (maybe_wrap_with_location): New function.
15173 (selftest::check_strip_nops): New function.
15174 (selftest::test_location_wrappers): New function.
15175 (selftest::tree_c_tests): Call it.
15176 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15177 (maybe_wrap_with_location): New decl.
15178 (EXPR_LOCATION_WRAPPER_P): New macro.
15179 (location_wrapper_p): New inline function.
15180 (tree_strip_any_location_wrapper): New inline function.
15182 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15185 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15186 stack_realign_offset for the largest alignment of stack slot
15188 (ix86_find_max_used_stack_alignment): New function.
15189 (ix86_finalize_stack_frame_flags): Use it. Set
15190 max_used_stack_alignment if we don't realign stack.
15191 * config/i386/i386.h (machine_function): Add
15192 max_used_stack_alignment.
15194 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15196 * config/arm/arm.opt (-mbranch-cost): New option.
15197 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15200 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15203 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15204 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15206 2018-01-10 Richard Biener <rguenther@suse.de>
15209 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15210 early out so it also covers the case where we have a non-NULL
15213 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15215 PR tree-optimization/83753
15216 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15217 for non-strided grouped accesses if the number of elements is 1.
15219 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15222 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15223 * i386.h (TARGET_USE_GATHER): Define.
15224 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15226 2018-01-10 Martin Liska <mliska@suse.cz>
15229 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15230 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15232 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15233 CLEANUP_NO_PARTITIONING is not set.
15235 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15237 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15238 for vectors, as a partial revert of r254296.
15239 * rtl.h (const_vec_p): Delete.
15240 (const_vec_duplicate_p): Don't test for vector CONSTs.
15241 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15242 * expmed.c (make_tree): Likewise.
15245 * common.md (E, F): Use CONSTANT_P instead of checking for
15247 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15248 checking for CONST_VECTOR.
15250 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15252 PR middle-end/83575
15253 * predict.c (force_edge_cold): Handle in more sane way edges
15254 with no prediction.
15256 2018-01-09 Carl Love <cel@us.ibm.com>
15258 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15260 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15261 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15262 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15263 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15264 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15265 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15266 * config/rs6000/rs6000-protos.h: Add extern defition for
15267 rs6000_generate_float2_double_code.
15268 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15270 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15271 (float2_v2df): Add define_expand.
15273 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15276 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15277 op_mode in the force_to_mode call.
15279 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15281 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15282 instead of checking each element individually.
15283 (aarch64_evpc_uzp): Likewise.
15284 (aarch64_evpc_zip): Likewise.
15285 (aarch64_evpc_ext): Likewise.
15286 (aarch64_evpc_rev): Likewise.
15287 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15288 instead of checking each element individually. Return true without
15290 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15291 whether all selected elements come from the same input, instead of
15292 checking each element individually. Remove calls to gen_rtx_REG,
15293 start_sequence and end_sequence and instead assert that no rtl is
15296 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15298 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15299 order of HIGH and CONST checks.
15301 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15303 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15304 if the destination isn't an SSA_NAME.
15306 2018-01-09 Richard Biener <rguenther@suse.de>
15308 PR tree-optimization/83668
15309 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15311 (canonicalize_loop_form): ... here, renamed from ...
15312 (canonicalize_loop_closed_ssa_form): ... this and amended to
15313 swap successor edges for loop exit blocks to make us use
15314 the RPO order we need for initial schedule generation.
15316 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15318 PR tree-optimization/64811
15319 * match.pd: When optimizing comparisons with Inf, avoid
15320 introducing or losing exceptions from comparisons with NaN.
15322 2018-01-09 Martin Liska <mliska@suse.cz>
15325 * asan.c (shadow_mem_size): Add gcc_assert.
15327 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15329 Don't save registers in main().
15332 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15333 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15334 * config/avr/avr.c (avr_set_current_function): Don't error if
15335 naked, OS_task or OS_main are specified at the same time.
15336 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15338 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15340 * common/config/avr/avr-common.c (avr_option_optimization_table):
15341 Switch on -mmain-is-OS_task for optimizing compilations.
15343 2018-01-09 Richard Biener <rguenther@suse.de>
15345 PR tree-optimization/83572
15346 * graphite.c: Include cfganal.h.
15347 (graphite_transform_loops): Connect infinite loops to exit
15348 and remove fake edges at the end.
15350 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15352 * ipa-inline.c (edge_badness): Revert accidental checkin.
15354 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15357 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15358 symbols; not inline clones.
15360 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15363 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15364 hard registers. Formatting fixes.
15366 PR preprocessor/83722
15367 * gcc.c (try_generate_repro): Pass
15368 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15369 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15372 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15373 Kito Cheng <kito.cheng@gmail.com>
15375 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15376 (riscv_leaf_function_p): Delete.
15377 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15379 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15381 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15383 (do_ifelse): New function.
15384 (do_isel): New function.
15385 (do_sub3): New function.
15386 (do_add3): New function.
15387 (do_load_mask_compare): New function.
15388 (do_overlap_load_compare): New function.
15389 (expand_compare_loop): New function.
15390 (expand_block_compare): Call expand_compare_loop() when appropriate.
15391 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15392 option description.
15393 (-mblock-compare-inline-loop-limit): New option.
15395 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15398 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15399 Reverse order of second and third operands in first alternative.
15400 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15401 of first and second elements in UNSPEC_VPERMR vector.
15402 (altivec_expand_vec_perm_le): Likewise.
15404 2018-01-08 Jeff Law <law@redhat.com>
15406 PR rtl-optimizatin/81308
15407 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15408 (process_switch): If group_case_labels makes a change, then set
15410 (pass_convert_switch::execute): If a switch is converted, then
15411 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15413 PR rtl-optimization/81308
15414 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15417 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15419 PR target/83663 - Revert r255946
15420 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15421 generation for cases where splatting a value is not useful.
15422 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15423 across a vec_duplicate and a paradoxical subreg forming a vector
15424 mode to a vec_concat.
15426 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15428 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15429 -march=armv8.3-a variants.
15430 * config/arm/t-multilib: Likewise.
15431 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15433 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15435 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15437 (cceq_ior_compare_complement): Give it a name so I can use it, and
15438 change boolean_or_operator predicate to boolean_operator so it can
15439 be used to generate a crand.
15440 (eqne): New code iterator.
15441 (bd/bd_neg): New code_attrs.
15442 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15443 a single define_insn.
15444 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15445 decrement (bdnzt/bdnzf/bdzt/bdzf).
15446 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15447 with the new names of the branch decrement patterns, and added the
15448 names of the branch decrement conditional patterns.
15450 2018-01-08 Richard Biener <rguenther@suse.de>
15452 PR tree-optimization/83563
15453 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15456 2018-01-08 Richard Biener <rguenther@suse.de>
15458 PR middle-end/83713
15459 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15461 2018-01-08 Richard Biener <rguenther@suse.de>
15463 PR tree-optimization/83685
15464 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15465 references to abnormals.
15467 2018-01-08 Richard Biener <rguenther@suse.de>
15470 * dwarf2out.c (output_indirect_strings): Handle empty
15471 skeleton_debug_str_hash.
15472 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15474 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15476 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15477 (emit_store_direct): Likewise.
15478 (arc_trampoline_adjust_address): Likewise.
15479 (arc_asm_trampoline_template): New function.
15480 (arc_initialize_trampoline): Use asm_trampoline_template.
15481 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15482 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15483 * config/arc/arc.md (flush_icache): Delete pattern.
15485 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15487 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15488 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15491 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15494 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15495 by not USED_FOR_TARGET.
15496 (make_pass_resolve_sw_modes): Likewise.
15498 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15500 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15503 2018-01-08 Richard Biener <rguenther@suse.de>
15505 PR middle-end/83580
15506 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15508 2018-01-08 Richard Biener <rguenther@suse.de>
15510 PR middle-end/83517
15511 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15513 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15515 PR middle-end/81897
15516 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15517 basic blocks with a small number of successors.
15518 (convert_control_dep_chain_into_preds): Improve handling of
15520 (dump_predicates): Split apart into...
15521 (dump_pred_chain): ...here...
15522 (dump_pred_info): ...and here.
15523 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15524 (can_chain_union_be_invalidated_p): Improve check for invalidation
15526 (uninit_uses_cannot_happen): Avoid unnecessary if
15527 convert_control_dep_chain_into_preds yielded nothing.
15529 2018-01-06 Martin Sebor <msebor@redhat.com>
15531 PR tree-optimization/83640
15532 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15533 subtracting negative offset from size.
15534 (builtin_access::overlap): Adjust offset bounds of the access to fall
15535 within the size of the object if possible.
15537 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15539 PR rtl-optimization/83699
15540 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15541 extract_bit_field_as_subreg to cases in which the extracted
15542 value is also a vector.
15544 * lra-constraints.c (process_alt_operands): Test for the equivalence
15545 substitutions when detecting a possible reload cycle.
15547 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15550 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15551 by default if flag_selective_schedling{,2}. Formatting fixes.
15553 PR rtl-optimization/83682
15554 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15555 if it has non-VECTOR_MODE element mode.
15556 (vec_duplicate_p): Likewise.
15558 PR middle-end/83694
15559 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15560 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15562 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15565 * config/i386/i386-builtin.def
15566 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15567 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15568 Require also OPTION_MASK_ISA_AVX512F in addition to
15569 OPTION_MASK_ISA_GFNI.
15570 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15571 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15572 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15573 to OPTION_MASK_ISA_GFNI.
15574 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15575 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15576 OPTION_MASK_ISA_AVX512BW.
15577 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15578 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15579 addition to OPTION_MASK_ISA_GFNI.
15580 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15581 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15582 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15583 to OPTION_MASK_ISA_GFNI.
15584 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15585 a requirement for all ISAs rather than any of them with a few
15587 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15589 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15590 bitmasks to be enabled with 3 exceptions, instead of requiring any
15591 enabled ISA with lots of exceptions.
15592 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15593 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15594 Change avx512bw in isa attribute to avx512f.
15595 * config/i386/sgxintrin.h: Add license boilerplate.
15596 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15597 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15598 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15599 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15601 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15602 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15603 temporarily sse2 rather than sse if not enabled already.
15606 * config/i386/sse.md (VI248_VLBW): Rename to ...
15607 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15608 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15609 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15610 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15611 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15612 mode iterator instead of VI248_VLBW.
15614 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15616 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15617 (record_modified): Skip clobbers; add debug output.
15618 (param_change_prob): Use sreal frequencies.
15620 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15622 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15623 punt for user-aligned variables.
15625 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15627 * tree-chrec.c (chrec_contains_symbols): Return true for
15630 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15633 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15634 of (x|y) == x for BICS pattern.
15636 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15638 PR tree-optimization/83605
15639 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15640 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15643 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15645 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15646 * config/epiphany/rtems.h: New file.
15648 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15649 Uros Bizjak <ubizjak@gmail.com>
15652 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15653 QIreg_operand instead of register_operand predicate.
15654 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15655 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15656 comments instead of -fmitigate[-_]rop.
15658 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15661 * cgraphunit.c (symbol_table::compile): Switch to text_section
15662 before calling assembly_start debug hook.
15663 * run-rtl-passes.c (run_rtl_passes): Likewise.
15666 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15668 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15669 range_int_cst_p rather than !symbolic_range_p before calling
15670 extract_range_from_multiplicative_op_1.
15672 2018-01-04 Jeff Law <law@redhat.com>
15674 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15675 redundant test in assertion.
15677 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15679 * doc/rtl.texi: Document machine_mode wrapper classes.
15681 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15683 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15684 using tree_to_uhwi.
15686 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15688 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15689 the VEC_PERM_EXPR fold to fail.
15691 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15694 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15695 to switched_sections.
15697 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15700 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15701 test for d.testing.
15703 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15706 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15707 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15709 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15712 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15713 is BLKmode and bitpos not zero or mode change is needed.
15715 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15718 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15721 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15724 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15725 instead of MULT rtx. Update all corresponding splitters.
15726 (*saddl_se): Ditto.
15727 (*ssub<modesuffix>): Ditto.
15728 (*ssubl_se): Ditto.
15729 (*cmp_sadd_di): Update split patterns.
15730 (*cmp_sadd_si): Ditto.
15731 (*cmp_sadd_sidi): Ditto.
15732 (*cmp_ssub_di): Ditto.
15733 (*cmp_ssub_si): Ditto.
15734 (*cmp_ssub_sidi): Ditto.
15735 * config/alpha/predicates.md (const23_operand): New predicate.
15736 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15737 Look for ASHIFT, not MULT inner operand.
15738 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15740 2018-01-04 Martin Liska <mliska@suse.cz>
15742 PR gcov-profile/83669
15743 * gcov.c (output_intermediate_file): Add version to intermediate
15745 * doc/gcov.texi: Document new field 'version' in intermediate
15746 file format. Fix location of '-k' option of gcov command.
15748 2018-01-04 Martin Liska <mliska@suse.cz>
15751 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15753 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15755 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15757 2018-01-03 Martin Sebor <msebor@redhat.com>
15759 PR tree-optimization/83655
15760 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15761 checking calls with invalid arguments.
15763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15765 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15766 (vectorizable_mask_load_store): Delete.
15767 (vectorizable_call): Return false for masked loads and stores.
15768 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15769 instead of gimple_assign_rhs1.
15770 (vectorizable_load): Handle IFN_MASK_LOAD.
15771 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15775 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15777 (vectorizable_mask_load_store): ...here.
15778 (vectorizable_load): ...and here.
15780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15782 * tree-vect-stmts.c (vect_build_all_ones_mask)
15783 (vect_build_zero_merge_argument): New functions, split out from...
15784 (vectorizable_load): ...here.
15786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15788 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15790 (vectorizable_mask_load_store): ...here.
15791 (vectorizable_store): ...and here.
15793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15795 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15797 (vectorizable_mask_load_store): ...here.
15799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15801 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15802 (vect_model_store_cost): Take a vec_load_store_type instead of a
15804 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15805 (vect_model_store_cost): Take a vec_load_store_type instead of a
15807 (vectorizable_mask_load_store): Update accordingly.
15808 (vectorizable_store): Likewise.
15809 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15813 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15814 IFN_MASK_LOAD calls here rather than...
15815 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
15822 fall back to extract_bit_field_as_subreg if vec_extract
15825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15826 Alan Hayward <alan.hayward@arm.com>
15827 David Sherwood <david.sherwood@arm.com>
15829 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15830 they are variable or constant sized.
15831 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15832 slots for constant-sized data.
15834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15835 Alan Hayward <alan.hayward@arm.com>
15836 David Sherwood <david.sherwood@arm.com>
15838 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15839 handling COND_EXPRs with boolean comparisons, try to find a better
15840 basis for the mask type than the boolean itself.
15842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15844 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15845 is calculated and how it can be overridden.
15846 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15847 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15849 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15853 Alan Hayward <alan.hayward@arm.com>
15854 David Sherwood <david.sherwood@arm.com>
15856 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15857 Remove the mode argument.
15858 (aarch64_simd_valid_immediate): Remove the mode and inverse
15860 * config/aarch64/iterators.md (bitsize): New iterator.
15861 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15862 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15863 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15864 aarch64_simd_valid_immediate.
15865 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15866 (aarch64_reg_or_bic_imm): Likewise.
15867 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15868 with an insn_type enum and msl with a modifier_type enum.
15869 Replace element_width with a scalar_mode. Change the shift
15870 to unsigned int. Add constructors for scalar_float_mode and
15871 scalar_int_mode elements.
15872 (aarch64_vect_float_const_representable_p): Delete.
15873 (aarch64_can_const_movi_rtx_p)
15874 (aarch64_simd_scalar_immediate_valid_for_move)
15875 (aarch64_simd_make_constant): Update call to
15876 aarch64_simd_valid_immediate.
15877 (aarch64_advsimd_valid_immediate_hs): New function.
15878 (aarch64_advsimd_valid_immediate): Likewise.
15879 (aarch64_simd_valid_immediate): Remove mode and inverse
15880 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15881 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15882 and aarch64_float_const_representable_p on the result.
15883 (aarch64_output_simd_mov_immediate): Remove mode argument.
15884 Update call to aarch64_simd_valid_immediate and use of
15885 simd_immediate_info.
15886 (aarch64_output_scalar_simd_mov_immediate): Update call
15889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15890 Alan Hayward <alan.hayward@arm.com>
15891 David Sherwood <david.sherwood@arm.com>
15893 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15894 (mode_nunits): Likewise CONST_MODE_NUNITS.
15895 * machmode.def (ADJUST_NUNITS): Document.
15896 * genmodes.c (mode_data::need_nunits_adj): New field.
15897 (blank_mode): Update accordingly.
15898 (adj_nunits): New variable.
15899 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15901 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15902 listed in adj_nunits.
15903 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15904 listed in adj_nunits. Don't emit case statements for such modes.
15905 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15906 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15907 nothing if adj_nunits is nonnull.
15908 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15909 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15910 (emit_mode_fbit): Update use of print_maybe_const_decl.
15911 (emit_move_size): Likewise. Treat the array as non-const
15913 (emit_mode_adjustments): Handle adj_nunits.
15915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15917 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15918 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15919 (VECTOR_MODES): Use it.
15920 (make_vector_modes): Take the prefix as an argument.
15922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15923 Alan Hayward <alan.hayward@arm.com>
15924 David Sherwood <david.sherwood@arm.com>
15926 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15927 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15928 for MODE_VECTOR_BOOL.
15929 * machmode.def (VECTOR_BOOL_MODE): Document.
15930 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15931 (make_vector_bool_mode): New function.
15932 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15934 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15935 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15937 * stor-layout.c (int_mode_for_mode): Likewise.
15938 * tree.c (build_vector_type_for_mode): Likewise.
15939 * varasm.c (output_constant_pool_2): Likewise.
15940 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15941 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15942 for MODE_VECTOR_BOOL.
15943 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15944 of mode class checks.
15945 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15946 instead of a list of mode class checks.
15947 (expand_vector_scalar_condition): Likewise.
15948 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15951 Alan Hayward <alan.hayward@arm.com>
15952 David Sherwood <david.sherwood@arm.com>
15954 * machmode.h (mode_size): Change from unsigned short to
15956 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15957 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15958 or if measurement_type is not polynomial.
15959 (fixed_size_mode::includes_p): Check for constant-sized modes.
15960 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15961 return a poly_uint16 rather than an unsigned short.
15962 (emit_mode_size): Change the type of mode_size from unsigned short
15963 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15964 (emit_mode_adjustments): Cope with polynomial vector sizes.
15965 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15967 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15969 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15970 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15971 * caller-save.c (setup_save_areas): Likewise.
15972 (replace_reg_with_saved_mem): Likewise.
15973 * calls.c (emit_library_call_value_1): Likewise.
15974 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15975 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15976 (gen_lowpart_for_combine): Likewise.
15977 * convert.c (convert_to_integer_1): Likewise.
15978 * cse.c (equiv_constant, cse_insn): Likewise.
15979 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15980 (cselib_subst_to_values): Likewise.
15981 * dce.c (word_dce_process_block): Likewise.
15982 * df-problems.c (df_word_lr_mark_ref): Likewise.
15983 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15984 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15985 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15986 (rtl_for_decl_location): Likewise.
15987 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15988 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15989 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15990 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15991 (expand_expr_real_1): Likewise.
15992 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15993 (pad_below): Likewise.
15994 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15995 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15996 * ira.c (get_subreg_tracking_sizes): Likewise.
15997 * ira-build.c (ira_create_allocno_objects): Likewise.
15998 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15999 (ira_sort_regnos_for_alter_reg): Likewise.
16000 * ira-costs.c (record_operand_costs): Likewise.
16001 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16002 (resolve_simple_move): Likewise.
16003 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16004 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16005 (lra_constraints): Likewise.
16006 (CONST_POOL_OK_P): Reject variable-sized modes.
16007 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16008 (add_pseudo_to_slot, lra_spill): Likewise.
16009 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16010 * optabs-query.c (get_best_extraction_insn): Likewise.
16011 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16012 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16013 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16014 * recog.c (offsettable_address_addr_space_p): Likewise.
16015 * regcprop.c (maybe_mode_change): Likewise.
16016 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16017 * regrename.c (build_def_use): Likewise.
16018 * regstat.c (dump_reg_info): Likewise.
16019 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16020 (find_reloads, find_reloads_subreg_address): Likewise.
16021 * reload1.c (eliminate_regs_1): Likewise.
16022 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16023 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16024 (simplify_binary_operation_1, simplify_subreg): Likewise.
16025 * targhooks.c (default_function_arg_padding): Likewise.
16026 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16027 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16028 (verify_gimple_assign_ternary): Likewise.
16029 * tree-inline.c (estimate_move_cost): Likewise.
16030 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16031 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16032 (get_address_cost_ainc): Likewise.
16033 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16034 (vect_supportable_dr_alignment): Likewise.
16035 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16036 (vectorizable_reduction): Likewise.
16037 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16038 (vectorizable_operation, vectorizable_load): Likewise.
16039 * tree.c (build_same_sized_truth_vector_type): Likewise.
16040 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16041 * var-tracking.c (emit_note_insn_var_location): Likewise.
16042 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16043 (ADDR_VEC_ALIGN): Likewise.
16045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16046 Alan Hayward <alan.hayward@arm.com>
16047 David Sherwood <david.sherwood@arm.com>
16049 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16051 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16052 or if measurement_type is polynomial.
16053 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16054 * combine.c (make_extraction): Likewise.
16055 * dse.c (find_shift_sequence): Likewise.
16056 * dwarf2out.c (mem_loc_descriptor): Likewise.
16057 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16058 (extract_bit_field, extract_low_bits): Likewise.
16059 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16060 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16061 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16062 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16063 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16064 * reload.c (find_reloads): Likewise.
16065 * reload1.c (alter_reg): Likewise.
16066 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16067 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16068 * tree-if-conv.c (predicate_mem_writes): Likewise.
16069 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16070 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16071 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16072 * valtrack.c (dead_debug_insert_temp): Likewise.
16073 * varasm.c (mergeable_constant_section): Likewise.
16074 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16077 Alan Hayward <alan.hayward@arm.com>
16078 David Sherwood <david.sherwood@arm.com>
16080 * expr.c (expand_assignment): Cope with polynomial mode sizes
16081 when assigning to a CONCAT.
16083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16084 Alan Hayward <alan.hayward@arm.com>
16085 David Sherwood <david.sherwood@arm.com>
16087 * machmode.h (mode_precision): Change from unsigned short to
16089 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16091 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16092 or if measurement_type is not polynomial.
16093 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16094 in which the mode is already known to be a scalar_int_mode.
16095 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16096 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16098 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16099 for GET_MODE_PRECISION.
16100 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16101 for GET_MODE_PRECISION.
16102 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16104 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16105 (expand_field_assignment, make_extraction): Likewise.
16106 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16107 (get_last_value): Likewise.
16108 * convert.c (convert_to_integer_1): Likewise.
16109 * cse.c (cse_insn): Likewise.
16110 * expr.c (expand_expr_real_1): Likewise.
16111 * lra-constraints.c (simplify_operand_subreg): Likewise.
16112 * optabs-query.c (can_atomic_load_p): Likewise.
16113 * optabs.c (expand_atomic_load): Likewise.
16114 (expand_atomic_store): Likewise.
16115 * ree.c (combine_reaching_defs): Likewise.
16116 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16117 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16118 * tree.h (type_has_mode_precision_p): Likewise.
16119 * ubsan.c (instrument_si_overflow): Likewise.
16121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16122 Alan Hayward <alan.hayward@arm.com>
16123 David Sherwood <david.sherwood@arm.com>
16125 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16126 polynomial numbers of units.
16127 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16128 (valid_vector_subparts_p): New function.
16129 (build_vector_type): Remove temporary shim and take the number
16130 of units as a poly_uint64 rather than an int.
16131 (build_opaque_vector_type): Take the number of units as a
16132 poly_uint64 rather than an int.
16133 * tree.c (build_vector_from_ctor): Handle polynomial
16134 TYPE_VECTOR_SUBPARTS.
16135 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16136 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16137 (build_vector_from_val): If the number of units is variable,
16138 use build_vec_duplicate_cst for constant operands and
16139 VEC_DUPLICATE_EXPR otherwise.
16140 (make_vector_type): Remove temporary is_constant ().
16141 (build_vector_type, build_opaque_vector_type): Take the number of
16142 units as a poly_uint64 rather than an int.
16143 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16145 * cfgexpand.c (expand_debug_expr): Likewise.
16146 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16147 (store_constructor, expand_expr_real_1): Likewise.
16148 (const_scalar_mask_from_tree): Likewise.
16149 * fold-const-call.c (fold_const_reduction): Likewise.
16150 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16151 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16152 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16153 (fold_relational_const): Likewise.
16154 (native_interpret_vector): Likewise. Change the size from an
16155 int to an unsigned int.
16156 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16157 TYPE_VECTOR_SUBPARTS.
16158 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16159 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16160 duplicating a non-constant operand into a variable-length vector.
16161 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16162 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16163 * ipa-icf.c (sem_variable::equals): Likewise.
16164 * match.pd: Likewise.
16165 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16166 * print-tree.c (print_node): Likewise.
16167 * stor-layout.c (layout_type): Likewise.
16168 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16169 * tree-cfg.c (verify_gimple_comparison): Likewise.
16170 (verify_gimple_assign_binary): Likewise.
16171 (verify_gimple_assign_ternary): Likewise.
16172 (verify_gimple_assign_single): Likewise.
16173 * tree-pretty-print.c (dump_generic_node): Likewise.
16174 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16175 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16176 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16177 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16178 (vect_shift_permute_load_chain): Likewise.
16179 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16180 (expand_vector_condition, optimize_vector_constructor): Likewise.
16181 (lower_vec_perm, get_compute_type): Likewise.
16182 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16183 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16184 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16185 (vect_recog_mask_conversion_pattern): Likewise.
16186 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16187 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16188 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16189 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16190 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16191 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16192 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16193 (supportable_widening_operation): Likewise.
16194 (supportable_narrowing_operation): Likewise.
16195 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16197 * varasm.c (output_constant): Likewise.
16199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16200 Alan Hayward <alan.hayward@arm.com>
16201 David Sherwood <david.sherwood@arm.com>
16203 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16204 so that both the length == 3 and length != 3 cases set up their
16205 own permute vectors. Add comments explaining why we know the
16206 number of elements is constant.
16207 (vect_permute_load_chain): Likewise.
16209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16210 Alan Hayward <alan.hayward@arm.com>
16211 David Sherwood <david.sherwood@arm.com>
16213 * machmode.h (mode_nunits): Change from unsigned char to
16215 (ONLY_FIXED_SIZE_MODES): New macro.
16216 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16217 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16218 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16220 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16221 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16222 or if measurement_type is not polynomial.
16223 * genmodes.c (ZERO_COEFFS): New macro.
16224 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16226 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16227 Use ZERO_COEFFS when emitting initializers.
16228 * data-streamer.h (bp_pack_poly_value): New function.
16229 (bp_unpack_poly_value): Likewise.
16230 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16231 for GET_MODE_NUNITS.
16232 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16233 for GET_MODE_NUNITS.
16234 * tree.c (make_vector_type): Remove temporary shim and make
16235 the real function take the number of units as a poly_uint64
16236 rather than an int.
16237 (build_vector_type_for_mode): Handle polynomial nunits.
16238 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16239 * emit-rtl.c (const_vec_series_p_1): Likewise.
16240 (gen_rtx_CONST_VECTOR): Likewise.
16241 * fold-const.c (test_vec_duplicate_folding): Likewise.
16242 * genrecog.c (validate_pattern): Likewise.
16243 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16244 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16245 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16246 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16247 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16248 * rtlanal.c (subreg_get_info): Likewise.
16249 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16250 (vect_grouped_load_supported): Likewise.
16251 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16252 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16253 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16254 (simplify_const_unary_operation, simplify_binary_operation_1)
16255 (simplify_const_binary_operation, simplify_ternary_operation)
16256 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16257 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16258 instead of CONST_VECTOR_NUNITS.
16259 * varasm.c (output_constant_pool_2): Likewise.
16260 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16261 explicit-encoded elements in the XVEC for variable-length vectors.
16263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16265 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16268 Alan Hayward <alan.hayward@arm.com>
16269 David Sherwood <david.sherwood@arm.com>
16271 * coretypes.h (fixed_size_mode): Declare.
16272 (fixed_size_mode_pod): New typedef.
16273 * builtins.h (target_builtins::x_apply_args_mode)
16274 (target_builtins::x_apply_result_mode): Change type to
16275 fixed_size_mode_pod.
16276 * builtins.c (apply_args_size, apply_result_size, result_vector)
16277 (expand_builtin_apply_args_1, expand_builtin_apply)
16278 (expand_builtin_return): Update accordingly.
16280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16282 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16283 * cselib.c (cselib_hash_rtx): Likewise.
16284 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16285 CONST_VECTOR encoding.
16287 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16288 Jeff Law <law@redhat.com>
16291 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16292 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16293 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16294 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16297 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16298 explicitly probe *sp in a noreturn function if there were any callee
16299 register saves or frame pointer is needed.
16301 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16304 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16305 BLKmode for ternary, binary or unary expressions.
16308 * var-tracking.c (delete_vta_debug_insn): New inline function.
16309 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16310 insns from get_insns () to NULL instead of each bb separately.
16311 Use delete_vta_debug_insn. No longer static.
16312 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16313 delete_vta_debug_insns callers.
16314 * rtl.h (delete_vta_debug_insns): Declare.
16315 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16316 instead of variable_tracking_main.
16318 2018-01-03 Martin Sebor <msebor@redhat.com>
16320 PR tree-optimization/83603
16321 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16322 arguments past the endof the argument list in functions declared
16323 without a prototype.
16324 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16325 Avoid checking when arguments are null.
16327 2018-01-03 Martin Sebor <msebor@redhat.com>
16330 * doc/extend.texi (attribute const): Fix a typo.
16331 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16332 issuing -Wsuggest-attribute for void functions.
16334 2018-01-03 Martin Sebor <msebor@redhat.com>
16336 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16337 offset_int::from instead of wide_int::to_shwi.
16338 (maybe_diag_overlap): Remove assertion.
16339 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16340 * gimple-ssa-sprintf.c (format_directive): Same.
16341 (parse_directive): Same.
16342 (sprintf_dom_walker::compute_format_length): Same.
16343 (try_substitute_return_value): Same.
16345 2018-01-03 Jeff Law <law@redhat.com>
16347 PR middle-end/83654
16348 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16349 non-constant residual for zero at runtime and avoid probing in
16350 that case. Reorganize code for trailing problem to mirror handling
16353 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16355 PR tree-optimization/83501
16356 * tree-ssa-strlen.c (get_string_cst): New.
16357 (handle_char_store): Call get_string_cst.
16359 2018-01-03 Martin Liska <mliska@suse.cz>
16361 PR tree-optimization/83593
16362 * tree-ssa-strlen.c: Include tree-cfg.h.
16363 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16364 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16365 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16367 (strlen_dom_walker::before_dom_children): Call
16368 gimple_purge_dead_eh_edges. Dump tranformation with details
16370 (strlen_dom_walker::before_dom_children): Update call by adding
16371 new argument cleanup_eh.
16372 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16374 2018-01-03 Martin Liska <mliska@suse.cz>
16377 * cif-code.def (VARIADIC_THUNK): New enum value.
16378 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16381 2018-01-03 Jan Beulich <jbeulich@suse.com>
16383 * sse.md (mov<mode>_internal): Tighten condition for when to use
16384 vmovdqu<ssescalarsize> for TI and OI modes.
16386 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16388 Update copyright years.
16390 2018-01-03 Martin Liska <mliska@suse.cz>
16393 * ipa-visibility.c (function_and_variable_visibility): Skip
16394 functions with noipa attribure.
16396 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16398 * gcc.c (process_command): Update copyright notice dates.
16399 * gcov-dump.c (print_version): Ditto.
16400 * gcov.c (print_version): Ditto.
16401 * gcov-tool.c (print_version): Ditto.
16402 * gengtype.c (create_file): Ditto.
16403 * doc/cpp.texi: Bump @copying's copyright year.
16404 * doc/cppinternals.texi: Ditto.
16405 * doc/gcc.texi: Ditto.
16406 * doc/gccint.texi: Ditto.
16407 * doc/gcov.texi: Ditto.
16408 * doc/install.texi: Ditto.
16409 * doc/invoke.texi: Ditto.
16411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16413 * vector-builder.h (vector_builder::m_full_nelts): Change from
16414 unsigned int to poly_uint64.
16415 (vector_builder::full_nelts): Update prototype accordingly.
16416 (vector_builder::new_vector): Likewise.
16417 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16418 (vector_builder::operator ==): Likewise.
16419 (vector_builder::finalize): Likewise.
16420 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16421 Take the number of elements as a poly_uint64 rather than an
16423 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16424 from unsigned int to poly_uint64.
16425 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16426 (vec_perm_indices::new_vector): Likewise.
16427 (vec_perm_indices::length): Likewise.
16428 (vec_perm_indices::nelts_per_input): Likewise.
16429 (vec_perm_indices::input_nelts): Likewise.
16430 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16431 number of elements per input as a poly_uint64 rather than an
16432 unsigned int. Use the original encoding for variable-length
16433 vectors, rather than clamping each individual element.
16434 For the second and subsequent elements in each pattern,
16435 clamp the step and base before clamping their sum.
16436 (vec_perm_indices::series_p): Handle polynomial element counts.
16437 (vec_perm_indices::all_in_range_p): Likewise.
16438 (vec_perm_indices_to_tree): Likewise.
16439 (vec_perm_indices_to_rtx): Likewise.
16440 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16441 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16442 (tree_vector_builder::new_binary_operation): Handle polynomial
16443 element counts. Return false if we need to know the number
16444 of elements at compile time.
16445 * fold-const.c (fold_vec_perm): Punt if the number of elements
16446 isn't known at compile time.
16448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16450 * vec-perm-indices.h (vec_perm_builder): Change element type
16451 from HOST_WIDE_INT to poly_int64.
16452 (vec_perm_indices::element_type): Update accordingly.
16453 (vec_perm_indices::clamp): Handle polynomial element_types.
16454 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16455 (vec_perm_indices::all_in_range_p): Likewise.
16456 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16458 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16459 polynomial vec_perm_indices element types.
16460 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16461 * fold-const.c (fold_vec_perm): Likewise.
16462 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16463 * tree-vect-generic.c (lower_vec_perm): Likewise.
16464 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16465 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16466 element type to HOST_WIDE_INT.
16468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16469 Alan Hayward <alan.hayward@arm.com>
16470 David Sherwood <david.sherwood@arm.com>
16472 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16473 rather than an int. Use plus_constant.
16474 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16475 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16478 Alan Hayward <alan.hayward@arm.com>
16479 David Sherwood <david.sherwood@arm.com>
16481 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16482 a HOST_WIDE_INT to a poly_int64.
16484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16485 Alan Hayward <alan.hayward@arm.com>
16486 David Sherwood <david.sherwood@arm.com>
16488 * calls.c (load_register_parameters): Cope with polynomial
16489 mode sizes. Require a constant size for BLKmode parameters
16490 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16491 forces a parameter to be padded at the lsb end in order to
16492 fill a complete number of words, require the parameter size
16493 to be ordered wrt UNITS_PER_WORD.
16495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16496 Alan Hayward <alan.hayward@arm.com>
16497 David Sherwood <david.sherwood@arm.com>
16499 * reload1.c (spill_stack_slot_width): Change element type
16500 from unsigned int to poly_uint64_pod.
16501 (alter_reg): Treat mode sizes as polynomial.
16503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16504 Alan Hayward <alan.hayward@arm.com>
16505 David Sherwood <david.sherwood@arm.com>
16507 * reload.c (complex_word_subreg_p): New function.
16508 (reload_inner_reg_of_subreg, push_reload): Use it.
16510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16511 Alan Hayward <alan.hayward@arm.com>
16512 David Sherwood <david.sherwood@arm.com>
16514 * lra-constraints.c (process_alt_operands): Reject matched
16515 operands whose sizes aren't ordered.
16516 (match_reload): Refer to this check here.
16518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16519 Alan Hayward <alan.hayward@arm.com>
16520 David Sherwood <david.sherwood@arm.com>
16522 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16523 that the mode size is in the set {1, 2, 4, 8, 16}.
16525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16526 Alan Hayward <alan.hayward@arm.com>
16527 David Sherwood <david.sherwood@arm.com>
16529 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16530 Use plus_constant instead of gen_rtx_PLUS.
16532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16533 Alan Hayward <alan.hayward@arm.com>
16534 David Sherwood <david.sherwood@arm.com>
16536 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16537 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16538 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16539 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16540 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16541 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16542 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16543 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16544 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16545 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16547 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16548 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16549 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16550 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16551 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16552 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16553 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16554 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16555 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16556 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16558 * expr.c (emit_move_resolve_push): Treat the input and result
16559 of PUSH_ROUNDING as a poly_int64.
16560 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16561 (emit_push_insn): Likewise.
16562 * lra-eliminations.c (mark_not_eliminable): Likewise.
16563 * recog.c (push_operand): Likewise.
16564 * reload1.c (elimination_effects): Likewise.
16565 * rtlanal.c (nonzero_bits1): Likewise.
16566 * calls.c (store_one_arg): Likewise. Require the padding to be
16567 known at compile time.
16569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16570 Alan Hayward <alan.hayward@arm.com>
16571 David Sherwood <david.sherwood@arm.com>
16573 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16574 Use plus_constant instead of gen_rtx_PLUS.
16576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16577 Alan Hayward <alan.hayward@arm.com>
16578 David Sherwood <david.sherwood@arm.com>
16580 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16581 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16588 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16589 via stack temporaries. Treat the mode size as polynomial too.
16591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16592 Alan Hayward <alan.hayward@arm.com>
16593 David Sherwood <david.sherwood@arm.com>
16595 * expr.c (expand_expr_real_2): When handling conversions involving
16596 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16597 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16598 as a poly_uint64 too.
16600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16601 Alan Hayward <alan.hayward@arm.com>
16602 David Sherwood <david.sherwood@arm.com>
16604 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16607 Alan Hayward <alan.hayward@arm.com>
16608 David Sherwood <david.sherwood@arm.com>
16610 * combine.c (can_change_dest_mode): Handle polynomial
16611 REGMODE_NATURAL_SIZE.
16612 * expmed.c (store_bit_field_1): Likewise.
16613 * expr.c (store_constructor): Likewise.
16614 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16615 and polynomial REGMODE_NATURAL_SIZE.
16616 (gen_lowpart_common): Likewise.
16617 * reginfo.c (record_subregs_of_mode): Likewise.
16618 * rtlanal.c (read_modify_subreg_p): Likewise.
16620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16621 Alan Hayward <alan.hayward@arm.com>
16622 David Sherwood <david.sherwood@arm.com>
16624 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16625 numbers of elements.
16627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16628 Alan Hayward <alan.hayward@arm.com>
16629 David Sherwood <david.sherwood@arm.com>
16631 * match.pd: Cope with polynomial numbers of vector elements.
16633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16634 Alan Hayward <alan.hayward@arm.com>
16635 David Sherwood <david.sherwood@arm.com>
16637 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16638 in a POINTER_PLUS_EXPR.
16640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16641 Alan Hayward <alan.hayward@arm.com>
16642 David Sherwood <david.sherwood@arm.com>
16644 * omp-simd-clone.c (simd_clone_subparts): New function.
16645 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16646 (ipa_simd_modify_function_body): Likewise.
16648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16649 Alan Hayward <alan.hayward@arm.com>
16650 David Sherwood <david.sherwood@arm.com>
16652 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16653 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16654 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16655 (expand_vector_condition, vector_element): Likewise.
16656 (subparts_gt): New function.
16657 (get_compute_type): Use subparts_gt.
16658 (count_type_subparts): Delete.
16659 (expand_vector_operations_1): Use subparts_gt instead of
16660 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
16667 (vect_compile_time_alias): ...this new function. Do the calculation
16668 on poly_ints rather than trees.
16669 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
16677 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
16684 constant and extern definitions for variable-length vectors.
16685 (vect_get_constant_vectors): Note that the number of units
16686 is known to be constant.
16688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16689 Alan Hayward <alan.hayward@arm.com>
16690 David Sherwood <david.sherwood@arm.com>
16692 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16693 of units as polynomial. Choose between WIDE and NARROW based
16696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16697 Alan Hayward <alan.hayward@arm.com>
16698 David Sherwood <david.sherwood@arm.com>
16700 * tree-vect-stmts.c (simd_clone_subparts): New function.
16701 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16704 Alan Hayward <alan.hayward@arm.com>
16705 David Sherwood <david.sherwood@arm.com>
16707 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16708 vectors as polynomial. Use build_index_vector for
16709 IFN_GOMP_SIMD_LANE.
16711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16712 Alan Hayward <alan.hayward@arm.com>
16713 David Sherwood <david.sherwood@arm.com>
16715 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16716 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16717 for variable-length vectors.
16718 (vectorizable_mask_load_store): Treat the number of units as
16719 polynomial, asserting that it is constant if the condition has
16720 already been enforced.
16721 (vectorizable_store, vectorizable_load): Likewise.
16723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16724 Alan Hayward <alan.hayward@arm.com>
16725 David Sherwood <david.sherwood@arm.com>
16727 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16728 of units as polynomial. Punt if we can't tell at compile time
16729 which vector contains the final result.
16731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16732 Alan Hayward <alan.hayward@arm.com>
16733 David Sherwood <david.sherwood@arm.com>
16735 * tree-vect-loop.c (vectorizable_induction): Treat the number
16736 of units as polynomial. Punt on SLP inductions. Use an integer
16737 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16738 cast of such a series for variable-length floating-point
16741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16742 Alan Hayward <alan.hayward@arm.com>
16743 David Sherwood <david.sherwood@arm.com>
16745 * tree.h (build_index_vector): Declare.
16746 * tree.c (build_index_vector): New function.
16747 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16748 of units as polynomial, forcibly converting it to a constant if
16749 vectorizable_reduction has already enforced the condition.
16750 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16751 to create a {1,2,3,...} vector.
16752 (vectorizable_reduction): Treat the number of units as polynomial.
16753 Choose vectype_in based on the largest scalar element size rather
16754 than the smallest number of units. Enforce the restrictions
16757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16758 Alan Hayward <alan.hayward@arm.com>
16759 David Sherwood <david.sherwood@arm.com>
16761 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16762 number of units as polynomial.
16764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16765 Alan Hayward <alan.hayward@arm.com>
16766 David Sherwood <david.sherwood@arm.com>
16768 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16769 * target.def (autovectorize_vector_sizes): Return the vector sizes
16770 by pointer, using vector_sizes rather than a bitmask.
16771 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16772 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16773 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16775 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16776 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16777 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16778 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16779 * omp-general.c (omp_max_vf): Likewise.
16780 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16781 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16782 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16783 * tree-vect-slp.c (vect_slp_bb): Likewise.
16784 * doc/tm.texi: Regenerate.
16785 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16787 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16788 the vector size as a poly_uint64 rather than an unsigned int.
16789 (current_vector_size): Change from an unsigned int to a poly_uint64.
16790 (get_vectype_for_scalar_type): Update accordingly.
16791 * tree.h (build_truth_vector_type): Take the size and number of
16792 units as a poly_uint64 rather than an unsigned int.
16793 (build_vector_type): Add a temporary overload that takes
16794 the number of units as a poly_uint64 rather than an unsigned int.
16795 * tree.c (make_vector_type): Likewise.
16796 (build_truth_vector_type): Take the number of units as a poly_uint64
16797 rather than an unsigned int.
16799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16800 Alan Hayward <alan.hayward@arm.com>
16801 David Sherwood <david.sherwood@arm.com>
16803 * target.def (get_mask_mode): Take the number of units and length
16804 as poly_uint64s rather than unsigned ints.
16805 * targhooks.h (default_get_mask_mode): Update accordingly.
16806 * targhooks.c (default_get_mask_mode): Likewise.
16807 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16808 * doc/tm.texi: Regenerate.
16810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16811 Alan Hayward <alan.hayward@arm.com>
16812 David Sherwood <david.sherwood@arm.com>
16814 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16815 * omp-general.c (omp_max_vf): Likewise.
16816 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16817 (expand_omp_simd): Handle polynomial safelen.
16818 * omp-low.c (omplow_simd_context): Add a default constructor.
16819 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16820 (lower_rec_simd_input_clauses): Update accordingly.
16821 (lower_rec_input_clauses): Likewise.
16823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16824 Alan Hayward <alan.hayward@arm.com>
16825 David Sherwood <david.sherwood@arm.com>
16827 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16828 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16829 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16830 (vect_analyze_slp_cost): Likewise.
16831 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16832 (vect_model_load_cost): Likewise.
16834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16835 Alan Hayward <alan.hayward@arm.com>
16836 David Sherwood <david.sherwood@arm.com>
16838 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16839 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16840 from an unsigned int * to a poly_uint64_pod *.
16841 (calculate_unrolling_factor): New function.
16842 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16845 Alan Hayward <alan.hayward@arm.com>
16846 David Sherwood <david.sherwood@arm.com>
16848 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16849 from an unsigned int to a poly_uint64.
16850 (_loop_vec_info::slp_unrolling_factor): Likewise.
16851 (_loop_vec_info::vectorization_factor): Change from an int
16853 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16854 (vect_get_num_vectors): New function.
16855 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16856 (vect_get_num_copies): Use vect_get_num_vectors.
16857 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16858 to an unsigned int *.
16859 (vect_analyze_data_refs): Change min_vf from an int * to a
16861 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16862 than an unsigned HOST_WIDE_INT.
16863 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16864 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16865 to an unsigned int *.
16866 (vect_analyze_data_ref_dependences): Likewise.
16867 (vect_compute_data_ref_alignment): Handle polynomial vf.
16868 (vect_enhance_data_refs_alignment): Likewise.
16869 (vect_prune_runtime_alias_test_list): Likewise.
16870 (vect_shift_permute_load_chain): Likewise.
16871 (vect_supportable_dr_alignment): Likewise.
16872 (dependence_distance_ge_vf): Take the vectorization factor as a
16873 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16874 (vect_analyze_data_refs): Change min_vf from an int * to a
16876 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16877 vfm1 as a poly_uint64 rather than an int. Make the same change
16878 for the returned bound_scalar.
16879 (vect_gen_vector_loop_niters): Handle polynomial vf.
16880 (vect_do_peeling): Likewise. Update call to
16881 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16882 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16884 * tree-vect-loop.c (vect_determine_vectorization_factor)
16885 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16886 (vect_get_known_peeling_cost): Likewise.
16887 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16888 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16889 (vect_transform_loop): Likewise. Use the lowest possible VF when
16890 updating the upper bounds of the loop.
16891 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16892 rather than an int.
16893 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16894 polynomial unroll factors.
16895 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16896 (vect_make_slp_decision): Likewise.
16897 (vect_supported_load_permutation_p): Likewise, and polynomial
16899 (vect_analyze_slp_cost): Handle polynomial vf.
16900 (vect_slp_analyze_node_operations): Likewise.
16901 (vect_slp_analyze_bb_1): Likewise.
16902 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16903 than an unsigned HOST_WIDE_INT.
16904 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16905 (vectorizable_load): Handle polynomial vf.
16906 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16908 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16911 Alan Hayward <alan.hayward@arm.com>
16912 David Sherwood <david.sherwood@arm.com>
16914 * match.pd: Handle bit operations involving three constants
16915 and try to fold one pair.
16917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16919 * tree-vect-loop-manip.c: Include gimple-fold.h.
16920 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16921 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16922 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16923 Add a path that uses a step of VF instead of 1, but disable it
16925 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16926 and niters_no_overflow parameters. Update calls to
16927 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16928 Create a new SSA name if the latter choses to use a ste other
16929 than zero, and return it via niters_vector_mult_vf_var.
16930 * tree-vect-loop.c (vect_transform_loop): Update calls to
16931 vect_do_peeling, vect_gen_vector_loop_niters and
16932 slpeel_make_loop_iterate_ntimes.
16933 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16934 (vect_gen_vector_loop_niters): Update declarations after above changes.
16936 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16938 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16939 128-bit round to integer instructions.
16940 (ceil<mode>2): Likewise.
16941 (btrunc<mode>2): Likewise.
16942 (round<mode>2): Likewise.
16944 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16946 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16947 unaligned VSX load/store on P8/P9.
16948 (expand_block_clear): Allow the use of unaligned VSX
16949 load/store on P8/P9.
16951 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16953 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16955 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16956 swap associated with both a load and a store.
16958 2018-01-02 Andrew Waterman <andrew@sifive.com>
16960 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16961 * config/riscv/riscv.md (clear_cache): Use it.
16963 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16965 * web.c: Remove out-of-date comment.
16967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16969 * expr.c (fixup_args_size_notes): Check that any existing
16970 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16971 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16972 (emit_single_push_insn): ...here.
16974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16976 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16977 (const_vector_encoded_nelts): New function.
16978 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16979 (const_vector_int_elt, const_vector_elt): Declare.
16980 * emit-rtl.c (const_vector_int_elt_1): New function.
16981 (const_vector_elt): Likewise.
16982 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16983 of CONST_VECTOR_ELT.
16985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16987 * expr.c: Include rtx-vector-builder.h.
16988 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16989 directly on the tree encoding.
16990 (const_vector_from_tree): Likewise.
16991 * optabs.c: Include rtx-vector-builder.h.
16992 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16993 sequence of "u" values.
16994 * vec-perm-indices.c: Include rtx-vector-builder.h.
16995 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16996 directly on the vec_perm_indices encoding.
16998 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17000 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17001 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17002 * rtx-vector-builder.h: New file.
17003 * rtx-vector-builder.c: Likewise.
17004 * rtl.h (rtx_def::u2): Add a const_vector field.
17005 (CONST_VECTOR_NPATTERNS): New macro.
17006 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17007 (CONST_VECTOR_DUPLICATE_P): Likewise.
17008 (CONST_VECTOR_STEPPED_P): Likewise.
17009 (CONST_VECTOR_ENCODED_ELT): Likewise.
17010 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17011 (unwrap_const_vec_duplicate): Likewise.
17012 (const_vec_series_p): Check for a non-duplicated vector encoding.
17013 Say that the function only returns true for integer vectors.
17014 * emit-rtl.c: Include rtx-vector-builder.h.
17015 (gen_const_vec_duplicate_1): Delete.
17016 (gen_const_vector): Call gen_const_vec_duplicate instead of
17017 gen_const_vec_duplicate_1.
17018 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17019 (gen_const_vec_duplicate): Use rtx_vector_builder.
17020 (gen_const_vec_series): Likewise.
17021 (gen_rtx_CONST_VECTOR): Likewise.
17022 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17023 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17024 Build a new vector rather than modifying a CONST_VECTOR in-place.
17025 (handle_special_swappables): Update call accordingly.
17026 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17027 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17028 Build a new vector rather than modifying a CONST_VECTOR in-place.
17029 (handle_special_swappables): Update call accordingly.
17031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17033 * simplify-rtx.c (simplify_const_binary_operation): Use
17034 CONST_VECTOR_ELT instead of XVECEXP.
17036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17038 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17039 the selector elements to be different from the data elements
17040 if the selector is a VECTOR_CST.
17041 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17042 ssizetype for the selector.
17044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17046 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17047 before testing each element individually.
17048 * tree-vect-generic.c (lower_vec_perm): Likewise.
17050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17052 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17053 * selftest-run-tests.c (selftest::run_tests): Call it.
17054 * vector-builder.h (vector_builder::operator ==): New function.
17055 (vector_builder::operator !=): Likewise.
17056 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17057 (vec_perm_indices::all_from_input_p): New function.
17058 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17059 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17060 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17061 instead of reading the VECTOR_CST directly. Detect whether both
17062 vector inputs are the same before constructing the vec_perm_indices,
17063 and update the number of inputs argument accordingly. Use the
17064 utility functions added above. Only construct sel2 if we need to.
17066 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17068 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17069 the broadcast of the low byte.
17070 (expand_mult_highpart): Use an explicit encoding for the permutes.
17071 * optabs-query.c (can_mult_highpart_p): Likewise.
17072 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17073 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17074 (vectorizable_bswap): Likewise.
17075 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17076 explicit encoding for the power-of-2 permutes.
17077 (vect_permute_store_chain): Likewise.
17078 (vect_grouped_load_supported): Likewise.
17079 (vect_permute_load_chain): Likewise.
17081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17083 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17084 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17085 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17086 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17087 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17088 (vect_gen_perm_mask_any): Likewise.
17090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17092 * int-vector-builder.h: New file.
17093 * vec-perm-indices.h: Include int-vector-builder.h.
17094 (vec_perm_indices): Redefine as an int_vector_builder.
17095 (auto_vec_perm_indices): Delete.
17096 (vec_perm_builder): Redefine as a stand-alone class.
17097 (vec_perm_indices::vec_perm_indices): New function.
17098 (vec_perm_indices::clamp): Likewise.
17099 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17100 (vec_perm_indices::new_vector): New function.
17101 (vec_perm_indices::new_expanded_vector): Update for new
17102 vec_perm_indices class.
17103 (vec_perm_indices::rotate_inputs): New function.
17104 (vec_perm_indices::all_in_range_p): Operate directly on the
17105 encoded form, without computing elided elements.
17106 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17107 encoding. Update for new vec_perm_indices class.
17108 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17109 the given vec_perm_builder.
17110 (expand_vec_perm_var): Update vec_perm_builder constructor.
17111 (expand_mult_highpart): Use vec_perm_builder instead of
17112 auto_vec_perm_indices.
17113 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17114 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17115 or double series encoding as appropriate.
17116 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17117 vec_perm_indices instead of auto_vec_perm_indices.
17118 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17119 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17120 (vect_permute_store_chain): Likewise.
17121 (vect_grouped_load_supported): Likewise.
17122 (vect_permute_load_chain): Likewise.
17123 (vect_shift_permute_load_chain): Likewise.
17124 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17125 (vect_transform_slp_perm_load): Likewise.
17126 (vect_schedule_slp_instance): Likewise.
17127 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17128 (vectorizable_mask_load_store): Likewise.
17129 (vectorizable_bswap): Likewise.
17130 (vectorizable_store): Likewise.
17131 (vectorizable_load): Likewise.
17132 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17133 vec_perm_indices instead of auto_vec_perm_indices. Use
17134 tree_to_vec_perm_builder to read the vector from a tree.
17135 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17136 vec_perm_builder instead of a vec_perm_indices.
17137 (have_whole_vector_shift): Use vec_perm_builder and
17138 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17139 truncation to calc_vec_perm_mask_for_shift.
17140 (vect_create_epilog_for_reduction): Likewise.
17141 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17142 from auto_vec_perm_indices to vec_perm_indices.
17143 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17144 instead of changing individual elements.
17145 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17146 the vector in d.perm.
17147 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17148 from auto_vec_perm_indices to vec_perm_indices.
17149 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17150 instead of changing individual elements.
17151 (arm_vectorize_vec_perm_const): Use new_vector to install
17152 the vector in d.perm.
17153 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17154 Update vec_perm_builder constructor.
17155 (rs6000_expand_interleave): Likewise.
17156 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17157 (rs6000_expand_interleave): Likewise.
17159 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17161 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17162 to qimode could truncate the indices.
17163 * optabs.c (expand_vec_perm_var): Likewise.
17165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17167 * Makefile.in (OBJS): Add vec-perm-indices.o.
17168 * vec-perm-indices.h: New file.
17169 * vec-perm-indices.c: Likewise.
17170 * target.h (vec_perm_indices): Replace with a forward class
17172 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17173 * optabs.h: Include vec-perm-indices.h.
17174 (expand_vec_perm): Delete.
17175 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17176 (expand_vec_perm_const): Declare.
17177 * target.def (vec_perm_const_ok): Replace with...
17178 (vec_perm_const): ...this new hook.
17179 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17180 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17181 * doc/tm.texi: Regenerate.
17182 * optabs.def (vec_perm_const): Delete.
17183 * doc/md.texi (vec_perm_const): Likewise.
17184 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17185 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17186 expand_vec_perm for constant permutation vectors. Assert that
17187 the mode of variable permutation vectors is the integer equivalent
17188 of the mode that is being permuted.
17189 * optabs-query.h (selector_fits_mode_p): Declare.
17190 * optabs-query.c: Include vec-perm-indices.h.
17191 (selector_fits_mode_p): New function.
17192 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17193 is defined, instead of checking whether the vec_perm_const_optab
17194 exists. Use targetm.vectorize.vec_perm_const instead of
17195 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17196 fit in the vector mode before using a variable permute.
17197 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17198 vec_perm_indices instead of an rtx.
17199 (expand_vec_perm): Replace with...
17200 (expand_vec_perm_const): ...this new function. Take the selector
17201 as a vec_perm_indices rather than an rtx. Also take the mode of
17202 the selector. Update call to shift_amt_for_vec_perm_mask.
17203 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17204 Use vec_perm_indices::new_expanded_vector to expand the original
17205 selector into bytes. Check whether the indices fit in the vector
17206 mode before using a variable permute.
17207 (expand_vec_perm_var): Make global.
17208 (expand_mult_highpart): Use expand_vec_perm_const.
17209 * fold-const.c: Includes vec-perm-indices.h.
17210 * tree-ssa-forwprop.c: Likewise.
17211 * tree-vect-data-refs.c: Likewise.
17212 * tree-vect-generic.c: Likewise.
17213 * tree-vect-loop.c: Likewise.
17214 * tree-vect-slp.c: Likewise.
17215 * tree-vect-stmts.c: Likewise.
17216 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17218 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17219 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17220 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17221 (aarch64_vectorize_vec_perm_const): ...this new function.
17222 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17223 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17224 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17225 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17226 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17227 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17228 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17230 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17231 check for NEON modes.
17232 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17233 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17234 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17235 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17237 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17238 the old VEC_PERM_CONST conditions.
17239 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17240 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17241 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17242 (ia64_vectorize_vec_perm_const_ok): Merge into...
17243 (ia64_vectorize_vec_perm_const): ...this new function.
17244 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17245 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17246 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17247 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17248 * config/mips/mips.c (mips_expand_vec_perm_const)
17249 (mips_vectorize_vec_perm_const_ok): Merge into...
17250 (mips_vectorize_vec_perm_const): ...this new function.
17251 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17252 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17253 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17254 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17255 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17256 (rs6000_expand_vec_perm_const): Delete.
17257 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17259 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17260 (altivec_expand_vec_perm_const_le): Take each operand individually.
17261 Operate on constant selectors rather than rtxes.
17262 (altivec_expand_vec_perm_const): Likewise. Update call to
17263 altivec_expand_vec_perm_const_le.
17264 (rs6000_expand_vec_perm_const): Delete.
17265 (rs6000_vectorize_vec_perm_const_ok): Delete.
17266 (rs6000_vectorize_vec_perm_const): New function.
17267 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17268 an element count and rtx array.
17269 (rs6000_expand_extract_even): Update call accordingly.
17270 (rs6000_expand_interleave): Likewise.
17271 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17272 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17273 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17274 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17275 (rs6000_expand_vec_perm_const): Delete.
17276 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17277 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17278 (altivec_expand_vec_perm_const_le): Take each operand individually.
17279 Operate on constant selectors rather than rtxes.
17280 (altivec_expand_vec_perm_const): Likewise. Update call to
17281 altivec_expand_vec_perm_const_le.
17282 (rs6000_expand_vec_perm_const): Delete.
17283 (rs6000_vectorize_vec_perm_const_ok): Delete.
17284 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17285 reference to the SPE evmerge intructions.
17286 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17287 an element count and rtx array.
17288 (rs6000_expand_extract_even): Update call accordingly.
17289 (rs6000_expand_interleave): Likewise.
17290 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17291 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17293 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17297 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17298 vector mode and that that mode matches the mode of the data
17300 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17301 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17302 directly using expand_vec_perm_1 when forcing selectors into
17304 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17306 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17308 * optabs-query.h (can_vec_perm_p): Delete.
17309 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17310 * optabs-query.c (can_vec_perm_p): Split into...
17311 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17312 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17313 particular selector is valid.
17314 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17315 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17316 (vect_grouped_load_supported): Likewise.
17317 (vect_shift_permute_load_chain): Likewise.
17318 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17319 (vect_transform_slp_perm_load): Likewise.
17320 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17321 (vectorizable_bswap): Likewise.
17322 (vect_gen_perm_mask_checked): Likewise.
17323 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17324 implementations of variable permutation vectors into account
17325 when deciding which selector to use.
17326 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17327 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17328 with a false third argument.
17329 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17330 to test whether the constant selector is valid and can_vec_perm_var_p
17331 to test whether a variable selector is valid.
17333 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17335 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17336 * optabs-query.c (can_vec_perm_p): Likewise.
17337 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17338 instead of vec_perm_indices.
17339 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17340 (vect_gen_perm_mask_checked): Likewise,
17341 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17342 (vect_gen_perm_mask_checked): Likewise,
17344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17346 * optabs-query.h (qimode_for_vec_perm): Declare.
17347 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17348 (qimode_for_vec_perm): ...this new function.
17349 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17351 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17353 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17354 does not have a conditional at the top.
17356 2018-01-02 Richard Biener <rguenther@suse.de>
17358 * ipa-inline.c (big_speedup_p): Fix expression.
17360 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17363 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17366 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17370 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17371 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17372 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17373 cond_taken_branch_cost 3->4.
17375 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17377 PR tree-optimization/83581
17378 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17379 TODO_cleanup_cfg if any changes have been made.
17381 PR middle-end/83608
17382 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17383 convert_modes if target mode has the right side, but different mode
17386 PR middle-end/83609
17387 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17388 last argument when extracting from CONCAT. If either from_real or
17389 from_imag is NULL, use expansion through memory. If result is not
17390 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17391 the parts directly to inner mode, if even that fails, use expansion
17394 PR middle-end/83623
17395 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17396 check for bswap in mode rather than HImode and use that in expand_unop
17399 Copyright (C) 2018 Free Software Foundation, Inc.
17401 Copying and distribution of this file, with or without modification,
17402 are permitted in any medium without royalty provided the copyright
17403 notice and this notice are preserved.