1 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4 behavior of vec_packsu (vector unsigned long long, vector unsigned
5 long long) to match behavior of vec_packs with same signature.
7 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8 Thomas Schwinge <thomas@codesourcery.com>
9 Cesar Philippidis <cesar@codesourcery.com>
11 * gimplify.c (gimplify_scan_omp_clauses): Add support for
12 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13 (gimplify_adjust_omp_clauses): Likewise.
14 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
16 (gimplify_omp_target_update): Update handling of acc update and
18 * omp-low.c (install_var_field): Remove unused parameter
19 base_pointers_restrict.
20 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
21 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
23 (omp_target_base_pointers_restrict_p): Delete.
24 (scan_omp_target): Update call to scan_sharing_clauses.
25 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
27 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
28 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
29 (convert_local_omp_clauses): Likewise.
30 * tree-pretty-print.c (dump_omp_clause): Likewise.
31 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
33 (omp_clause_code_name): Likewise.
35 2018-06-20 Jakub Jelinek <jakub@redhat.com>
38 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
41 PR tree-optimization/86231
42 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
43 anti-range don't overwrite *vr0min before using it to compute *vr0max.
45 2018-06-20 Tom de Vries <tdevries@suse.de>
47 PR tree-optimization/86097
48 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
49 iv type if signedness of iv type is not the same as that of *nit.
51 2018-06-20 Jakub Jelinek <jakub@redhat.com>
53 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
54 EDGE_EH edges, verify they are all EDGE_EH.
56 2018-06-20 Maya Rashish <coypu@sdf.org>
58 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
60 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
62 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
63 * config/aarch64/aarch64.c (xgene1_tunings): Add
64 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
65 (aarch64_mode_valid_for_sched_fusion_p):
67 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
68 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
70 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
72 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
73 * config/aarch64/iterators.md (VQ2): New mode iterator.
75 2018-06-20 Martin Liska <mliska@suse.cz>
77 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
78 Change default ratio from 10 to 8.
80 2018-06-20 Martin Liska <mliska@suse.cz>
82 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
84 (bit_test_cluster::find_bit_tests): Likewise.
85 (switch_decision_tree::analyze_switch_statement): Find clusters.
86 * tree-switch-conversion.h (struct jump_table_cluster): Document
89 2018-06-20 Martin Liska <mliska@suse.cz>
91 * tree-switch-conversion.c (switch_conversion::collect):
92 Record m_uniq property.
93 (switch_conversion::expand): Bail out for special conditions.
94 (group_cluster::~group_cluster): New.
95 (group_cluster::group_cluster): Likewise.
96 (group_cluster::dump): Likewise.
97 (jump_table_cluster::emit): New.
98 (switch_decision_tree::fix_phi_operands_for_edges): New.
99 (struct case_node): Remove struct.
100 (jump_table_cluster::can_be_handled): New.
101 (case_values_threshold): Moved to header.
102 (reset_out_edges_aux): Likewise.
103 (jump_table_cluster::is_beneficial): New.
104 (bit_test_cluster::can_be_handled): Likewise.
105 (add_case_node): Remove.
106 (bit_test_cluster::is_beneficial): New.
107 (case_bit_test::cmp): New.
108 (bit_test_cluster::emit): New.
109 (expand_switch_as_decision_tree_p): Remove.
110 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
111 (fix_phi_operands_for_edge): Likewise.
112 (switch_decision_tree::analyze_switch_statement): New.
113 (compute_cases_per_edge): Move ...
114 (switch_decision_tree::compute_cases_per_edge): ... here.
115 (try_switch_expansion): Likewise.
116 (switch_decision_tree::try_switch_expansion): Likewise.
117 (record_phi_operand_mapping): Likewise.
118 (switch_decision_tree::record_phi_operand_mapping): Likewise.
119 (emit_case_decision_tree): Likewise.
120 (switch_decision_tree::emit): Likewise.
121 (balance_case_nodes): Likewise.
122 (switch_decision_tree::balance_case_nodes): Likewise.
123 (dump_case_nodes): Likewise.
124 (switch_decision_tree::dump_case_nodes): Likewise.
125 (emit_jump): Likewise.
126 (switch_decision_tree::emit_jump): Likewise.
127 (emit_cmp_and_jump_insns): Likewise.
128 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
129 (emit_case_nodes): Likewise.
130 (switch_decision_tree::emit_case_nodes): Likewise.
131 (conditional_probability): Remove.
132 * tree-switch-conversion.h (enum cluster_type): New.
134 (struct cluster): Likewise.
135 (cluster::cluster): Likewise.
136 (struct simple_cluster): Likewise.
137 (simple_cluster::simple_cluster): Likewise.
138 (struct group_cluster): Likewise.
139 (struct jump_table_cluster): Likewise.
140 (struct bit_test_cluster): Likewise.
141 (struct min_cluster_item): Likewise.
142 (struct case_tree_node): Likewise.
143 (case_tree_node::case_tree_node): Likewise.
144 (jump_table_cluster::case_values_threshold): Likewise.
145 (struct case_bit_test): Likewise.
146 (struct switch_decision_tree): Likewise.
147 (struct switch_conversion): Likewise.
148 (switch_decision_tree::reset_out_edges_aux): Likewise.
150 2018-06-20 Martin Liska <mliska@suse.cz>
152 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
153 (hoist_edge_and_branch_if_true): Likewise.
154 (expand_switch_using_bit_tests_p): Likewise.
155 (struct case_bit_test): Likewise.
156 (case_bit_test_cmp): Likewise.
157 (emit_case_bit_tests): Likewise.
158 (switch_conversion::switch_conversion): New class.
159 (struct switch_conv_info): Remove old struct.
160 (collect_switch_conv_info): More to ...
161 (switch_conversion::collect): ... this.
162 (check_range): Likewise.
163 (switch_conversion::check_range): Likewise.
164 (check_all_empty_except_final): Likewise.
165 (switch_conversion::check_all_empty_except_final): Likewise.
166 (check_final_bb): Likewise.
167 (switch_conversion::check_final_bb): Likewise.
168 (create_temp_arrays): Likewise.
169 (switch_conversion::create_temp_arrays): Likewise.
170 (free_temp_arrays): Likewise.
171 (gather_default_values): Likewise.
172 (switch_conversion::gather_default_values): Likewise.
173 (build_constructors): Likewise.
174 (switch_conversion::build_constructors): Likewise.
175 (constructor_contains_same_values_p): Likewise.
176 (switch_conversion::contains_same_values_p): Likewise.
177 (array_value_type): Likewise.
178 (switch_conversion::array_value_type): Likewise.
179 (build_one_array): Likewise.
180 (switch_conversion::build_one_array): Likewise.
181 (build_arrays): Likewise.
182 (switch_conversion::build_arrays): Likewise.
183 (gen_def_assigns): Likewise.
184 (switch_conversion::gen_def_assigns): Likewise.
185 (prune_bbs): Likewise.
186 (switch_conversion::prune_bbs): Likewise.
187 (fix_phi_nodes): Likewise.
188 (switch_conversion::fix_phi_nodes): Likewise.
189 (gen_inbound_check): Likewise.
190 (switch_conversion::gen_inbound_check): Likewise.
191 (process_switch): Use the newly created class.
192 (switch_conversion::expand): New.
193 (switch_conversion::~switch_conversion): New.
194 * tree-switch-conversion.h: New file.
196 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
198 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
199 tree-vect-patterns.c.
200 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
201 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
202 (vect_recog_sad_pattern): Likewise.
203 (vect_recog_widen_sum_pattern): Likewise.
204 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
205 (vect_recog_widen_shift_pattern): Remove the type_in argument.
206 (vect_recog_rotate_pattern): Likewise.
207 (vect_recog_mult_pattern): Likewise.
208 (vect_recog_vector_vector_shift_pattern): Likewise.
209 (vect_recog_divmod_pattern): Likewise.
210 (vect_recog_mixed_size_cond_pattern): Likewise.
211 (vect_recog_bool_pattern): Likewise.
212 (vect_recog_mask_conversion_pattern): Likewise.
213 (vect_try_gather_scatter_pattern): Likewise.
214 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
215 (vect_recog_over_widening_pattern): Likewise.
216 (vect_recog_gather_scatter_pattern): Likewise.
217 (vect_recog_func_ptr): Move from tree-vectorizer.h
218 (vect_vect_recog_func_ptrs): Move further down the file.
219 (vect_recog_func): Likewise. Remove the third argument.
220 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
221 (vect_pattern_recog_1): Expect the pattern function to do any
222 necessary target tests. Also expect it to provide a vector type.
223 Remove the type_in handling.
225 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
227 * tree-vect-patterns.c (vect_pattern_detected): New function.
228 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
229 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
230 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
231 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
232 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
233 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
234 (vect_recog_mask_conversion_pattern)
235 (vect_try_gather_scatter_pattern): Likewise.
237 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
239 * tree-vect-patterns.c (vect_get_internal_def): New function.
240 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
241 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
242 (search_type_for_mask_1): Use it.
244 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
246 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
247 redundant WIDEN_SUM_EXPR handling.
248 (vect_recog_sad_pattern): Likewise.
250 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
252 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
253 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
254 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
255 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
256 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
258 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
260 * tree-vect-stmts.c (vectorizable_call): Make sure that we
261 use the stmt_vec_info of the original bb statement for the
262 new zero assignment, even if the call is part of a pattern.
264 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
266 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
267 that the sequence is attached to the original statement rather
268 than the pattern statement.
269 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
270 PATTERN_DEF_SEQ from the original statement rather than
271 the main pattern statement.
272 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
273 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
274 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
276 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
278 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
279 definition statements before the early exit for statements that aren't
281 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
283 (vect_transform_loop): ...here. Process pattern definition
284 statements without first checking whether the main pattern
285 statement is live or relevant.
287 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
289 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
290 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
292 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
294 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
295 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
296 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
297 (expand_block_compare): Change select_block_compare_mode call.
298 (expand_strncmp_align_check): Use new functions, fix comment.
299 (emit_final_str_compare_gpr): New function.
300 (expand_strn_compare): Refactor and clean up code.
301 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
303 2018-06-19 Tony Reix <tony.reix@atos.com>
304 Damien Bergamini <damien.bergamini@atos.com>
305 David Edelsohn <dje.gcc@gmail.com>
307 * collect2.c (static_obj): New variable.
308 (static_libs): New variable.
309 (is_in_list): Uncomment declaration.
310 (main): Track AIX libraries linked statically.
311 (is_in_list): Uncomment definition.
312 (scan_prog_file): Don't add AIX shared libraries initializer
313 to constructor list if linking statically.
315 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
317 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
319 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
321 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
323 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
326 2018-06-19 Martin Liska <mliska@suse.cz>
328 * config/i386/i386.c (ix86_can_inline_p): Do not use
329 ipa_fn_summaries::get_create.
330 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
332 (devirtualization_time_bonus): Likewise.
333 (ipcp_propagate_stage): Likewise.
334 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
335 (edge_set_predicate): Likewise.
336 (evaluate_conditions_for_known_args): Likewise.
337 (evaluate_properties_for_edge): Likewise.
338 (ipa_call_summary::reset): Tranform to ...
339 (ipa_call_summary::~ipa_call_summary): ... this.
340 (ipa_fn_summary::reset): Transform to ...
341 (ipa_fn_summary::~ipa_fn_summary): ... this.
342 (ipa_fn_summary_t::remove): Rename to ...
343 (ipa_fn_summary_t::remove_callees): ... this.
344 (ipa_fn_summary_t::duplicate): Use placement new
345 instead of memory copy.
346 (ipa_call_summary_t::duplicate): Likewise.
347 (ipa_call_summary_t::remove): Remove.
348 (dump_ipa_call_summary): Change get_create to get.
349 (ipa_dump_fn_summary): Dump only when summary exists.
350 (analyze_function_body): Use symbol_summary::get instead
352 (compute_fn_summary): Likewise.
353 (estimate_edge_devirt_benefit): Likewise.
354 (estimate_edge_size_and_time): Likewise.
355 (inline_update_callee_summaries): Likewise.
356 (remap_edge_change_prob): Likewise.
357 (remap_edge_summaries): Likewise.
358 (ipa_merge_fn_summary_after_inlining): Likewise.
359 (write_ipa_call_summary): Likewise.
360 (ipa_fn_summary_write): Likewise.
361 (ipa_free_fn_summary): Likewise.
362 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
363 (struct ipa_call_summary): Likewise.
364 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
366 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
367 (estimate_size_after_inlining): Likewise.
368 (estimate_growth): Likewise.
369 (growth_likely_positive): Likewise.
370 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
371 (inline_call): Likewise.
372 * ipa-inline.c (caller_growth_limits): Likewise.
373 (can_inline_edge_p): Likewise.
374 (can_inline_edge_by_limits_p): Likewise.
375 (compute_uninlined_call_time): Likewise.
376 (compute_inlined_call_time): Likewise.
377 (want_inline_small_function_p): Likewise.
378 (edge_badness): Likewise.
379 (update_caller_keys): Likewise.
380 (update_callee_keys): Likewise.
381 (inline_small_functions): Likewise.
382 (inline_to_all_callers_1): Likewise.
383 (dump_overall_stats): Likewise.
384 (early_inline_small_functions): Likewise.
385 (early_inliner): Likewise.
386 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
387 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
388 * ipa-pure-const.c (malloc_candidate_p): Likewise.
389 * ipa-split.c (execute_split_functions): Likewise.
390 * symbol-summary.h: Likewise.
391 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
393 2018-06-19 Richard Biener <rguenther@suse.de>
395 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
396 (vectorize_loops): ... here. Fix dbgcnt handling.
397 (try_vectorize_loop): Wrap try_vectorize_loop_1.
399 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
402 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
403 ieee128 argument takes up only one (vector) register, not two (floating
406 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
408 * gimplify.c (gimplify_init_constructor): Really never clear for an
409 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
411 2018-06-19 Richard Biener <rguenther@suse.de>
413 PR tree-optimization/86179
414 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
415 after failed recognition.
417 2018-06-18 Martin Sebor <msebor@redhat.com>
420 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
421 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
422 Handle integer subtraction.
423 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
424 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
426 2018-06-18 David Malcolm <dmalcolm@redhat.com>
428 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
429 param from rtx to rtx_insn *.
430 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
432 (frv_ifcvt_modify_insn): Likwise.
433 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
434 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
435 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
436 as_a <rtx_insn *> cast to local "unprotected_region" once
437 it's been established that it's not NULL or pc_rtx.
438 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
439 param "sethi" from rtx to rtx_insn *.
440 (nds32_group_float_insns): Likewise for param "insn".
441 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
443 (vax_output_int_subtract): Likewise.
444 * config/vax/vax.c (vax_output_int_add): Likewise for param
446 (vax_output_int_subtract): Likewise.
447 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
448 (emit_pattern_after): Likewise for param "after".
449 (emit_insn_after): Likewise.
450 (emit_jump_insn_after): Likewise.
451 (emit_call_insn_after): Likewise.
452 (emit_debug_insn_after): Likewise.
453 (emit_pattern_before): Likewise for param "before".
454 (emit_insn_before): Likewise.
455 (emit_jump_insn_before): Likewise.
456 * final.c (get_insn_template): Likewise for param "insn", removing
458 * output.h (get_insn_template): Likewise for 2nd param.
459 * rtl.h (emit_insn_before): Likewise.
460 (emit_jump_insn_before): Likewise.
461 (emit_debug_insn_before_noloc): Likewise.
462 (emit_insn_after): Likewise.
463 (emit_jump_insn_after): Likewise.
464 (emit_call_insn_after): Likewise.
465 (emit_debug_insn_after): Likewise.
466 (set_insn_deleted): Likewise for param.
468 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
471 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
472 floating point modes, so that IFmode is numerically greater than
473 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
474 to declare the ordering. This prevents IFmode from being
475 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
476 machine. Include rs6000-modes.h to share the fractional values
477 between genmodes* and the rest of the compiler.
481 * config/rs6000/rs6000-modes.h: New file.
482 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
483 meaning of rs6000_long_double_size so that 126..128 selects an
484 appropriate 128-bit floating point type.
485 (rs6000_option_override_internal): Likewise.
486 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
487 (TARGET_LONG_DOUBLE_128): Change the meaning of
488 rs6000_long_double_size so that 126..128 selects an appropriate
489 128-bit floating point type.
490 (LONG_DOUBLE_TYPE_SIZE): Update comment.
491 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
492 source and destination to match the standard usage.
493 (truncifkf2): Likewise.
494 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
495 ISA 2.07 to use an explicit clobber, instead of passing in a
497 (copysign<mode>3_soft): Likewise.
499 2018-06-18 David Malcolm <dmalcolm@redhat.com>
501 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
502 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
503 (vect_slp_analyze_instance_dependence): Likewise.
504 (vect_enhance_data_refs_alignment): Likewise.
505 (vect_analyze_data_refs_alignment): Likewise.
506 (vect_slp_analyze_and_verify_instance_alignment
507 (vect_analyze_data_ref_accesses): Likewise.
508 (vect_prune_runtime_alias_test_list): Likewise.
509 (vect_analyze_data_refs): Likewise.
510 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
511 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
512 (vect_analyze_scalar_cycles_1): Likewise.
513 (vect_get_loop_niters): Likewise.
514 (vect_analyze_loop_form_1): Likewise.
515 (vect_update_vf_for_slp): Likewise.
516 (vect_analyze_loop_operations): Likewise.
517 (vect_analyze_loop): Likewise.
518 (vectorizable_induction): Likewise.
519 (vect_transform_loop): Likewise.
520 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
521 * tree-vect-slp.c (vect_analyze_slp): Likewise.
522 (vect_make_slp_decision): Likewise.
523 (vect_detect_hybrid_slp): Likewise.
524 (vect_slp_analyze_operations): Likewise.
525 (vect_slp_bb): Likewise.
526 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
527 (vectorizable_bswap): Likewise.
528 (vectorizable_call): Likewise.
529 (vectorizable_simd_clone_call): Likewise.
530 (vectorizable_conversion): Likewise.
531 (vectorizable_assignment): Likewise.
532 (vectorizable_shift): Likewise.
533 (vectorizable_operation): Likewise.
534 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
536 2018-06-18 Martin Sebor <msebor@redhat.com>
538 PR tree-optimization/81384
539 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
540 * builtins.c (expand_builtin_strnlen): New function.
541 (expand_builtin): Call it.
542 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
543 * builtins.def (BUILT_IN_STRNLEN): New.
544 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
545 Warn for bounds in excess of maximum object size.
546 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
547 single-value ranges. Handle strnlen.
548 (handle_builtin_strlen): Handle strnlen.
549 (strlen_check_and_optimize_stmt): Same.
550 * doc/extend.texi (Other Builtins): Document strnlen.
552 2018-06-18 Maya Rashish <coypu@sdf.org>
554 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
555 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
556 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
558 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
560 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
562 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
564 * tree.c (escaped_string::escape): Replace cast to char * by
565 const_cast<char *> (unescaped).
567 2018-06-18 Nick Clifton <nickc@redhat.com>
570 * tree.c (escaped_string): New class. Converts an unescaped
571 string into its escaped equivalent.
572 (warn_deprecated_use): Use the new class to convert the
573 deprecation message, if present.
574 (test_escaped_strings): New self test.
575 (test_c_tests): Add test_escaped_strings.
576 * doc/extend.texi (deprecated): Add a note that the
577 deprecation message is affected by the -fmessage-length
578 option, and that control characters will be escaped.
579 (#pragma GCC error): Document this pragma.
580 (#pragma GCC warning): Likewise.
581 * doc/invoke.texi (-fmessage-length): Document this option's
582 effect on the #warning and #error preprocessor directives and
583 the deprecated attribute.
585 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
587 * tree.c (decl_value_expr_lookup): Revert latest change.
588 (decl_value_expr_insert): Likewise.
590 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
592 * gimplify.c (nonlocal_vlas): Delete.
593 (nonlocal_vla_vars): Likewise.
594 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
596 (gimplify_body): Do not create and destroy nonlocal_vlas.
597 * tree-nested.c: Include diagnostic.h.
598 (use_pointer_in_frame): Tweak.
599 (lookup_field_for_decl): Add assertion and declare the transformation.
600 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
601 internal error when the reference is in a wrong context. Do not
602 create a debug decl by default.
603 (note_nonlocal_block_vlas): Delete.
604 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
605 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
606 create a debug decl by default.
607 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
608 call is in a wrong context.
609 (fixup_vla_decls): New function.
610 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
611 debug variables were created.
612 * tree.c (decl_value_expr_lookup): Add checking assertion.
613 (decl_value_expr_insert): Likewise.
615 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
618 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
619 * tree-scalar-evolution.c (interpret_expr): Likewise.
620 (expression_expensive_p): Likewise.
621 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
622 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
623 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
624 (ssa_defined_by_minus_one_stmt_p): New.
626 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
629 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
630 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
631 * dojump.c (do_jump): Likewise.
632 * expr.c (expand_expr_real_2): Check operand type's sign.
633 * fold-const.c (const_unop): Handle ABSU_EXPR.
634 (fold_abs_const): Likewise.
635 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
636 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
637 (strip_sign_op_1): Likesise.
638 * match.pd: Add new pattern to generate ABSU_EXPR.
639 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
640 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
641 * tree-eh.c (operation_could_trap_helper_p): Likewise.
642 * tree-inline.c (estimate_operator_cost): Likewise.
643 * tree-pretty-print.c (dump_generic_node): Likewise.
644 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
645 * tree.def (ABSU_EXPR): New.
647 2018-06-16 Jakub Jelinek <jakub@redhat.com>
650 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
651 documented as preserved for backward compatibility only.
652 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
654 PR rtl-optimization/86108
655 * bb-reorder.c (create_forwarder_block): Renamed to ...
656 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
657 jump from new landing pad to the second part.
658 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
661 2018-06-15 Jakub Jelinek <jakub@redhat.com>
664 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
665 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
666 Only call store_expr for halves if the mode is the same.
669 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
670 Fix up comment formatting.
672 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
674 * typed-splay-tree.h (typed_splay_tree::remove): New function.
675 (typed_splay_tree::closure,
676 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
677 (typed_splay_tree::typed_splay_tree,
678 typed_splay_tree::operator =): Declared private.
679 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
680 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
681 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
682 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
683 typed_splay_tree::splay_tree_splay,
684 typed_splay_tree::splay_tree_foreach_helper,
685 typed_splay_tree::splay_tree_insert,
686 typed_splay_tree::splay_tree_remove,
687 typed_splay_tree::splay_tree_lookup,
688 typed_splay_tree::splay_tree_predecessor,
689 typed_splay_tree::splay_tree_successor,
690 typed_splay_tree::splay_tree_min,
691 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
692 (typed_splay_tree::root, typed_splay_tree::comp,
693 typed_splay_tree::delete_key,
694 typed_splay_tree::delete_value): New data members.
695 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
696 typed_splay_tree::remove.
698 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
700 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
701 -mginv and -mno-ginv to the assembler.
702 * config/mips/mips.opt (-mcrc): New option.
704 * doc/invoke.text (-mcrc): Document.
707 2018-06-15 Nick Clifton <nickc@redhat.com>
710 * tree.c (escaped_string): New class. Converts an unescaped
711 string into its escaped equivalent.
712 (warn_deprecated_use): Use the new class to convert the
713 deprecation message, if present.
714 (test_escaped_strings): New self test.
715 (test_c_tests): Add test_escaped_strings.
716 * doc/extend.texi (deprecated): Add a note that the
717 deprecation message is affected by the -fmessage-length
718 option, and that control characters will be escaped.
719 (#pragma GCC error): Document this pragma.
720 (#pragma GCC warning): Likewise.
721 * doc/invoke.texi (-fmessage-length): Document this option's
722 effect on the #warning and #error preprocessor directives and
723 the deprecated attribute.
725 2018-06-15 Richard Biener <rguenther@suse.de>
727 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
728 here, also noting vector size used.
729 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
730 size used in MSG_OPTIMIZED_LOCATIONS dump.
731 (pass_slp_vectorize::execute): Adjust.
733 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
736 * config/arc/arc.c (arc_return_address_register): Fix
739 2018-06-15 Richard Biener <rguenther@suse.de>
742 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
743 leave useless conversion stripping to force_gimple_operand_gsi.
744 (gimplify_build2): Likewise.
745 (gimplify_build1): Likewise.
747 2018-06-15 Richard Biener <rguenther@suse.de>
750 * tree-cfg.c (move_stmt_op): unshare invariant addresses
751 before adjusting their block.
753 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
755 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
756 multilibs for *-*-rtems*.
757 * config/riscv/t-rtems: New file.
759 2018-06-14 Jakub Jelinek <jakub@redhat.com>
762 * match.pd ((A +- CST1) +- CST2): Punt if last resort
763 unsigned_type_for returns NULL.
766 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
767 subregs of multi-word pseudos unless the float mode has word size.
769 2018-06-14 Richard Biener <rguenther@suse.de>
772 * tree-vect-generic.c (build_word_mode_vector_type): Remove
773 duplicate and harmful type_hash_canon.
774 * tree.c (type_hash_canon): Assert we didn't find ourselves.
776 2018-06-14 Richard Biener <rguenther@suse.de>
779 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
782 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
784 * config/rtems.h (STDINT_LONG32): Define.
786 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
787 Prachi Godbole <prachi.godbole@imgtec.com>
789 * config/mips/mips-cpus.def: Define P6600.
790 * config/mips/mips-tables.opt: Regenerate.
791 * config/mips/mips.c (mips_ucbranch_type): New enum.
792 (mips_rtx_cost_data): Add support for P6600.
793 (mips_issue_rate): Likewise.
794 (mips_multipass_dfa_lookahead): Likewise.
795 (mips_avoid_hazard): Likewise.
796 (mips_reorg_process_insns): Likewise.
797 (mips_classify_branch_p6600): New function.
798 * config/mips/mips.h (TUNE_P6600): New define.
799 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
800 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
801 * config/mips/mips.md: Include p6600.md.
802 (processor): Add p6600.
803 * config/mips/p6600.md: New file.
804 * doc/invoke.texi: Add p6600 to supported architectures.
806 2018-06-13 Martin Sebor <msebor@redhat.com>
808 PR tree-optimization/86114
809 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
811 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
813 2018-06-13 Richard Biener <rguenther@suse.de>
815 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
816 Properly set vector type of the intermediate stmt.
817 * tree-vect-stmts.c (vectorizable_operation): The destination
818 var always has vectype_out type.
820 2018-06-13 Jeff Law <law@redhat.com>
822 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
823 integer 0 for argument to print_rtl_with_bb.
824 (rl78_reorg): Likewise.
826 2018-06-13 David Malcolm <dmalcolm@redhat.com>
828 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
829 from rtx to rtx_insn *.
830 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
832 (add_sched_insns_for_speculation): Likewise for local "target",
833 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
834 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
835 from rtx_insn ** to rtx_code_label **.
836 (reorg_emit_nops): Likewise.
837 (c6x_reorg): Likewise for local "call_labels".
838 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
840 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
841 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
842 the loops over LABEL_REFS.
843 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
845 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
846 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
847 (split_branches): Strengthen local "olabel" from rtx to
848 rtx_insn *, adding a safe_as_a cast.
849 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
851 (add_insn_after): Likewise for first two params.
852 (add_insn_before): Likewise.
853 (remove_insn): Likewise for param.
854 (emit_pattern_before_noloc): Likewise for second and third params.
855 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
856 (emit_call_insn_before_noloc): Likewise.
857 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
859 (emit_barrier_before): Likewise.
860 (emit_label_before): Strengthen "label" param from "rtx" to
861 "rtx_code_label *". Strengthen "before" param from "rtx" to
863 (emit_insn_after_1): Strengthen "after" param from "rtx" to
865 (emit_pattern_after_noloc): Likewise.
866 (emit_insn_after_noloc): Likewise.
867 (emit_jump_insn_after_noloc): Likewise.
868 (emit_call_insn_after_noloc): Likewise.
869 (emit_debug_insn_after_noloc): Likewise.
870 (emit_barrier_after): Likewise.
871 (emit_label_after): Likewise for both params.
872 (emit_pattern_after_setloc): Likewise for "after" param. Convert
873 "loc" param from "int" to "location_t".
874 (emit_insn_after_setloc): Likewise.
875 (emit_jump_insn_after_setloc): Likewise.
876 (emit_call_insn_after_setloc): Likewise.
877 (emit_debug_insn_after_setloc): Likewise.
878 (emit_pattern_before_setloc): Likewise for "before" param. Convert
879 "loc" param from "int" to "location_t".
880 (emit_pattern_before): Convert NULL_RTX to NULL.
881 (emit_insn_before_setloc): Convert "loc" param from "int" to
883 (emit_jump_insn_before_setloc): Likewise.
884 (emit_call_insn_before_setloc): Likewise.
885 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
886 rtx_insn *. Convert "loc" param from "int" to "location_t".
887 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
888 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
889 Convert 3rd param from "int" to "location_t".
890 (emit_barrier_before, emit_barrier_after, next_real_insn):
891 Strengthen param from rtx to rtx_insn *.
892 (emit_label_before): Strengthen 1st param from "rtx" to
893 "rtx_code_label *". Strengthen 2nd param from "rtx" to
895 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
896 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
897 Strengthen 2nd param from "rtx" to "rtx_insn *".
898 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
899 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
900 Likewise. Convert 3rd param from "int" to "location_t".
901 (emit_label_after): Strengthen 1st param from "rtx" to
903 (next_real_insn, remove_insn): Strengthen param from "rtx" to
905 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
906 from "rtx" to "rtx_insn *".
908 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
910 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
911 bodies streamed in with -Q.
912 * dumpfile.c (dump_files): Add lto-stream-out dump file.
913 * dumpfile.h (tree_dump_index): Add lto_stream_out.
914 * gimple-streamer-out.c: Include gimple-pretty-print.h
915 (output_bb): Dump stmts streamed.
916 * lto-section-out.c: Include print-tree.h
917 (lto_begin_section): Dump sections created.
918 (lto_output_decl_index): Dump decl encoded.
919 * lto-streamer-out.c: Include print-tree.h
920 (create_output_block): Dump output block created.
921 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
922 (output_function): Dump function output.
923 (output_constructor): Dump constructor streamed.
924 (write_global_stream): Output indexes encoded.
925 (produce_asm_for_decls): Dump streams encoded.
926 * lto-streamer.c (streamer_dump_file): New global var.
927 * lto-streamer.h (streamer_dump_file): Declare.
928 * passes.c (ipa_write_summaries): Initialize streamer dump.
929 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
932 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
935 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
936 offsets for register save directives. Emit a second batch of save
937 directives, if need be, when the function accesses prior frames.
939 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
941 * config/arc/fpu.md (fmasf4): Force operand to register.
944 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
946 * config/arc/arc-protos.h (arc_pad_return): Remove.
947 * config/arc/arc.c (machine_function): Remove force_short_suffix
949 (arc_print_operand): Adjust printing of '&'.
950 (arc_verify_short): Remove conditional printing of short suffix.
951 (arc_final_prescan_insn): Remove reference to size_reason.
952 (pad_return): New function.
953 (arc_reorg): Call pad_return.
954 (arc_pad_return): Remove.
955 (arc_init_machine_status): Remove reference to force_short_suffix.
956 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
957 (attr length): When attribute iscompact is true force to 2
958 regardless; in the case of maybe check if we want to force the
959 instruction to have 4 bytes length.
960 (nopv): Change it to generate 4 byte long nop as well.
961 (blockage): New pattern.
962 (simple_return): Remove call to arc_pad_return.
963 (p_return_i): Likewise.
965 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
967 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
969 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
971 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
974 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
976 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
977 for ARC700 and ARCv2.
979 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
982 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
983 operands[2] instead of operands[1].
986 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
988 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
989 case, check whether the outer register overlaps an unallocatable
990 register, not just whether it fits the required class.
992 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
994 * poly-int.h (can_div_trunc_p): Add new overload in which all values
996 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
997 (memrefs_conflict_p): Likewise.
998 (init_alias_analysis): Likewise.
999 * cfgexpand.c (expand_debug_expr): Likewise.
1000 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1001 * cse.c (fold_rtx): Likewise.
1002 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1003 * expr.c (emit_block_move_hints): Likewise.
1004 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1005 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1006 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1007 (emit_group_store): Likewise.
1008 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1009 to read the PRE/POST_MODIFY increment.
1010 * calls.c (store_one_arg): Use strip_offset.
1011 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1013 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1015 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1016 (simplify_binary_operation_1): Extend CONST_INT handling to
1018 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1019 than a HOST_WIDE_INT.
1020 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1022 (adjust_mems, add_stores): Update accodingly.
1023 (vt_canonicalize_addr): Track polynomial offsets.
1024 (emit_note_insn_var_location): Likewise.
1025 (vt_add_function_parameter): Likewise.
1026 (vt_initialize): Likewise.
1028 2018-06-12 Jeff Law <law@redhat.com>
1030 * config.gcc (alpha*-*-freebsd*): Remove.
1031 * config/alpha/freebsd.h: Remove.
1033 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1036 * spellcheck-tree.c (levenshtein_distance): Rename to...
1037 (get_edit_distance): ...this, and update for underlying renaming.
1038 * spellcheck-tree.h (levenshtein_distance): Rename to...
1039 (get_edit_distance): ...this.
1040 * spellcheck.c (levenshtein_distance): Rename to...
1041 (get_edit_distance): ...this. Convert from Levenshtein distance
1042 to Damerau-Levenshtein distance by supporting transpositions of
1043 adjacent characters. Rename "v1" to "v_next" and "v0" to
1045 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1046 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1047 update for underlying renaming.
1048 (selftest::levenshtein_distance_unit_test): Rename to...
1049 (selftest::test_get_edit_distance_unit): ...this, and update for
1050 underlying renaming.
1051 (selftest::test_find_closest_string): Add example from PR 69968
1052 where transposition helps
1053 (selftest::test_metric_conditions): Update for renaming.
1054 (selftest::test_metric_conditions): Likewise.
1055 (selftest::spellcheck_c_tests): Likewise.
1056 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1057 (get_edit_distance): ...this.
1058 (best_match::consider): Update for renaming.
1060 2018-06-12 Martin Sebor <msebor@redhat.com>
1062 PR tree-optimization/85259
1063 * builtins.c (compute_objsize): Handle constant offsets.
1064 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1065 true iff a warning has been issued.
1066 * gimple.h (gimple_nonartificial_location): New function.
1067 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1068 gimple_nonartificial_location and handle -Wno-system-headers.
1069 (handle_builtin_stxncpy): Same.
1071 2018-06-12 Martin Sebor <msebor@redhat.com>
1074 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1076 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1078 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1079 BUILTIN_VEC_XST entries for pointer to double and long long.
1081 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1084 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1086 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1089 2018-06-12 Martin Liska <mliska@suse.cz>
1091 * doc/options.texi: Document IntegerRange.
1093 2018-06-12 Martin Liska <mliska@suse.cz>
1095 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1096 * opt-functions.awk: Handle Deprecated flag.
1097 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1099 (read_cmdline_option): Report warning for a deprecated option.
1100 * opts.h (struct cl_option): Add new field cl_deprecated.
1101 (CL_ERR_DEPRECATED): New.
1103 2018-06-12 Martin Liska <mliska@suse.cz>
1105 * doc/options.texi: Document Deprecated option flag.
1107 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1109 * config/arc/arc-arch.h (arc_extras): New enum.
1110 (arc_cpu_t):Add field extra.
1111 (arc_cpu_types): Consider the extras.
1112 * config/arc/arc-cpus.def: Add extras info.
1113 * config/arc/arc-opts.h (processor_type): Consider extra field.
1114 * config/arc/arc.c (arc_override_options): Handle extra field.
1116 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1118 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1119 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1121 * config/arc/arc.c (arc_init): Update pic warning.
1122 (irq_range): Update irq range parsing warnings.
1123 (arc_override_options): Update various warning messages.
1124 (arc_handle_aux_attribute): Likewise.
1126 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1128 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1130 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1132 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1133 compiler messages directives.
1135 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1137 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1138 * config/mips/mips-tables.opt: Regenerate.
1139 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1141 * doc/invoke.texi: Document -march=i6500.
1143 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1145 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1146 (i6400_gpmul): Add cpu_unit.
1147 (i6400_gpdiv): Likewise.
1148 (i6400_msa_add_d): Update reservations.
1149 (i6400_msa_int_add) Likewise.
1150 (i6400_msa_short_logic3) Likewise.
1151 (i6400_msa_short_logic2) Likewise.
1152 (i6400_msa_short_logic) Likewise.
1153 (i6400_msa_move) Likewise.
1154 (i6400_msa_cmp) Likewise.
1155 (i6400_msa_short_float2) Likewise.
1156 (i6400_msa_div_d) Likewise.
1157 (i6400_msa_long_logic1) Likewise.
1158 (i6400_msa_long_logic2) Likewise.
1159 (i6400_msa_mult) Likewise.
1160 (i6400_msa_long_float2) Likewise.
1161 (i6400_msa_long_float4) Likewise.
1162 (i6400_msa_long_float5) Likewise.
1163 (i6400_msa_long_float8) Likewise.
1164 (i6400_fpu_fadd): Include frint type.
1165 (i6400_fpu_store): New define_insn_reservation.
1166 (i6400_fpu_load): Likewise.
1167 (i6400_fpu_move): Likewise.
1168 (i6400_fpu_fcmp): Likewise.
1169 (i6400_fpu_fmadd): Likewise.
1170 (i6400_int_mult): Include imul3nc type and update reservation.
1171 (i6400_int_div): Include idiv3 type and update reservation.
1172 (i6400_int_load): Update to check type not move_type.
1173 (i6400_int_store): Likewise.
1174 (i6400_int_prefetch): Set zero latency.
1176 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1178 * gcc.c: Document new %@{...} sequence.
1179 (LINK_COMMAND_SPEC): Use it for the -L switches.
1180 (cpp_unique_options): Use it for the -I switches.
1181 (at_file_argbuf): New global variable.
1182 (in_at_file): Likewise.
1183 (alloc_args): Create at_file_argbuf.
1184 (clear_args): Truncate at_file_argbuf.
1185 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1186 (open_at_file): New function.
1187 (close_at_file): Likewise.
1188 (create_at_file): Delete.
1189 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1192 (validate_switches_from_spec): Deal with %@{...} sequence.
1193 (validate_switches): Likewise.
1194 (driver::finalize): Call clear_args.
1196 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1198 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1200 2018-06-11 Martin Sebor <msebor@redhat.com>
1202 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1203 (Wno-class-memaccess): Add @opindex.
1204 (Wno-templates, Wno-multiple-inheritance): Same.
1205 (Wno-virtual-inheritance, Wno-namespaces): Same.
1206 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1207 (Wno-format-overflow, Wno-format-truncation): Same.
1208 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1209 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1210 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1211 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1212 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1213 (Wno-misspelled-isr): Same.
1215 2018-06-11 Martin Sebor <msebor@redhat.com>
1217 * PR tree-optimization/86083
1218 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1220 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1222 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1224 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1227 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1228 on the correct operand.
1229 (*movdi_internal64): Ditto.
1231 2018-06-11 Martin Liska <mliska@suse.cz>
1233 PR tree-optimization/86089
1234 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1236 2018-06-11 Julia Koval <julia.koval@intel.com>
1238 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1239 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1240 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1242 2018-06-11 Olivier Hainque <hainque@adacore.com>
1244 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1245 for Ada with strict dwarf2.
1247 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1250 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1253 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1255 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1257 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1259 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1262 2018-06-08 Martin Liska <mliska@suse.cz>
1264 * tree-cfg.h (debug_function): Fix argument type to match
1267 2018-06-08 Martin Liska <mliska@suse.cz>
1269 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1270 Remove usage of MPX-related (and removed) fields.
1271 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1273 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1275 * cfg.c (debug): Use TDF_NONE rather than 0.
1276 * cfghooks.c (debug): Likewise.
1277 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1278 (struct dump_option_value_info): Convert to...
1279 (struct kv_pair): ...this template type.
1280 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1282 (optinfo_verbosity_options): Likewise.
1283 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1285 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1286 than int for "optgroup_flags" param.
1287 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1289 (dump_dec): Likewise.
1290 (dump_finish): Use TDF_NONE rather than 0.
1291 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1292 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1293 than 0. Update for change to option_ptr.
1294 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1295 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1296 0. Update for changes to optinfo_verbosity_options and
1298 (opt_info_switch_p): Convert optgroup_flags from int to
1300 (dump_basic_block): Use dump_flags_t rather than int
1301 for "dump_kind" param.
1302 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1303 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1304 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1305 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1306 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1307 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1308 TDF_NONE): Convert from macros to...
1309 (enum dump_flag): ...this new enum.
1310 (dump_flags_t): Update to use enum.
1311 (operator|, operator&, operator~, operator|=, operator&=):
1312 Implement for dump_flags_t.
1313 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1314 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1315 Convert from macros to...
1316 (enum optgroup_flag): ...this new enum.
1317 (optgroup_flags_t): New typedef.
1318 (operator|, operator|=): Implement for optgroup_flags_t.
1319 (struct dump_file_info): Convert field "alt_flags" to
1320 dump_flags_t. Convert field "optgroup_flags" to
1322 (dump_basic_block): Use dump_flags_t rather than int for param.
1323 (dump_generic_expr_loc): Likewise.
1324 (dump_dec): Likewise.
1325 (dump_register): Convert param "optgroup_flags" to
1327 (opt_info_enable_passes): Likewise.
1328 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1330 * gimple-pretty-print.c (debug): Likewise.
1331 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1332 (merged_store_group::apply_stores): Likewise.
1333 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1334 * gimple.c (verify_gimple_pp): Likewise.
1335 * graphite-poly.c (print_pbb_body): Likewise.
1336 * passes.c (pass_manager::register_one_dump_file): Convert
1337 local "optgroup_flags" to optgroup_flags_t.
1338 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1340 (debug_body): Likewise.
1341 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1342 to optgroup_flags_t.
1343 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1345 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1346 (convert_mult_to_fma): Likewise.
1347 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1348 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1349 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1350 "dump_kind" to dump_flags_t.
1352 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1354 * config/rs6000/rs6000.c (min, max): Delete.
1356 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1358 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1359 -mabi=spe and -mabi=no-spe.
1361 2018-06-08 Martin Liska <mliska@suse.cz>
1363 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1364 where we expect an existing summary.
1366 2018-06-08 Martin Liska <mliska@suse.cz>
1368 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1369 * ipa-inline.h (estimate_edge_growth): Likewise.
1371 2018-06-08 Martin Liska <mliska@suse.cz>
1373 * cgraph.c (function_version_hasher::hash): Use
1374 cgraph_node::get_uid ().
1375 (function_version_hasher::equal):
1376 * cgraph.h (cgraph_node::get_uid): New method.
1377 * ipa-inline.c (update_caller_keys): Use
1378 cgraph_node::get_uid ().
1379 (update_callee_keys): Likewise.
1380 * ipa-utils.c (searchc): Likewise.
1381 (ipa_reduced_postorder): Likewise.
1382 * lto-cgraph.c (input_node): Likewise.
1383 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1384 * symbol-summary.h (symtab_insertion): Likewise.
1385 (symtab_removal): Likewise.
1386 (symtab_duplication): Likewise.
1387 * tree-pretty-print.c (dump_function_header): Likewise.
1388 * tree-sra.c (convert_callers_for_node): Likewise.
1390 2018-06-08 Martin Liska <mliska@suse.cz>
1392 * cgraph.c (symbol_table::create_edge): Always assign a new
1394 (symbol_table::free_edge): Do not recycle numbers.
1395 * cgraph.h (cgraph_edge::get): New method.
1396 * symbol-summary.h (symtab_removal): Use it.
1397 (symtab_duplication): Likewise.
1398 (call_summary::hashable_uid): Remove.
1400 2018-06-08 Martin Liska <mliska@suse.cz>
1402 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1403 (initialize_growth_caches): Remove.
1404 (free_growth_caches): Likewise.
1405 (do_estimate_edge_time): Use edge_growth_cache.
1406 (do_estimate_edge_size): Likewise.
1407 (do_estimate_edge_hints): Likewise.
1408 * ipa-inline.c (reset_edge_caches): Likewise.
1409 (recursive_inlining): Likewise.
1410 (inline_small_functions): Likewise.
1411 * ipa-inline.h (initialize_growth_caches): Remove.
1412 (estimate_edge_size): Likewise.
1413 (estimate_edge_time): Likewise.
1414 (estimate_edge_hints): Likewise.
1415 (reset_edge_growth_cache): Likewise.
1416 * symbol-summary.h (call_summary::remove): New method.
1418 2018-06-08 Martin Liska <mliska@suse.cz>
1420 * ipa-cp.c (class edge_clone_summary): New summary.
1421 (grow_edge_clone_vectors): Remove.
1422 (ipcp_edge_duplication_hook): Remove.
1423 (class edge_clone_summary_t): New call_summary class.
1424 (ipcp_edge_removal_hook): Remove.
1425 (edge_clone_summary_t::duplicate): New function.
1426 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1427 (create_specialized_node): Likewise.
1428 (ipcp_driver): Initialize edge_clone_summaries and do not
1431 2018-06-08 Martin Liska <mliska@suse.cz>
1433 * symbol-summary.h (get): New function.
1434 (call_summary::m_initialize_when_cloning): New class member.
1436 2018-06-08 Martin Liska <mliska@suse.cz>
1438 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1439 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1440 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1441 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1442 (remove_cgraph_node_from_order): Use the removed_nodes set.
1443 (do_per_function_toporder): Likwise.
1444 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1445 instead of summary_uid.
1446 (symtab_removal): Likewise.
1447 (symtab_duplication): Likewise.
1449 2018-06-08 Martin Liska <mliska@suse.cz>
1451 * ipa-cp.c (ipcp_store_bits_results): Use
1452 ipcp_transformation_sum.
1453 (ipcp_store_vr_results): Likewise.
1454 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1456 (ipcp_transformation_initialize): ... this.
1457 (ipa_set_node_agg_value_chain):
1458 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1459 (write_ipcp_transformation_info): Likewise.
1460 (read_ipcp_transformation_info): Likewise.
1461 (ipcp_update_bits): Likewise.
1462 (ipcp_update_vr): Likewise.
1463 (ipcp_transform_function): Likewise.
1464 * ipa-prop.h: Rename ipcp_transformation_summary to
1465 ipcp_transformation.
1466 (class ipcp_transformation_t): New function summary.
1467 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1468 (ipa_get_agg_replacements_for_node): Likewise.
1470 2018-06-08 Martin Liska <mliska@suse.cz>
1472 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1474 (class funct_state_summary_t): New function_summary class.
1475 (has_function_state): Remove.
1476 (get_function_state): Likewise.
1477 (set_function_state): Likewise.
1478 (add_new_function): Likewise.
1479 (funct_state_summary_t::insert): New function.
1480 (duplicate_node_data): Remove.
1481 (remove_node_data): Remove.
1482 (funct_state_summary_t::duplicate): New function.
1483 (register_hooks): Create new funct_state_summaries.
1484 (pure_const_generate_summary): Use it.
1485 (pure_const_write_summary): Likewise.
1486 (pure_const_read_summary): Likewise.
1487 (propagate_pure_const): Likewise.
1488 (propagate_nothrow): Likewise.
1489 (dump_malloc_lattice): Likewise.
1490 (propagate_malloc): Likewise.
1491 (execute): Do not register hooks, just remove summary
1493 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1496 2018-06-08 Martin Liska <mliska@suse.cz>
1498 * ipa-reference.c (remove_node_data): Remove.
1499 (duplicate_node_data): Likewise.
1500 (class ipa_ref_var_info_summary_t): New class.
1501 (class ipa_ref_opt_summary_t): Likewise.
1502 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1503 (get_reference_optimization_summary): Use
1504 ipa_ref_opt_sum_summaries.
1505 (set_reference_vars_info): Remove.
1506 (set_reference_optimization_summary): Likewise.
1507 (ipa_init): Create summaries.
1508 (init_function_info): Use function summary.
1509 (ipa_ref_opt_summary_t::duplicate): New function.
1510 (ipa_ref_opt_summary_t::remove): New function.
1511 (get_read_write_all_from_node): Fix GNU coding style.
1512 (propagate): Use function summary.
1513 (write_node_summary_p): Fix GNU coding style.
1514 (stream_out_bitmap): Likewise.
1515 (ipa_reference_read_optimization_summary): Use function summary.
1516 (ipa_reference_c_finalize): Do not release hooks.
1518 2018-06-08 Martin Liska <mliska@suse.cz>
1520 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1521 (analyze_function_body): Extract multiple calls of get_create.
1522 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1523 * ipa-inline.c (recursive_inlining): Use ::get method.
1524 * ipa-inline.h (estimate_edge_growth): Likewise.
1526 2018-06-08 Martin Liska <mliska@suse.cz>
1528 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1530 (hsa_function_summary::hsa_function_summary): Use the new enum
1532 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1533 * hsa-gen.c (hsa_get_host_function): Likewise.
1534 (get_brig_function_name): Likewise.
1535 * ipa-hsa.c (process_hsa_functions): Likewise.
1536 (ipa_hsa_write_summary): Likewise.
1537 * symbol-summary.h (symtab_duplication): Use ::get function/
1538 (get): New function.
1540 2018-06-08 Martin Liska <mliska@suse.cz>
1542 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1544 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1545 (hsa_register_kernel): Likewise.
1546 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1547 * hsa-gen.c (hsa_get_host_function): Likewise.
1548 (get_brig_function_name): Likewise.
1549 (generate_hsa): Likewise.
1550 (pass_gen_hsail::execute): Likewise.
1551 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1552 (devirtualization_time_bonus): Likewise.
1553 (ipcp_propagate_stage): Likewise.
1554 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1555 (edge_set_predicate): Likewise.
1556 (evaluate_conditions_for_known_args): Likewise.
1557 (evaluate_properties_for_edge): Likewise.
1558 (ipa_fn_summary::reset): Likewise.
1559 (ipa_fn_summary_t::duplicate): Likewise.
1560 (dump_ipa_call_summary): Likewise.
1561 (ipa_dump_fn_summary): Likewise.
1562 (analyze_function_body): Likewise.
1563 (compute_fn_summary): Likewise.
1564 (estimate_edge_devirt_benefit): Likewise.
1565 (estimate_edge_size_and_time): Likewise.
1566 (estimate_calls_size_and_time): Likewise.
1567 (estimate_node_size_and_time): Likewise.
1568 (inline_update_callee_summaries): Likewise.
1569 (remap_edge_change_prob): Likewise.
1570 (remap_edge_summaries): Likewise.
1571 (ipa_merge_fn_summary_after_inlining): Likewise.
1572 (ipa_update_overall_fn_summary): Likewise.
1573 (read_ipa_call_summary): Likewise.
1574 (inline_read_section): Likewise.
1575 (write_ipa_call_summary): Likewise.
1576 (ipa_fn_summary_write): Likewise.
1577 (ipa_free_fn_summary): Likewise.
1578 * ipa-hsa.c (process_hsa_functions): Likewise.
1579 (ipa_hsa_write_summary): Likewise.
1580 (ipa_hsa_read_section): Likewise.
1581 * ipa-icf.c (sem_function::merge): Likewise.
1582 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1583 (do_estimate_edge_time): Likewise.
1584 (estimate_size_after_inlining): Likewise.
1585 (estimate_growth): Likewise.
1586 (growth_likely_positive): Likewise.
1587 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1588 (inline_call): Likewise.
1589 * ipa-inline.c (caller_growth_limits): Likewise.
1590 (can_inline_edge_p): Likewise.
1591 (can_inline_edge_by_limits_p): Likewise.
1592 (compute_uninlined_call_time): Likewise.
1593 (compute_inlined_call_time): Likewise.
1594 (want_inline_small_function_p): Likewise.
1595 (edge_badness): Likewise.
1596 (update_caller_keys): Likewise.
1597 (update_callee_keys): Likewise.
1598 (recursive_inlining): Likewise.
1599 (inline_small_functions): Likewise.
1600 (inline_to_all_callers_1): Likewise.
1601 (dump_overall_stats): Likewise.
1602 (early_inline_small_functions): Likewise.
1603 (early_inliner): Likewise.
1604 * ipa-inline.h (estimate_edge_growth): Likewise.
1605 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1606 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1607 * ipa-prop.h (IPA_NODE_REF): Likewise.
1608 (IPA_EDGE_REF): Likewise.
1609 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1610 (propagate_malloc): Likewise.
1611 * ipa-split.c (execute_split_functions): Likewise.
1612 * symbol-summary.h: Rename get to get_create.
1614 (get_create): Likewise.
1615 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1617 2018-06-08 Martin Liska <mliska@suse.cz>
1619 * symbol-summary.h (release): Move definition out of class
1621 (symtab_removal): Likewise.
1622 (symtab_duplication): Likewise.
1624 2018-06-08 Martin Liska <mliska@suse.cz>
1626 * symbol-summary.h (function_summary): Move constructor
1627 implementation out of class declaration.
1628 (release): Likewise.
1629 (symtab_insertion): Likewise.
1630 (symtab_removal): Likewise.
1631 (symtab_duplication): Likewise.
1634 2018-06-08 Martin Liska <mliska@suse.cz>
1636 * Makefile.in: Remove support for MPX (macros, related functions,
1637 fields in cgraph_node, ...).
1638 * builtin-types.def (BT_BND): Likewise.
1639 (BT_FN_BND_CONST_PTR): Likewise.
1640 (BT_FN_CONST_PTR_BND): Likewise.
1641 (BT_FN_VOID_PTR_BND): Likewise.
1642 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1643 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1644 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1645 (expand_builtin_mempcpy_with_bounds): Likewise.
1646 (expand_builtin_memset_with_bounds): Likewise.
1647 (expand_builtin_memset_args): Likewise.
1648 (std_expand_builtin_va_start): Likewise.
1649 (expand_builtin): Likewise.
1650 (expand_builtin_with_bounds): Likewise.
1651 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1652 (DEF_LIB_BUILTIN_CHKP): Likewise.
1653 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1654 (DEF_CHKP_BUILTIN): Likewise.
1655 (BUILT_IN_MEMCPY): Likewise.
1656 (BUILT_IN_MEMMOVE): Likewise.
1657 (BUILT_IN_MEMPCPY): Likewise.
1658 (BUILT_IN_MEMSET): Likewise.
1659 (BUILT_IN_STPCPY): Likewise.
1660 (BUILT_IN_STRCAT): Likewise.
1661 (BUILT_IN_STRCHR): Likewise.
1662 (BUILT_IN_STRCPY): Likewise.
1663 (BUILT_IN_STRLEN): Likewise.
1664 (BUILT_IN_MEMCPY_CHK): Likewise.
1665 (BUILT_IN_MEMMOVE_CHK): Likewise.
1666 (BUILT_IN_MEMPCPY_CHK): Likewise.
1667 (BUILT_IN_MEMSET_CHK): Likewise.
1668 (BUILT_IN_STPCPY_CHK): Likewise.
1669 (BUILT_IN_STRCAT_CHK): Likewise.
1670 (BUILT_IN_STRCPY_CHK): Likewise.
1671 * calls.c (store_bounds): Likewise.
1672 (emit_call_1): Likewise.
1673 (special_function_p): Likewise.
1674 (maybe_warn_nonstring_arg): Likewise.
1675 (initialize_argument_information): Likewise.
1676 (finalize_must_preallocate): Likewise.
1677 (compute_argument_addresses): Likewise.
1678 (expand_call): Likewise.
1679 * cfgexpand.c (expand_call_stmt): Likewise.
1680 (expand_return): Likewise.
1681 (expand_gimple_stmt_1): Likewise.
1682 (pass_expand::execute): Likewise.
1683 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1684 (cgraph_node::remove): Likewise.
1685 (cgraph_node::dump): Likewise.
1686 (cgraph_node::verify_node): Likewise.
1687 * cgraph.h (chkp_function_instrumented_p): Likewise.
1688 (symtab_node::get_alias_target): Likewise.
1689 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1690 (cgraph_local_p): Likewise.
1691 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1692 (cgraph_edge::rebuild_references): Likewise.
1693 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1694 (walk_polymorphic_call_targets): Likewise.
1695 (cgraph_node::expand_thunk): Likewise.
1696 (symbol_table::output_weakrefs): Likewise.
1697 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1698 (ix86_handle_option): Likewise.
1699 * config/i386/constraints.md: Likewise.
1700 * config/i386/i386-builtin-types.def (BND): Likewise.
1704 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1705 (BDESC_FIRST): Likewise.
1707 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1708 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1709 * config/i386/i386.c (enum reg_class): Likewise.
1710 (ix86_target_string): Likewise.
1711 (ix86_option_override_internal): Likewise.
1712 (ix86_conditional_register_usage): Likewise.
1713 (ix86_valid_target_attribute_inner_p): Likewise.
1714 (ix86_set_indirect_branch_type): Likewise.
1715 (ix86_set_current_function): Likewise.
1716 (ix86_function_arg_regno_p): Likewise.
1717 (init_cumulative_args): Likewise.
1718 (ix86_function_arg_advance): Likewise.
1719 (ix86_function_arg): Likewise.
1720 (ix86_pass_by_reference): Likewise.
1721 (ix86_function_value_regno_p): Likewise.
1722 (ix86_function_value_1): Likewise.
1723 (ix86_function_value_bounds): Likewise.
1724 (ix86_return_in_memory): Likewise.
1725 (ix86_setup_incoming_vararg_bounds): Likewise.
1726 (ix86_va_start): Likewise.
1727 (indirect_thunk_need_prefix): Likewise.
1728 (print_reg): Likewise.
1729 (ix86_print_operand): Likewise.
1730 (ix86_expand_call): Likewise.
1731 (ix86_output_function_return): Likewise.
1732 (reg_encoded_number): Likewise.
1733 (BDESC_VERIFYS): Likewise.
1734 (ix86_init_mpx_builtins): Likewise.
1735 (ix86_init_builtins): Likewise.
1736 (ix86_emit_cmove): Likewise.
1737 (ix86_emit_move_max): Likewise.
1738 (ix86_expand_builtin): Likewise.
1739 (ix86_builtin_mpx_function): Likewise.
1740 (ix86_get_arg_address_for_bt): Likewise.
1741 (ix86_load_bounds): Likewise.
1742 (ix86_store_bounds): Likewise.
1743 (ix86_load_returned_bounds): Likewise.
1744 (ix86_store_returned_bounds): Likewise.
1745 (ix86_class_likely_spilled_p): Likewise.
1746 (ix86_hard_regno_mode_ok): Likewise.
1747 (x86_order_regs_for_local_alloc): Likewise.
1748 (ix86_mitigate_rop): Likewise.
1749 (ix86_bnd_prefixed_insn_p): Likewise.
1750 (ix86_mpx_bound_mode): Likewise.
1751 (ix86_make_bounds_constant): Likewise.
1752 (ix86_initialize_bounds): Likewise.
1753 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1754 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1755 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1756 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1757 (TARGET_CHKP_BOUND_MODE): Likewise.
1758 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1759 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1760 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1761 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1762 * config/i386/i386.h (TARGET_MPX): Likewise.
1763 (TARGET_MPX_P): Likewise.
1764 (VALID_BND_REG_MODE): Likewise.
1765 (FIRST_BND_REG): Likewise.
1766 (LAST_BND_REG): Likewise.
1767 (enum reg_class): Likewise.
1768 (BND_REG_P): Likewise.
1769 (BND_REGNO_P): Likewise.
1770 (BNDmode): Likewise.
1771 (ADJUST_INSN_LENGTH): Likewise.
1772 * config/i386/i386.md: Likewise.
1773 * config/i386/i386.opt: Likewise.
1774 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1775 (defined): Likewise.
1776 (LINK_MPX): Likewise.
1777 (MPX_SPEC): Likewise.
1778 (LIBMPX_SPEC): Likewise.
1779 (LIBMPXWRAPPERS_SPEC): Likewise.
1780 (CHKP_SPEC): Likewise.
1781 * config/i386/predicates.md: Likewise.
1782 * dbxout.c (dbxout_type): Likewise.
1783 * doc/extend.texi: Likewise.
1784 * doc/invoke.texi: Likewise.
1785 * doc/md.texi: Likewise.
1786 * doc/tm.texi: Likewise.
1787 * doc/tm.texi.in: Likewise.
1788 * dwarf2out.c (is_base_type): Likewise.
1789 (gen_formal_types_die): Likewise.
1790 (gen_subprogram_die): Likewise.
1791 (gen_type_die_with_usage): Likewise.
1792 (gen_decl_die): Likewise.
1793 (dwarf2out_late_global_decl): Likewise.
1794 * expr.c (expand_assignment): Likewise.
1795 (emit_storent_insn): Likewise.
1796 (store_expr_with_bounds): Likewise.
1797 (store_expr): Likewise.
1798 (expand_expr_real_1): Likewise.
1799 * expr.h (store_expr_with_bounds): Likewise.
1800 * function.c (use_register_for_decl): Likewise.
1801 (struct bounds_parm_data): Likewise.
1802 (assign_parms_augmented_arg_list): Likewise.
1803 (assign_parm_find_entry_rtl): Likewise.
1804 (assign_parm_is_stack_parm): Likewise.
1805 (assign_parm_load_bounds): Likewise.
1806 (assign_bounds): Likewise.
1807 (assign_parms): Likewise.
1808 (expand_function_start): Likewise.
1809 * gcc.c (CHKP_SPEC): Likewise.
1810 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1811 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1812 (wrestrict_dom_walker::check_call): Likewise.
1813 * gimple.c (gimple_build_call_from_tree): Likewise.
1814 * gimple.h (enum gf_mask): Likewise.
1815 (gimple_call_with_bounds_p): Likewise.
1816 (gimple_call_set_with_bounds): Likewise.
1817 * gimplify.c (gimplify_init_constructor): Likewise.
1818 * ipa-cp.c (initialize_node_lattices): Likewise.
1819 (propagate_constants_across_call): Likewise.
1820 (find_more_scalar_values_for_callers_subset): Likewise.
1821 * ipa-hsa.c (process_hsa_functions): Likewise.
1822 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1823 * ipa-icf.c (sem_function::merge): Likewise.
1824 * ipa-inline.c (early_inliner): Likewise.
1825 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1826 (warn_function_cold): Likewise.
1827 (propagate_pure_const): Likewise.
1828 * ipa-ref.h (enum GTY): Likewise.
1829 * ipa-split.c (find_retbnd): Likewise.
1830 (consider_split): Likewise.
1831 (split_function): Likewise.
1832 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1833 * ipa.c (walk_polymorphic_call_targets): Likewise.
1834 (symbol_table::remove_unreachable_nodes): Likewise.
1835 (process_references): Likewise.
1836 (cgraph_build_static_cdtor_1): Likewise.
1837 * lto-cgraph.c (lto_output_node): Likewise.
1838 (output_refs): Likewise.
1839 (compute_ltrans_boundary): Likewise.
1840 (input_overwrite_node): Likewise.
1841 (input_node): Likewise.
1842 (input_cgraph_1): Likewise.
1843 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1844 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1845 (class pass_chkp_instrumentation_passes): Likewise.
1846 (make_pass_chkp_instrumentation_passes): Likewise.
1847 * passes.def: Likewise.
1848 * rtl.h (struct GTY): Likewise.
1849 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1850 * stor-layout.c (layout_type): Likewise.
1851 * symtab.c: Likewise.
1852 * target.def: Likewise.
1853 * targhooks.c (default_chkp_bound_type): Likewise.
1854 (default_chkp_bound_mode): Likewise.
1855 (default_builtin_chkp_function): Likewise.
1856 (default_chkp_function_value_bounds): Likewise.
1857 (default_chkp_make_bounds_constant): Likewise.
1858 (default_chkp_initialize_bounds): Likewise.
1859 * targhooks.h (default_chkp_bound_type): Likewise.
1860 (default_chkp_bound_mode): Likewise.
1861 (default_builtin_chkp_function): Likewise.
1862 (default_chkp_function_value_bounds): Likewise.
1863 (default_chkp_make_bounds_constant): Likewise.
1864 (default_chkp_initialize_bounds): Likewise.
1865 * toplev.c (compile_file): Likewise.
1866 (process_options): Likewise.
1867 * tree-core.h (DEF_BUILTIN): Likewise.
1868 (DEF_BUILTIN_CHKP): Likewise.
1869 * tree-inline.c (declare_return_variable): Likewise.
1870 (remap_gimple_stmt): Likewise.
1871 (copy_bb): Likewise.
1872 (initialize_inlined_parameters): Likewise.
1873 (expand_call_inline): Likewise.
1874 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1875 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1876 (make_pass_ipa_chkp_produce_thunks): Likewise.
1877 (make_pass_chkp): Likewise.
1878 (make_pass_chkp_opt): Likewise.
1879 (make_pass_chkp_instrumentation_passes): Likewise.
1880 * tree-pretty-print.c (dump_generic_node): Likewise.
1881 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1882 * tree-ssa-dce.c (propagate_necessity): Likewise.
1883 (eliminate_unnecessary_stmts): Likewise.
1884 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1885 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1886 * tree-ssa-sccvn.h: Likewise.
1887 * tree-ssa-strlen.c (get_string_length): Likewise.
1888 (valid_builtin_call): Likewise.
1889 (adjust_last_stmt): Likewise.
1890 (handle_builtin_strchr): Likewise.
1891 (handle_builtin_strcpy): Likewise.
1892 (handle_builtin_stxncpy): Likewise.
1893 (handle_builtin_memcpy): Likewise.
1894 (handle_builtin_strcat): Likewise.
1895 (strlen_check_and_optimize_stmt): Likewise.
1896 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1897 * tree-streamer-in.c: Likewise.
1898 * tree-streamer.c (record_common_node): Likewise.
1899 * tree.c (tree_code_size): Likewise.
1900 (wide_int_to_tree_1): Likewise.
1901 (type_contains_placeholder_1): Likewise.
1902 (build_common_tree_nodes): Likewise.
1903 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1904 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1905 (POINTER_BOUNDS_P): Likewise.
1906 (BOUNDED_TYPE_P): Likewise.
1907 (BOUNDED_P): Likewise.
1908 (CALL_WITH_BOUNDS_P): Likewise.
1909 (pointer_bounds_type_node): Likewise.
1910 * value-prof.c (gimple_ic): Likewise.
1911 * var-tracking.c (vt_add_function_parameters): Likewise.
1912 * varasm.c (make_decl_rtl): Likewise.
1913 (assemble_start_function): Likewise.
1914 (output_constant): Likewise.
1915 (maybe_assemble_visibility): Likewise.
1916 * varpool.c (ctor_for_folding): Likewise.
1917 * chkp-builtins.def: Remove.
1918 * ipa-chkp.c: Remove.
1919 * ipa-chkp.h: Remove.
1920 * rtl-chkp.c: Remove.
1921 * rtl-chkp.h: Remove.
1922 * tree-chkp-opt.c: Remove.
1923 * tree-chkp.c: Remove.
1924 * tree-chkp.h: Remove.
1926 2018-06-07 Carl Love <cel@us.ibm.com>
1928 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1929 vextract_fp_from_shortl): Add BE support.
1931 2018-06-07 Paul Koning <ni1d@arrl.net>
1933 * compare-elim.c (try_merge_compare): Don't merge compare if
1934 address contains a side effect.
1935 (try_eliminate_compare): Likewise.
1937 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1939 * config.gcc: Support "tremont".
1940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1941 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1943 * config/i386/i386.c (m_TREMONT): Define.
1944 (processor_target_table): Add "tremont".
1945 (PTA_TREMONT): Define.
1946 (ix86_lea_outperforms): Add TARGET_TREMONT.
1947 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1948 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1949 and M_INTEL_GOLDMONT_PLUS.
1950 (fold_builtin_cpu): Add "tremont".
1951 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1952 (ix86_option_override_internal): Add "tremont".
1953 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1954 (processor_type): Add PROCESSOR_TREMONT.
1955 * config/i386/x86-tune.def: Add m_TREMONT.
1956 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1958 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1960 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1961 symbol defined for msp430i* devices to be lower case.
1963 gcc/testsuite/gcc.target/msp430/
1964 * msp430i-device-symbol.c: New test.
1965 * msp430f-device-symbol.c: New test.
1966 * msp430.h: New test header file.
1968 2018-06-07 Richard Biener <rguenther@suse.de>
1970 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1971 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1972 Properly wrap signed arithmetic if overflow wraps.
1974 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1976 PR tree-optimization/69615
1977 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1978 of a cast from a same precision integral SSA_NAME in a bb dominated
1979 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1980 cast to utype if rhs2 has already a compatible type.
1982 2018-06-07 Richard Biener <rguenther@suse.de>
1984 PR tree-optimization/85935
1985 * graphite-scop-detection.c (find_params_in_bb): Analyze
1986 condition operands with respect to the correct loop. Assert
1987 the analysis doesn't fail.
1989 2018-06-04 Carl Love <cel@us.ibm.com>
1991 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1992 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1993 as it is slightly cheaper.
1994 (first_match_or_eos_index_<mode>):
1995 Calculate index using natural element order.
1996 (first_match_index_<mode>):
1997 Calculate index using natural element order.
1998 (first_match_or_eos_index_<mode>):
1999 Calculate index using natural order.
2000 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2001 for BE and LE modes.
2002 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2003 P9V_BUILTIN_VCLZLSBB_V16QI.
2004 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2007 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2009 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2010 indentation and line wrap for many prototypes. Add missing
2011 @smallexample directives around block of prototypes for vec_xl and
2014 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2016 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2017 track if we pass or return IEEE 128-bit floating point.
2018 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2019 C++ mangling that is compatible with GCC 8.1.
2020 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2021 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2022 floating point types.
2023 (rs6000_function_arg_advance_1): Likewise.
2024 (rs6000_mangle_type): Optionally generate mangled names that match
2025 what GCC 8.1 generated for IEEE 128-bit floating point types.
2026 (rs6000_globalize_decl_name): If we have an external function that
2027 passes or returns IEEE 128-bit types, generate a weak reference
2028 from the mangled name used in GCC 8.1 to the current mangled
2030 (rs6000_init_builtins): Make __ibm128 use the long double type if
2031 long double is IBM extended double. Make __float128 use the long
2032 double type if long double is IEEE 128-bit.
2035 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2036 macro for __ibm128 built-in functions.
2037 (PACK_IF): Add __ibm128 pack/unpack functions.
2038 (UNPACK_IF): Likewise.
2039 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2040 enable long double built-in functions if long double is IEEE
2041 128-bit floating point.
2042 (rs6000_invalid_builtin): Update long double built-in function
2044 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2045 functions, adjust the built-in function to use the long double
2046 built-in function if __ibm128 and long double are the same type.
2047 * doc/extend.texi (PowerPC builtins): Update documention for
2048 __builtin_{,un}pack_longdouble. Add documentation for
2049 __builtin_{,un}pack_ibm128.
2051 2018-06-06 Jim Wilson <jimw@sifive.com>
2053 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2054 (struct machine_function): New field interrupt_mode.
2055 (riscv_handle_type_attribute): New function. Add forward declaration.
2056 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2057 (riscv_expand_epilogue): Check interrupt_mode field.
2058 (riscv_set_current_function): Check interrupt attribute args and
2059 set interrupt_mode field.
2060 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2061 (riscv_sret, riscv_uret): New.
2062 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2063 new arguments to interrupt attribute.
2065 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2068 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2069 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2071 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2073 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2074 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2075 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2076 several redundant entries.
2078 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2080 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2081 type from "rtx" to "rtx_insn *".
2082 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2083 for local "call_insn", removing cast.
2084 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2086 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2088 PR tree-optimization/86066
2089 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2090 for BIT_INSERT_EXPR stores.
2092 2018-06-06 Richard Biener <rguenther@suse.de>
2094 PR tree-optimization/86062
2095 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2096 component refs ontop
2097 of to be offsetted base.
2099 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2101 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2102 to be static and remove check on interrupt attribute name.
2104 gcc/testsuite/gcc.target/msp430/
2105 * function-attributes-4.c: New test.
2106 * static-interrupts.c: New test.
2108 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2110 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2111 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2113 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2116 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2118 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2119 Remove second argument, change how error is called.
2120 (aarch64_layout_arg): Remove second argument from
2121 aarch64_err_no_fpadvsimd call.
2122 (aarch64_init_cumulative_args): Ditto.
2123 (aarch64_gimplify_va_arg_expr): Ditto.
2124 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2126 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2128 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2129 (*simple_return_indirect_internal<mode>): Rename from
2130 simple_return_indirect_internal. Use W mode iterator.
2131 (rstorssp): New expander.
2132 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2133 (clrssbsy): New expander.
2134 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2136 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2138 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2140 (cmse_check_pointed_object): Likewise.
2142 2018-06-05 Martin Liska <mliska@suse.cz>
2144 PR gcov-profile/47618
2145 * doc/invoke.texi: Document how -fprofile-dir format
2148 2018-06-05 Richard Biener <rguenther@suse.de>
2150 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2151 removal pretend DOM info isn't available so we do not update
2152 it and only remove edges, not dominated blocks. Actually free
2153 DOM info in case we removed something. Remove unreachable blocks.
2154 (mfb_keep_latches): Work with either DOM info or marked backedges.
2155 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2156 first. Mark backedges if DOM info isn't available.
2157 (Re-)compute DOM info after cleanup_control_flow_pre.
2159 2018-06-05 Richard Biener <rguenther@suse.de>
2161 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2162 (locus_discrim_hasher::hash): Adjust.
2163 (locus_discrim_hasher::equal): Likewise.
2164 (next_discriminator_for_locus): Work on line directly.
2165 (same_line_p): Pass in expanded locus1 as well.
2166 (assign_discriminators): Avoid redundant location expansions.
2168 2018-06-05 Richard Biener <rguenther@suse.de>
2170 PR tree-optimization/86046
2171 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2172 if required after clearing TREE_ADDRESSABLE.
2174 2018-06-05 Richard Biener <rguenther@suse.de>
2176 PR tree-optimization/86047
2177 * tree-ssa-loop.c (for_each_index): Glob handling of all
2178 decls and constants and really handle all of them.
2180 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2183 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2184 qualifier_void_pointer and qualifier_const_void_pointer.
2185 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2186 (arm_init_builtins): Handle the above.
2187 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2188 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2189 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2192 2018-06-05 Martin Liska <mliska@suse.cz>
2194 * auto-profile.c (read_autofdo_file): Do not use
2195 gcov_ctr_summary struct.
2196 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2197 * coverage.c (struct counts_entry): Likewise.
2198 (read_counts_file): Read just single summary entry.
2199 (get_coverage_counts): Use gcov_summary struct.
2200 * coverage.h (get_coverage_counts): Likewise.
2201 * gcov-dump.c (dump_working_sets): Likewise.
2202 (tag_summary): Dump just single summary.
2203 * gcov-io.c (gcov_write_summary): Write just histogram
2205 (gcov_read_summary): Read just single summary.
2206 (compute_working_sets): Use gcov_summary struct.
2207 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2208 of GCOV_COUNTERS_SUMMABLE.
2209 (GCOV_COUNTERS_SUMMABLE): Remove.
2210 (GCOV_FIRST_VALUE_COUNTER): Replace with
2211 GCOV_COUNTER_V_INTERVAL.
2212 (struct gcov_ctr_summary): Remove.
2213 (struct gcov_summary): Directly use fields of former
2215 (compute_working_sets): Use gcov_summary struct.
2216 * gcov.c (read_count_file): Do not use ctrs fields.
2217 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2219 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2221 * profile.c: Likewise.
2222 * profile.h: Likewise.
2224 2018-06-05 Martin Liska <mliska@suse.cz>
2226 PR gcov-profile/84846
2227 * gcov.c (output_lines): Print working directory only
2228 in intermediate format.
2230 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2232 * config/s390/s390-builtin-types.def: Add void function type.
2233 * config/s390/s390-builtins.def: Use the function type for the
2236 2018-06-04 Jim Wilson <jimw@sifive.com>
2238 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2240 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2241 and maybe_eh_return. Change regno to unsigned int. Use new args to
2242 handle EH_RETURN_DATA_REGNO registers properly.
2243 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2244 (riscv_expand_epilogue): Update comment. Change argument name and
2245 type. Update code to use new name and type. Pass new args to
2246 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2248 * config/riscv/riscv.md (NORMAL_RETURN): New.
2249 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2250 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2251 (eh_return): Call gen_eh_return_internal and emit barrier.
2252 (eh_return_internal): Call riscv_expand_epilogue.
2254 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2256 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2257 bit_insertion field and declare can_be_merged_into method.
2258 (merged_store_group::can_be_merged_into): New method.
2259 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2260 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2261 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2263 2018-06-04 Richard Biener <rguenther@suse.de>
2265 PR tree-optimization/85955
2266 * builtins.c (fold_builtin_sincos): Convert pointers to
2267 destination to appropriate type before dereferencing.
2269 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2271 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2273 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2275 * expr.c (expand_expr_real_1): Force the operand into memory if
2276 its TYPE_MODE is BLKmode and if there is no integer mode for
2277 the number of bits being extracted.
2279 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2283 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2284 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2286 2018-06-04 Richard Biener <rguenther@suse.de>
2288 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2289 (cleanup_tree_cfg_noloop): ... single caller. Do
2290 start_recording_case_labels later.
2292 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2294 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2295 to _IMMINTRIN_H_INCLUDED.
2296 * config/i386/pconfigintrin.h: Ditto.
2297 * config/i386/waitpkgintrin.h: Ditto.
2298 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2299 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2300 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2301 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2302 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2303 waitpkgintrin.h and cldemoteintrin.h.
2305 2018-06-04 Richard Biener <rguenther@suse.de>
2307 PR tree-optimization/86038
2308 * tracer.c (find_best_successor): Check probability for
2309 being initialized, bail out if not.
2311 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2314 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2315 of bits to ignore when comparing architectures.
2317 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2319 PR tree-optimization/69615
2320 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2321 maximum or minimum of the type, try to merge it also as if
2322 range1 is + [-, x - 1] or + [x + 1, -].
2325 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2327 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2329 PR tree-optimization/86034
2330 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2331 the unsigned bitfield type in a bit insertion sequence if it does not
2332 have a larger precision than the bitfield size.
2333 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2335 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2337 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2339 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2341 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2342 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2343 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2344 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2346 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2348 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2349 Disable -fdelete-null-pointer-checks for ELF toolchain.
2351 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2352 Kito Cheng <kito.cheng@gmail.com>
2354 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2355 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2356 (nds32*-*-*): Add glibc and uclibc conditions.
2357 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2358 (TARGET_EXCEPT_UNWIND_INFO): Define.
2359 * config/nds32/elf.h: New file.
2360 * config/nds32/linux.h: New file.
2361 * config/nds32/nds32-elf.opt: New file.
2362 * config/nds32/nds32-linux.opt: New file.
2363 * config/nds32/nds32-fp-as-gp.c
2364 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2365 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2367 (nds32_asm_file_end): Ditto.
2368 (nds32_print_operand): Ditto.
2369 (nds32_insert_attributes): Ditto.
2370 (nds32_init_libfuncs): New function.
2371 (TARGET_HAVE_TLS): Define.
2372 (TARGET_INIT_LIBFUNCS): Define.
2373 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2375 (TARGET_ELF): Apply different mcmodel setting.
2376 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2377 been migrated into elf.h and linux.h files.
2378 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2379 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2380 (mcmodel): The content has been migrated into nds32-elf.opt and
2381 nds32-linux.opt files.
2382 * config/nds32/t-elf: New file.
2383 * config/nds32/t-linux: New file.
2385 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2386 Shiva Chen <shiva0217@gmail.com>
2388 * config/nds32/constants.md (unspec_volatile_element): Add
2389 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2390 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2392 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2393 (make_pass_nds32_fp_as_gp): Declare.
2394 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2396 (nds32_asm_function_end_prologue): Remove unused asm output.
2397 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2398 (nds32_asm_file_start): Output necessary fp_as_gp information.
2399 (nds32_option_override): Adjust register usage.
2400 (nds32_expand_prologue): Consider fp_as_gp situation.
2401 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2402 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2405 (simple_return): Ditto.
2406 (omit_fp_begin): Output special directive for fp_as_gp.
2407 (omit_fp_end): Output special directive for fp_as_gp.
2408 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2409 mforbid-fp-as-gp): New options.
2411 2018-06-01 Mark Wielaard <mark@klomp.org>
2413 * dwarf2out.c (dwarf2out_finish): Remove generation of
2414 DW_AT_loclists_base.
2416 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2418 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2419 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2420 (struct merged_store_group): Add bit_insertion field.
2421 (dump_char_array): Use standard hexadecimal format.
2422 (merged_store_group::merged_store_group): Set bit_insertion to false.
2423 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2424 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2425 also print the mask in the dump file.
2426 (pass_store_merging::gate): Minor tweak.
2427 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2428 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2429 stores with INTEGER_CST stores.
2430 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2431 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2432 and use it throughout. Generate bit insertion sequences if need be.
2433 (pass_store_merging::process_store): Remove redundant condition.
2434 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2436 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2438 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2439 the 128-bit floating point types. Fix function comment.
2441 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2443 * config/aarch64/aarch64-simd.md
2444 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2446 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2449 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2451 PR tree-optimization/85989
2452 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2454 (backprop::intersect_uses): Check it when deciding whether this
2455 is a backedge reference.
2456 (backprop::process_block): Add each phi to m_visited_phis
2457 after visiting it, then clear it at the end.
2459 2018-06-01 Richard Biener <rguenther@suse.de>
2461 * tree-vectorizer.h (vect_dr_stmt): New function.
2462 (vect_get_load_cost): Adjust.
2463 (vect_get_store_cost): Likewise.
2464 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2465 Use vect_dr_stmt instead of DR_SMTT.
2466 (vect_record_base_alignments): Likewise.
2467 (vect_calculate_target_alignment): Likewise.
2468 (vect_compute_data_ref_alignment): Likewise and make static.
2469 (vect_update_misalignment_for_peel): Likewise.
2470 (vect_verify_datarefs_alignment): Likewise.
2471 (vector_alignment_reachable_p): Likewise.
2472 (vect_get_data_access_cost): Likewise. Pass down
2473 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2474 (vect_get_peeling_costs_all_drs): Likewise.
2475 (vect_peeling_hash_get_lowest_cost): Likewise.
2476 (vect_enhance_data_refs_alignment): Likewise.
2477 (vect_find_same_alignment_drs): Likewise.
2478 (vect_analyze_data_refs_alignment): Likewise.
2479 (vect_analyze_group_access_1): Likewise.
2480 (vect_analyze_group_access): Likewise.
2481 (vect_analyze_data_ref_access): Likewise.
2482 (vect_analyze_data_ref_accesses): Likewise.
2483 (vect_vfa_segment_size): Likewise.
2484 (vect_small_gap_p): Likewise.
2485 (vectorizable_with_step_bound_p): Likewise.
2486 (vect_prune_runtime_alias_test_list): Likewise.
2487 (vect_analyze_data_refs): Likewise.
2488 (vect_supportable_dr_alignment): Likewise.
2489 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2490 (vect_gen_prolog_loop_niters): Likewise.
2491 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2492 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2494 (vect_recog_mask_conversion_pattern): Likewise.
2495 (vect_try_gather_scatter_pattern): Likewise.
2496 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2497 to vect_get_store_cost.
2498 (vect_get_store_cost): Get stmt_info instead of DR.
2499 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2500 (vect_get_load_cost): Get stmt_info instead of DR.
2502 2018-06-01 Richard Biener <rguenther@suse.de>
2505 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2506 vars wrapped in MEM_REFs.
2508 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2510 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2511 Fix subreg tests so that we only return a choice between
2512 GENERAL_REGS and FP_REGS if the original classes included both.
2514 2018-06-01 Richard Biener <rguenther@suse.de>
2517 * tree-ssa-structalias.c (get_function_part_constraint):
2518 Handle NULL fi->decl.
2519 (find_func_aliases_for_call): Properly handle indirect
2520 fi from direct call.
2521 (find_func_clobbers): Likewise.
2522 (ipa_pta_execute): Likewise.
2523 (create_variable_info_for): For functions that are ifunc_resolver
2524 resolve to a varinfo that contains the result of the resolver call.
2525 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2528 2018-05-31 Michael Collison <michael.collison@arm.com>
2530 * config/aarch64/aarch64.md:
2531 (*fix_to_zero_extenddfdi2): New pattern.
2532 * gcc.target/aarch64/fix_extend1.c: New testcase.
2534 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2538 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2539 and BUILT_IN_STRNCMP_EQ.
2540 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2541 BUILT_IN_STRNCMP_EQ.
2542 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2543 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2544 (gimple_fold_builtin): Likewise.
2545 * tree-ssa-strlen.c (compute_string_length): New function.
2546 (determine_min_obsize): New function.
2547 (handle_builtin_string_cmp): New function to handle calls to
2548 string compare functions.
2549 (strlen_optimize_stmt): Add handling to builtin string compare
2551 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2552 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2553 * tree.c (build_common_builtin_nodes): Add new defines of
2554 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2556 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2559 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2560 functions with naked attribute.
2562 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2564 * config/i386/sse.md (avx_vec_concat<mode>):
2565 Substitute concat_tg_mode mode attribute with xtg_mode.
2566 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2567 (concat_tg_mode): Remove mode attribute.
2569 2018-05-31 Martin Sebor <msebor@redhat.com>
2572 * calls.c (alloc_max_size): Correct a logic error/typo.
2573 Treat excessive arguments as infinite. Warn for invalid arguments.
2574 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2576 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2579 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2580 and movx for Haswell.
2582 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2583 Cesar Philippidis <cesar@codesourcery.com>
2586 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2587 when emitting error on private/firstprivate reductions.
2588 * omp-low.c (lower_omp_target): Avoid reference-type processing
2589 on pointers for firstprivate clause.
2591 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2593 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2596 * config/aarch64/aarch64-simd.md
2597 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2598 (aarch64_ld1_x3_<mode>): Likewise
2599 (aarch64_st1x2<VALLDIF:mode>): Likewise
2600 (aarch64_st1_x2_<mode>): Likewise
2601 (aarch64_st1x3<VALLDIF:mode>): Likewise
2602 (aarch64_st1_x3_<mode>): Likewise
2603 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2604 (vld1_s8_x3): Likewise.
2605 (vld1_u16_x3): Likewise.
2606 (vld1_s16_x3): Likewise.
2607 (vld1_u32_x3): Likewise.
2608 (vld1_s32_x3): Likewise.
2609 (vld1_u64_x3): Likewise.
2610 (vld1_s64_x3): Likewise.
2611 (vld1_f16_x3): Likewise.
2612 (vld1_f32_x3): Likewise.
2613 (vld1_f64_x3): Likewise.
2614 (vld1_p8_x3): Likewise.
2615 (vld1_p16_x3): Likewise.
2616 (vld1_p64_x3): Likewise.
2617 (vld1q_u8_x3): Likewise.
2618 (vld1q_s8_x3): Likewise.
2619 (vld1q_u16_x3): Likewise.
2620 (vld1q_s16_x3): Likewise.
2621 (vld1q_u32_x3): Likewise.
2622 (vld1q_s32_x3): Likewise.
2623 (vld1q_u64_x3): Likewise.
2624 (vld1q_s64_x3): Likewise.
2625 (vld1q_f16_x3): Likewise.
2626 (vld1q_f32_x3): Likewise.
2627 (vld1q_f64_x3): Likewise.
2628 (vld1q_p8_x3): Likewise.
2629 (vld1q_p16_x3): Likewise.
2630 (vld1q_p64_x3): Likewise.
2631 (vst1_s64_x2): Likewise.
2632 (vst1_u64_x2): Likewise.
2633 (vst1_f64_x2): Likewise.
2634 (vst1_s8_x2): Likewise.
2635 (vst1_p8_x2): Likewise.
2636 (vst1_s16_x2): Likewise.
2637 (vst1_p16_x2): Likewise.
2638 (vst1_s32_x2): Likewise.
2639 (vst1_u8_x2): Likewise.
2640 (vst1_u16_x2): Likewise.
2641 (vst1_u32_x2): Likewise.
2642 (vst1_f16_x2): Likewise.
2643 (vst1_f32_x2): Likewise.
2644 (vst1_p64_x2): Likewise.
2645 (vst1q_s8_x2): Likewise.
2646 (vst1q_p8_x2): Likewise.
2647 (vst1q_s16_x2): Likewise.
2648 (vst1q_p16_x2): Likewise.
2649 (vst1q_s32_x2): Likewise.
2650 (vst1q_s64_x2): Likewise.
2651 (vst1q_u8_x2): Likewise.
2652 (vst1q_u16_x2): Likewise.
2653 (vst1q_u32_x2): Likewise.
2654 (vst1q_u64_x2): Likewise.
2655 (vst1q_f16_x2): Likewise.
2656 (vst1q_f32_x2): Likewise.
2657 (vst1q_f64_x2): Likewise.
2658 (vst1q_p64_x2): Likewise.
2659 (vst1_s64_x3): Likewise.
2660 (vst1_u64_x3): Likewise.
2661 (vst1_f64_x3): Likewise.
2662 (vst1_s8_x3): Likewise.
2663 (vst1_p8_x3): Likewise.
2664 (vst1_s16_x3): Likewise.
2665 (vst1_p16_x3): Likewise.
2666 (vst1_s32_x3): Likewise.
2667 (vst1_u8_x3): Likewise.
2668 (vst1_u16_x3): Likewise.
2669 (vst1_u32_x3): Likewise.
2670 (vst1_f16_x3): Likewise.
2671 (vst1_f32_x3): Likewise.
2672 (vst1_p64_x3): Likewise.
2673 (vst1q_s8_x3): Likewise.
2674 (vst1q_p8_x3): Likewise.
2675 (vst1q_s16_x3): Likewise.
2676 (vst1q_p16_x3): Likewise.
2677 (vst1q_s32_x3): Likewise.
2678 (vst1q_s64_x3): Likewise.
2679 (vst1q_u8_x3): Likewise.
2680 (vst1q_u16_x3): Likewise.
2681 (vst1q_u32_x3): Likewise.
2682 (vst1q_u64_x3): Likewise.
2683 (vst1q_f16_x3): Likewise.
2684 (vst1q_f32_x3): Likewise.
2685 (vst1q_f64_x3): Likewise.
2686 (vst1q_p64_x3): Likewise.
2688 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2690 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2691 user_label_prefix to name.
2693 * tree-core.h: Update comment about the format of NAME string
2694 passed to handler in attribute_spec.
2696 * config/msp430/msp430.md: Remove erroneous subreg expression from
2697 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2698 zero_extend{q,h}isi2.
2700 2018-05-30 Borislav Petkov <bp@suse.de>
2702 * doc/extend.texi: Document some architecture specific
2703 constraints and sort entries.
2705 2018-05-30 Martin Sebor <msebor@redhat.com>
2708 * builtins.c (expand_builtin_stpcpy_1): New function.
2709 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2710 only if the former succeeds.
2712 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2714 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2717 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2719 * doc/invoke.texi (-flinker-output): Document
2721 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2723 * passes.c (ipa_write_summaries): Only modify statements if body
2725 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2726 incrementally linking.
2727 (ipa_passes): Likewise.
2728 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2729 pass down resolution info.
2730 * common.opt (flag_incremental_link): Update info.
2731 * gcc.c (plugin specs): Turn flinker-output=* to
2732 -plugin-opt=-linker-output-known
2733 * toplev.c (compile_file): Also cut compilation when doing incremental
2735 * flag-types. (enum lto_partition_model): Add
2736 LTO_LINKER_OUTPUT_NOLTOREL.
2737 (invoke.texi): Add -flinker-output docs.
2738 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2739 link same way as WPA; do not stream in dead initializers.
2741 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2742 darf2out_register_external_decl): Support incremental link.
2744 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2746 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2748 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2750 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2751 it down to simple_object_copy_lto_debug_sections.
2752 (run_gcc): Determine incremental LTO link time and configure
2753 lto1 into non-wpa mode, disable renaming of debug sections.
2755 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2757 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2758 descriptions of various incorrectly documented functions.
2760 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2763 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2764 address check not strict.
2766 2018-05-30 Richard Biener <rguenther@suse.de>
2768 PR tree-optimization/85964
2769 * tracer.c (better_p): Drop initialized count check, we only
2770 call the function with initialized counts now.
2771 (find_best_successor): Do find a best edge if one
2772 has uninitialized count.
2773 (find_best_predecessor): Likewise. Do BB frequency check only
2774 if count is initialized.
2776 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2778 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2779 (aarch64_ldrstr_offset_compare): New.
2780 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2781 load/store orderings.
2782 (aarch64_gen_adjusted_ldpstp): Likewise.
2784 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2786 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2787 Check for subset of GENERAL_REGS and FP_REGS.
2788 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2791 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2793 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2794 and wi::to_poly_offset. Add the current offset and then check
2795 whether the sum fits, rather than using an unchecked addition of
2796 a checked term. Check for a shwi rather than a uhwi.
2797 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2798 (store_constructor): Use poly_int_tree_p.
2799 (expand_expr_real_1): Likewise.
2800 * function.c (assign_temp): Likewise.
2801 * fold-const.c (const_binop): Use poly_int_tree_p and
2803 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2805 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2806 to_poly_offset for MEM offsets.
2807 * ipa-icf.c (sem_variable::equals): Likewise.
2808 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2809 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2810 wi::to_poly_offset for BIT_FIELD_REF offsets.
2811 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2813 * var-tracking.c (emit_note_insn_var_location): Use
2814 tree_to_poly_uint64.
2816 2018-05-29 Jim Wilson <jimw@sifive.com>
2818 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2820 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2823 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2824 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2826 (sse4_1_round<mode>2): Use nonimmediate_operand
2827 for operand 1 predicate.
2829 2018-05-29 Martin Sebor <msebor@redhat.com>
2830 Richard Biener <rguenther@suse.de>
2833 * calls.c (get_size_range): Call determine_value_range instead
2834 of get_value_range..
2835 * tree-vrp.h (determine_value_range): Declared new function.
2836 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2838 2018-05-29 Richard Biener <rguenther@suse.de>
2840 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2841 sure to use non-pattern stmts for get_earlier_stmt arguments.
2842 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2843 called on pattern stmts.
2844 (get_later_stmt): Likewise.
2846 2018-05-29 Martin Liska <mliska@suse.cz>
2848 PR gcov-profile/85759
2849 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2852 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2854 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2856 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2860 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2861 VEC_PACK_FLOAT_EXPR): New tree codes.
2862 * tree-pretty-print.c (op_code_prio): Handle
2863 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2864 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2865 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2866 * tree-inline.c (estimate_operator_cost): Likewise.
2867 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2868 * fold-const.c (const_binop): Likewise.
2869 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2870 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2871 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2872 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2873 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2874 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2875 * expr.c (expand_expr_real_2): Likewise.
2876 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2877 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2878 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2880 * optabs.c (expand_widen_pattern_expr): For
2881 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2882 sign from result type rather than operand's type.
2883 (expand_binop_directly): For vec_packu_float_optab and
2884 vec_packs_float_optab allow result type to be different from operand's
2886 * optabs-tree.c (optab_for_tree_code): Handle
2887 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2888 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2889 * tree-vect-generic.c (expand_vector_operations_1): Handle
2890 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2891 VEC_PACK_FLOAT_EXPR.
2892 * tree-vect-stmts.c (supportable_widening_operation): Handle
2894 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2895 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2896 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2897 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2898 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2900 (vec_pack<floatprefix>_float_<mode>): New expander.
2901 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2903 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2904 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2905 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2906 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2907 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2909 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2910 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2911 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2912 VEC_PACK_FLOAT_EXPR): Document.
2914 2018-05-29 Richard Biener <rguenther@suse.de>
2916 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2918 (stmt_vec_info_vec): Make pointer.
2919 (init_stmt_vec_info_vec): Remove.
2920 (free_stmt_vec_info_vec): Likewise.
2921 (set_stmt_vec_info_vec): New function.
2922 (free_stmt_vec_infos): Likewise.
2923 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2924 (set_vinfo_for_stmt): Likewise.
2925 (get_earlier_stmt): Likewise.
2926 (get_later_stmt): Likewise.
2927 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2928 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2929 (vec_info::~vec_info): Free stmt_vec_infos.
2930 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2931 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2932 (pass_slp_vectorize::execute): Likewise.
2933 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2934 (free_stmt_vec_info_vec): Likewise.
2935 (set_stmt_vec_info_vec): New function.
2936 (free_stmt_vec_infos): Likewise.
2937 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2938 the global stmt_vec_info_vec.
2939 * tree-parloops.c (gather_scalar_reductions): Use
2940 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2943 2018-05-29 Richard Biener <rguenther@suse.de>
2945 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2947 2018-05-29 Martin Liska <mliska@suse.cz>
2948 David Malcolm <dmalcolm@redhat.com>
2950 * vec.c (test_reverse): New.
2951 (vec_c_tests): Add new test.
2952 * vec.h (vl_ptr>::reverse): New function.
2954 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2956 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2958 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2961 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2963 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2965 2018-05-28 Richard Biener <rguenther@suse.de>
2967 PR tree-optimization/85933
2968 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2969 look at stmts marked as vectorizable.
2971 2018-05-28 Richard Biener <rguenther@suse.de>
2973 PR tree-optimization/85934
2974 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2975 vector boolean check before scalar optimization.
2977 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2979 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2982 2018-05-28 Mark Wielaard <mark@klomp.org>
2984 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2985 if it is an expression containing a minus sign.
2987 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2989 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2991 2018-05-27 Paul Koning <ni1d@arrl.net>
2993 * config/pdp11/pdp11.md (truncsihi2): Remove.
2995 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2996 Chung-Ju Wu <jasonwucj@gmail.com>
2998 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3000 (unaligned_store_dw): Ditto.
3001 * config/nds32/nds32-memory-manipulation.c
3002 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3003 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3004 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3005 (emit_setmem_word_loop): Rename to ...
3006 (emit_setmem_doubleword_loop): ... this.
3007 (nds32_gen_dup_4_byte_to_word_value): New function.
3008 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3009 (nds32_expand_setmem_loop): Refine implementation.
3010 (nds32_expand_setmem_loop_v3m): Ditto.
3011 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3014 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3016 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3018 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3020 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3021 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3022 machine->attr_no_prologue_p.
3023 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3024 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3025 (nds32_expand_epilogue): Consider attr_naked_p.
3026 (nds32_expand_epilogue_v3pop): Likewise.
3027 (nds32_can_use_return_insn): Likewise.
3028 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3029 attr_no_prologue_p fields.
3030 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3032 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3035 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3037 * config/i386/sse.md
3038 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3040 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3042 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3044 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3046 (*<floatsuffix>floatv2div2sf2): Rename to ...
3047 (*float<floatunssuffix>v2div2sf2): ... this.
3048 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3049 (float<floatunssuffix>v2div2sf2_mask): ... this.
3050 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3051 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3052 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3054 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3056 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3058 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3060 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3062 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3064 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3065 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3066 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3067 gen_ufix_truncv8dfv8si2.
3068 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3069 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3070 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3071 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3072 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3073 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3074 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3075 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3077 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3081 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3083 2018-05-25 Jim Wilson <jimw@sifive.com>
3085 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3086 * config/riscv/riscv.c (struct machine_function): Add
3087 interrupt_handler_p and attribute_checked_p fields.
3088 (riscv_attribute_table): Add interrupt.
3089 (riscv_interrupt_type_p): New.
3090 (riscv_save_reg_p): Save extra regs for interrupt handler.
3091 (riscv_use_save_libcall): Return false for interrupt handler.
3092 (riscv_first_stack_step): Add forward declaration.
3093 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3094 for interrupt handler with large frame. Use it for saved reg list.
3095 (riscv_expand_prologue): Move flag_stack_usage_info support to
3096 eliminate duplication.
3097 (riscv_expand_epilogue): Generate mret for interrupt handler.
3098 (riscv_epilogue_uses): New.
3099 (riscv_can_use_return_insn): Return false for interrupt handler.
3100 (riscv_function_ok_for_sibcall): Likewise.
3101 (riscv_set_current_function): Add interrupt handler support.
3102 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3103 * config/riscv/riscv.md (UNSPECV_MRET): New.
3105 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3107 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3109 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3111 PR tree-optimization/85712
3112 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3113 this candidate has already been replaced in-situ by a copy.
3115 2018-05-25 Jason Merrill <jason@redhat.com>
3117 PR c++/80485 - inline function non-zero address.
3118 * symtab.c (nonzero_address): Check DECL_COMDAT.
3120 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3123 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3124 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3125 extension of SImode operation. Use const123_operand predicate.
3126 (*saddsi_1): Remove.
3127 (*saddl_se_1): Ditto.
3129 (*ssubl_se_1): Ditto.
3130 * config/alpha/predicates.md (const123_operand): New predicate.
3131 * config/alpha/constraints.md (P): Use IN_RANGE.
3133 2018-05-25 Richard Biener <rguenther@suse.de>
3135 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3137 (ref_maybe_used_by_stmt_p): Likewise.
3138 (stmt_may_clobber_ref_p): Likewise.
3139 (stmt_may_clobber_ref_p_1): Likewise.
3140 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3142 (ref_maybe_used_by_stmt_p): Likewise.
3143 (stmt_may_clobber_ref_p): Likewise.
3144 (stmt_may_clobber_ref_p_1): Likewise.
3145 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3146 the alias oracle to disambiguate DRs with stmts DR analysis
3148 (vect_analyze_data_refs): Do not give up on not analyzable
3149 DRs for BB vectorization. Remove code truncating the dataref
3152 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3155 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3156 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3157 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3159 2018-05-25 Richard Biener <rguenther@suse.de>
3161 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3162 function, combining stmt data ref gathering and fatal analysis
3164 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3165 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3166 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3167 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3168 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3170 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3172 PR tree-optimization/85720
3173 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3174 SCC if all partitions are builtins.
3175 (version_loop_by_alias_check): New parameter. Generate cancelable
3176 runtime alias check if all partitions are builtins.
3177 (distribute_loop): Update call to above function.
3179 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3181 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3182 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3183 (parm_default_def_partition_arg): Ditto.
3184 (set_parm_default_def_partition): Ditto.
3185 (get_parm_default_def_partitions): Ditto and make it static.
3186 (get_undefined_value_partitions): Ditto and make it static.
3187 (remove_ssa_form): Refactor call to init_var_map here.
3188 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3189 computation for loop region.
3190 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3191 (register_default_def): Delete.
3192 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3193 (parm_default_def_partition_arg): Ditto.
3194 (set_parm_default_def_partition): Ditto.
3195 (get_parm_default_def_partitions): Ditto and make it static.
3196 (get_undefined_value_partitions): Ditto and make it static.
3197 (coalesce_with_default, coalesce_with_default): Update comment.
3198 (create_coalesce_list_for_region): New func factored out from
3199 create_outofssa_var_map.
3200 (populate_coalesce_list_for_outofssa): New func factored out from
3201 create_outofssa_var_map and coalesce_ssa_name.
3202 (create_outofssa_var_map): Delete.
3203 (coalesce_ssa_name): Refactor to support live range computation.
3204 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3205 (get_parm_default_def_partitions): Delete.
3206 (get_undefined_value_partitions): Ditto.
3207 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3208 computation for loop region.
3209 (new_tree_live_info, loe_visit_block): Ditto.
3210 (live_worklist, set_var_live_on_entry): Ditto.
3211 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3212 * tree-ssa-live.h (struct _var_map): New fields.
3213 (init_var_map): Change decl.
3214 (region_contains_p): New.
3216 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3218 * tree-ssa-live.h (live_merge_and_clear): Delete.
3220 2018-05-25 Richard Biener <rguenther@suse.de>
3223 * tree-dump.c (dequeue_and_dump): Remove access to removed
3224 operand 2 of a SWITCH_EXPR.
3226 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3228 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3229 multiplication and division.
3230 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3231 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3232 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3233 (cond_udiv_optab, cond_umod_optab): New optabs.
3234 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3235 (IFN_COND_RDIV): New internal functions.
3236 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3237 TRUNC_MOD_EXPR and RDIV_EXPR.
3238 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3239 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3241 (SVE_INT_BINARY): Include mult.
3242 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3243 (optab, sve_int_op): Handle mult.
3244 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3246 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3247 for SVE_INT_BINARY_SD.
3249 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3251 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3252 (optab, sve_int_op): Handle div and udiv.
3253 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3254 for SVE_INT_BINARY_SD.
3255 (*<optab><mode>3): New insn for the same.
3257 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3259 * tree-vect-patterns.c: Include predict.h.
3260 (vect_recog_divmod_pattern): Restrict check for division support
3261 to when optimizing for size.
3263 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3265 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3266 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3267 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3268 (gimple_match_op::set_op): Likewise.
3269 (gimple_resimplify4): Declare.
3270 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3271 (expr::gen_transform): Likewise.
3272 (decision_tree::gen): Generate a simplification routine for 4 operands.
3273 * gimple-match-head.c (gimple_simplify): Add an overload for
3274 4 operands. In the top-level function, handle up to 4 call
3275 arguments and call gimple_resimplify4.
3276 (gimple_resimplify4): New function.
3277 (build_call_internal): Pass a fourth operand.
3278 (maybe_push_to_seq): Likewise.
3279 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3280 Fold VEC_COND_EXPRs of an operation and a default value into
3281 an IFN_COND_* function if possible.
3282 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3284 (SVE_COND_FP_BINARY): Include them.
3285 (optab, sve_fp_op): Handle them.
3286 (SVE_INT_BINARY_REV): New code iterator.
3287 (SVE_COND_FP_BINARY_REV): New int iterator.
3288 (commutative): New int attribute.
3289 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3291 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3293 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3294 (*cond_<optab><mode>): New patterns for reversed operands.
3296 2018-05-25 Richard Biener <rguenther@suse.de>
3298 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3299 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3300 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3301 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3302 * tree-vect-data-refs.c (everywhere): Adjust users.
3303 * tree-vect-loop.c (everywhere): Likewise.
3304 * tree-vect-slp.c (everywhere): Likewise.
3305 * tree-vect-stmts.c (everywhere): Likewise.
3306 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3308 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3310 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3312 (gcc_cv_as_section_exclude): ... this.
3313 Try Solaris as #exclude syntax.
3314 * configure: Regenerate.
3315 * config.in: Regenerate.
3316 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3318 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3319 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3321 * varasm.c (default_elf_asm_named_section): Don't check if
3322 HAVE_GAS_SECTION_EXCLUDE is defined.
3324 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3326 * doc/md.texi: Update the documentation of the cond_* optabs
3327 to mention the new final operand. Fix GET_MODE_NUNITS call.
3328 Describe the scalar case too.
3329 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3330 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3332 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3333 (get_conditional_internal_fn): Update comment.
3334 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3335 accumulator value as a final argument to conditional functions.
3336 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3337 a define_expand and add an "else" operand. Assert for now that
3338 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3339 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3340 (*cond_<optab><mode>): New patterns.
3341 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3342 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3343 (UNSPEC_COND_EOR): Delete.
3344 (optab): Remove associated mappings.
3345 (SVE_INT_BINARY): New code iterator.
3346 (sve_int_op): Remove int attribute and add "minus" to the code
3348 (SVE_COND_INT_OP): Delete.
3349 (SVE_COND_FP_OP): Rename to...
3350 (SVE_COND_FP_BINARY): ...this.
3352 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3354 * optabs.c (can_reuse_operands_p): New function.
3355 (maybe_legitimize_operands): Try to reuse the results for
3358 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3360 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3361 Add {q} suffix to insn mnemonic.
3363 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3365 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3366 (msp430_warn_func_return): New.
3368 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3370 * fold-const.c (tree_nonzero_bits): New function.
3371 * fold-const.h (tree_nonzero_bits): Likewise.
3372 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3373 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3375 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3379 * varasm.c (assemble_alias): Check ifunc_resolver only on
3382 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3385 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3386 when memory input operand is handled.
3388 2018-05-24 Luis Machado <luis.machado@linaro.org>
3390 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3392 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3394 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3396 * match.pd: Delay FMA folds until after vectorization.
3398 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3401 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3402 address check not strict.
3404 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3406 * gimple-match.h (gimple_match_op): New class.
3407 (mprts_hook): Replace parameters with a gimple_match_op *.
3408 (maybe_build_generic_op): Likewise.
3409 (gimple_simplified_result_is_gimple_val): Replace parameters with
3410 a const gimple_match_op *.
3411 (gimple_simplify): Replace code_helper * and tree * parameters with
3412 a gimple_match_op * parameter.
3413 (gimple_resimplify1): Replace code_helper *, tree and tree *
3414 parameters with a gimple_match_op * parameter.
3415 (gimple_resimplify2): Likewise.
3416 (gimple_resimplify3): Likewise.
3417 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3418 parameters with a gimple_match_op * parameter.
3419 * gimple-match-head.c (gimple_simplify): Change prototypes of
3420 auto-generated functions to take a gimple_match_op * instead of
3421 separate code_helper * and tree * parameters. Make the same
3422 change in the top-level overload and update calls to the
3423 gimple_resimplify routines. Update calls to the auto-generated
3424 functions and to maybe_push_res_to_seq in the publicly-facing
3425 operation-specific gimple_simplify overloads.
3426 (gimple_match_op::MAX_NUM_OPS): Define.
3427 (gimple_resimplify1): Replace rcode and ops with a single res_op
3428 parameter. Update call to gimple_simplify.
3429 (gimple_resimplify2): Likewise.
3430 (gimple_resimplify3): Likewise.
3431 (mprts_hook): Replace parameters with a gimple_match_op *.
3432 (maybe_build_generic_op): Likewise.
3433 (build_call_internal): Replace type, nargs and ops with
3434 a gimple_match_op *.
3435 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3436 with a single gimple_match_op *. Update calls to mprts_hook,
3437 build_call_internal and gimple_simplified_result_is_gimple_val.
3438 Factor out code that is common to the tree_code and combined_fn cases.
3439 * genmatch.c (expr::gen_transform): Replace tem_code and
3440 tem_ops with a gimple_match_op called tem_op. Update calls
3441 to the gimple_resimplify functions and maybe_push_res_to_seq.
3442 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3443 res_ops. Update call to the gimple_resimplify functions.
3444 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3445 (decision_tree::gen): Make the functions take a gimple_match_op *
3446 called res_op instead of separate res_code and res_ops parameters.
3447 Update call accordingly.
3448 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3449 and ops with a single res_op parameter. Update calls to
3450 maybe_build_generic_op and maybe_push_res_to_seq.
3451 (fold_stmt_1): Update calls to gimple_simplify and
3452 replace_stmt_with_simplification.
3453 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3454 and gimple_simplified_result_is_gimple_val.
3455 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3457 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3458 with a gimple_match_op *.
3459 (vn_nary_build_or_lookup): Likewise. Update call to
3460 vn_nary_build_or_lookup_1.
3461 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3462 gimple_match_op *. Update calls to the gimple_resimplify routines
3463 and to gimple_simplified_result_is_gimple_val.
3464 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3465 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3466 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3467 (visit_nary_op): Likewise.
3468 (visit_reference_op_load): Likewise.
3470 2018-05-23 Luis Machado <luis.machado@linaro.org>
3472 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3473 modifier for printing the step amount.
3475 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3478 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3481 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3483 * doc/sourcebuild.texi (Endianness): New subsubsection.
3485 2018-05-23 Luis Machado <luis.machado@linaro.org>
3487 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3488 <prefetch_dynamic_strides>: New const bool field.
3489 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3490 prefetch_dynamic_strides.
3491 (exynosm1_prefetch_tune): Likewise.
3492 (thunderxt88_prefetch_tune): Likewise.
3493 (thunderx_prefetch_tune): Likewise.
3494 (thunderx2t99_prefetch_tune): Likewise.
3495 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3497 (aarch64_override_options_internal): Update to set
3498 PARAM_PREFETCH_DYNAMIC_STRIDES.
3499 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3500 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3501 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3502 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3503 prefetch-dynamic-strides setting.
3505 2018-05-23 Luis Machado <luis.machado@linaro.org>
3507 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3508 <minimum_stride>: New const int field.
3509 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3510 minimum_stride field defaulting to -1.
3511 (exynosm1_prefetch_tune): Likewise.
3512 (thunderxt88_prefetch_tune): Likewise.
3513 (thunderx_prefetch_tune): Likewise.
3514 (thunderx2t99_prefetch_tune): Likewise.
3515 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3516 <default_opt_level>: Set to 3.
3517 (aarch64_override_options_internal): Update to set
3518 PARAM_PREFETCH_MINIMUM_STRIDE.
3519 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3520 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3521 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3522 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3523 stride is constant and is below the minimum stride threshold.
3525 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3527 * config/arm/arm-cpus.in (mode26): Delete.
3528 (armv4): Delete mode26 reference.
3529 * config/arm/arm.c (arm_configure_build_target): Delete use of
3532 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3534 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3536 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3537 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3539 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3540 (floatunsdidf2): Ditto.
3542 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3544 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3545 (fixuns_trunc<mode>si2_avx512f): Ditto.
3546 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3547 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3548 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3550 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3552 PR rtl-optimization/79985
3553 * df-scan.c (df_insn_refs_collect): Remove special case for
3554 global registers and asm statements.
3556 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3558 * extend.texi (Global Register Variables): Rewrite the bullet list.
3559 Note that the register is available for allocation. Note that access
3560 via inline asm must use constraints. Add note about async-signal
3561 handlers. Remove paragraph about automagic register selection.
3563 2018-05-23 Richard Biener <rguenther@suse.de>
3565 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3566 of fixed offset from memset VN.
3568 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3570 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3572 (alloc_cand_and_find_basis): Initialize first_interp field.
3573 (slsr_process_mul): Modify first_interp field.
3574 (slsr_process_add): Likewise.
3575 (slsr_process_cast): Modify first_interp field for each new
3577 (slsr_process_copy): Likewise.
3578 (dump_candidate): Dump first_interp field.
3579 (replace_mult_candidate): Process all interpretations, not just
3581 (replace_rhs_if_not_dup): Likewise.
3582 (replace_one_candidate): Likewise.
3584 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3586 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3588 (aarch64_needs_frame_chain): New function.
3589 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3591 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3594 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3595 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3596 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3597 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3598 as true for strict-align.
3599 (aarch64_can_inline_p): Perform checks even when callee has no
3600 attributes to check for strict alignment.
3601 * doc/extend.texi (AArch64 Function Attributes): Document
3603 * doc/invoke.texi: (AArch64 Options): Likewise.
3605 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3607 PR tree-optimization/85853
3608 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3609 the handling of the root of the node to...
3610 (vect_slp_analyze_node_operations_1): ...this new function,
3611 and run the whole thing with the child nodes' def types
3612 set according to their SLP node's def type.
3614 2018-05-23 Richard Biener <rguenther@suse.de>
3617 * tree-data-ref.c (create_runtime_alias_checks): Defer
3618 and ignore overflow warnings.
3620 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3622 PR tree-optimization/85822
3623 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3626 2018-05-23 Richard Biener <rguenther@suse.de>
3628 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3629 memset constants via native_interpret_expr.
3631 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3634 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3636 (cgraph_node::create_alias): Likewise.
3637 (cgraph_node::get_availability): Check ifunc_resolver instead
3638 of looking up ifunc attribute.
3639 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3640 * varasm.c (do_assemble_alias): Likewise.
3641 (assemble_alias): Likewise.
3642 (default_binds_local_p_3): Likewise.
3643 * cgraph.h (cgraph_node): Add ifunc_resolver.
3644 (cgraph_node::only_called_directly_or_aliased_p): Return false
3646 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3648 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3649 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3650 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3651 instead of looking up ifunc attribute.
3653 2018-05-22 Luis Machado <luis.machado@linaro.org>
3655 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3657 2018-05-22 Martin Sebor <msebor@redhat.com>
3660 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3661 only when expasion succeeds.
3662 (expand_builtin_strcmp): Same.
3663 (expand_builtin_strncmp): Same.
3665 2018-05-22 Martin Sebor <msebor@redhat.com>
3667 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3669 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3670 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3672 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3673 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3675 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3676 Add check for legitimate_address.
3677 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3678 (aarch64_swap_ldrstr_operands): New.
3679 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3682 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3683 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3685 * config/aarch64/aarch64.md: New patterns to generate stp
3687 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3688 single words and double words.
3689 (load_pair_sw, load_pair_dw): Likewise.
3690 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3692 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3694 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3695 for different mode ldpstp and add peephole for merged zero stores.
3697 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3699 (aarch64_gen_store_pair): Rename calls to match new patterns.
3700 (aarch64_gen_load_pair): Rename calls to match new patterns.
3701 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3702 (load_pair<DREG:mode><DREG2:mode>): ... This.
3703 (store_pair<mode>): Rename to...
3704 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3705 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3707 (V_INT_EQUIV): Handle SImode.
3708 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3711 2018-05-22 Martin Sebor <msebor@redhat.com>
3714 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3715 or ajust the presumed bound on an operation to avoid unnecessary
3718 2018-05-22 Martin Sebor <msebor@redhat.com>
3720 PR tree-optimization/85826
3721 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3722 assuming that a DECL necesarily has a constant size.
3724 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3727 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3729 2018-05-22 Richard Biener <rguenther@suse.de>
3731 PR tree-optimization/85834
3732 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3733 non-constant and non-zero memset arguments.
3735 2018-05-22 Martin Liska <mliska@suse.cz>
3738 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3740 2018-05-22 Richard Biener <rguenther@suse.de>
3742 PR tree-optimization/85863
3743 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3744 comparisons when vectype is specified.
3745 (vectorizable_condition): Do not specify vectype for
3746 vect_is_simple_cond when SLP vectorizing.
3748 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3751 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3752 define __ibm128 as long double.
3753 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3754 as a distinct type when IEEE 128-bit support is enabled.
3755 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3756 128-bit types to use the correct functions.
3757 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3758 convert between 128-bit floating point types that have different
3759 modes but the same representation, instead of using gen_lowpart to
3761 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3763 (IFKF_reg): New attributes to give the register constraints for
3765 (extend<mode>tf2_internal): New insns to mark an explicit
3766 conversion between 128-bit floating point types that have a
3767 different mode but share the same representation.
3769 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3771 PR tree-optimization/85814
3772 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3773 a null return from get_strinfo when unsharing the next
3774 strinfo in the chain.
3776 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3779 * varasm.c (weak_finish): Clean up weak_decls.
3781 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3783 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3784 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3785 UNSPEC_UADALP values.
3786 * config/aarch64/iterators.md (ABAL): New int iterator.
3789 (sur): Add mappings for the above.
3790 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3792 (aarch64_<sur>abal<mode>_4): Likewise.
3793 (aarch64_<sur>adalp<mode>_3): Likewise.
3794 (<sur>sadv16qi): New define_expand.
3796 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3798 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3799 (*movdf_internal): Ditto.
3800 (*rcpsf2_sse): Ditto.
3801 (*rsqrtsf2_sse): Ditto.
3802 (*sqrt<mode>2_sse): Ditto.
3804 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3806 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3808 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3809 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3810 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3811 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3813 * config/aarch64/arm_neon.h: Likewise.
3814 * config/aarch64/iterators.md (VQ_I): New.
3816 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3818 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3820 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3823 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3825 * config/nds32/constraints.md (S): New constraint.
3826 * config/nds32/nds32.md (call_internal): Use constraint S.
3827 (call_value_internal): Likewise.
3828 (sibcall_internal): Likewise.
3829 (sibcall_value_internal): Likewise.
3831 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3832 Chung-Ju Wu <jasonwucj@gmail.com>
3834 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3837 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3838 Chung-Ju Wu <jasonwucj@gmail.com>
3840 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3841 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3842 (nds32_rtx_costs_impl): Simplify.
3843 (nds32_address_cost_impl): Simplify.
3844 (nds32_init_rtx_costs): New function.
3845 (nds32_rtx_costs_speed_prefer): Likewise.
3846 (nds32_rtx_costs_size_prefer): Likewise.
3847 (nds32_address_cost_speed_prefer): Likewise.
3848 (nds32_address_cost_speed_fwprop): Likewise.
3849 (nds32_address_cost_size_prefer): Likewise.
3850 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3851 * config/nds32/nds32.c (nds32_option_override): Use
3852 nds32_init_rtx_costs function.
3854 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3856 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3857 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3858 (TARGET_PIPELINE_N8): Likewise.
3859 (TARGET_PIPELINE_N10): Likewise.
3860 (TARGET_PIPELINE_N13): Likewise.
3861 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3863 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3865 * config/nds32/nds32-fpu.md: Update copyright year.
3867 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3869 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3871 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3873 * config/nds32/nds32.c
3874 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3875 * config/nds32/nds32.opt (minline-asm-r15): New option.
3877 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3879 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3881 * config/nds32/nds32.md (abssi2): New pattern.
3883 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3885 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3886 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3887 Merge insn pattern from sse_cvtsi2ss<round_name> and
3888 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3889 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3890 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3891 using SWI48 mode iterator.
3892 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3893 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3894 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3895 pattern from sse_cvttss2si<round_saeonly_name>
3896 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3897 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3898 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3899 using SWI48 mode iterator.
3900 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3901 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3902 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3903 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3904 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3905 using SWI48 mode iterator.
3906 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3907 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3908 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3909 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3910 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3911 SWI48 mode iterator.
3912 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3913 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3914 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3915 pattern from sse_cvttsd2si<round_saeonly_name>
3916 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3918 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3920 * config/nds32/nds32-md-auxiliary.c
3921 (nds32_valid_smw_lwm_base_p): Refine.
3922 (nds32_output_smw_single_word): Refine.
3923 (nds32_output_smw_double_word): New.
3924 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3926 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3928 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3929 (nds32_output_stack_pop): Refine.
3930 (nds32_expand_unaligned_load): Refine.
3931 (nds32_expand_unaligned_store): Refine.
3933 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3934 Chung-Ju Wu <jasonwucj@gmail.com>
3936 * config/nds32/constants.md: Add TP_REGNUM constant.
3937 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3938 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3940 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3941 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3942 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3943 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3944 and PIC code generation.
3945 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3947 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3949 * config/nds32/nds32.md: Support TLS and PIC.
3950 * config/nds32/nds32.c: Support TLS and PIC.
3951 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3952 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3955 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3957 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3958 mode with E_ prefix.
3960 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3961 Chung-Ju Wu <jasonwucj@gmail.com>
3963 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3964 * config/nds32/nds32-md-auxiliary.c
3965 (symbolic_reference_mentioned_p): New.
3966 (nds32_legitimize_ict_address): New.
3967 (nds32_expand_ict_move): New.
3968 (nds32_indirect_call_referenced_p): New.
3969 (nds32_symbol_binds_local_p): Delete.
3970 (nds32_long_call_p): Modify.
3971 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3972 * config/nds32/nds32-protos.h
3973 (symbolic_reference_mentioned_p): Declare.
3974 (nds32_legitimize_ict_address): Declare.
3975 (nds32_expand_ict_move): Declare.
3976 (nds32_indirect_call_referenced_p): Declare.
3977 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3978 (nds32_relax_group): Use nds32_ict_const_p as condition.
3979 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3980 (nds32_asm_file_start): Output ict_model directive in asm code.
3981 (nds32_legitimate_address_p): Consider indirect call.
3982 (nds32_print_operand): Consider indirect call.
3983 (nds32_print_operand_address): Consider indirect call.
3984 (nds32_insert_attributes): Handle "indirect_call" attribute.
3985 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3986 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3987 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3988 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3989 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3990 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3991 (TARGET_ICT_MODEL_SMALL): Define.
3992 (TARGET_ICT_MODEL_LARGE): Define.
3993 * config/nds32/nds32.md (movsi): Consider ict model.
3994 (call, call_value): Consider ict model.
3995 (sibcall, sibcall_value): Consider ict model.
3996 * config/nds32/nds32.opt (mict-model): New option.
3997 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4000 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4001 Monk Chiang <sh.chiang04@gmail.com>
4002 Jim Wilson <jimw@sifive.com>
4004 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4005 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4006 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4007 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4008 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4009 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4010 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4011 compute save_libcall_adjustment properly.
4012 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4013 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4014 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4015 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4016 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4017 (ABI_SPEC): Handle mabi=ilp32e.
4018 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4019 (RVE): Add RVE mask.
4020 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4021 <-march>: Add rv32e as an example.
4023 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4026 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4027 (intra_create_variable_infos): Handle C++ constructors.
4029 2018-05-18 Martin Liska <mliska@suse.cz>
4031 * passes.def: Remove a redundant pass.
4033 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4036 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4038 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4040 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4042 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4044 (armv2, armv2a, armv3, armv3m): Delete architectures.
4045 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4046 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4047 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4049 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4050 (*mulsidi3adddi): Likewise.
4051 (mulsidi3): Likewise.
4052 (*mulsidi3_nov6): Likewise.
4053 (umulsidi3): Likewise.
4054 (umulsidi3_nov6): Likewise.
4055 (umaddsidi4): Likewise.
4056 (*umulsidi3adddi): Likewise.
4057 (smulsi3_highpart): Likewise.
4058 (*smulsi3_highpart_nov6): Likewise.
4059 (umulsi3_highpart): Likewise.
4060 (*umulsi3_highpart_nov6): Likewise.
4061 * config/arm/arm.h (arm_arch3m): Delete.
4062 * config/arm/arm.c (arm_arch3m): Delete.
4063 (arm_option_override_internal): Update armv3-related comment.
4064 (arm_configure_build_target): Delete use of isa_bit_mode32.
4065 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4066 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4067 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4069 (mulusa3): Likewise.
4070 * config/arm/arm-protos.h (arm_arch3m): Delete.
4071 * config/arm/arm-tables.opt: Regenerate.
4072 * config/arm/arm-tune.md: Likewise.
4073 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4074 deleted architectures.
4076 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4078 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4079 (armv5t, armv5te): New features.
4080 (ARMv5, ARMv5e): Delete fgroups.
4081 (ARMv5t, ARMv5te): Adjust for above changes.
4083 (armv5, armv5e): Delete arches.
4084 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4086 (*call_reg_arm): Likewise.
4087 (*call_value_reg_armv5): Likewise.
4088 (*call_value_reg_arm): Likewise.
4089 (*call_symbol): Likewise.
4090 (*call_value_symbol): Likewise.
4091 (*sibcall_insn): Likewise.
4092 (*sibcall_value_insn): Likewise.
4094 (prefetch): Likewise.
4095 (define_split and define_peephole2 dependent on arm_arch5):
4097 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4099 (TARGET_ARM_QBIT): Likewise.
4100 (TARGET_DSP_MULTIPLY): Likewise.
4101 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4102 (arm_arch5, arm_arch5e): Delete.
4103 (arm_arch5t, arm_arch5te): Declare.
4104 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4105 (arm_arch5t): Declare.
4106 (arm_option_reconfigure_globals): Update for the above.
4107 (arm_options_perform_arch_sanity_checks): Update comment, replace
4108 use of arm_arch5 with arm_arch5t.
4109 (use_return_insn): Likewise.
4110 (arm_emit_call_insn): Likewise.
4111 (output_return_instruction): Likewise.
4112 (arm_final_prescan_insn): Likewise.
4113 (arm_coproc_builtin_available): Likewise.
4114 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4115 arm_arch5e with arm_arch5t and arm_arch5te.
4116 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4117 (arm_arch5t, arm_arch5te): Declare.
4118 * config/arm/arm-tables.opt: Regenerate.
4119 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4120 * config/arm/t-multilib: Likewise.
4121 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4122 instead of arm_arch5.
4123 (*call_reg_thumb1): Likewise.
4124 (*call_value_reg_thumb1_v5): Likewise.
4125 (*call_value_reg_thumb1): Likewise.
4126 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4128 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4130 2018-05-18 Martin Liska <mliska@suse.cz>
4132 PR gcov-profile/84846
4133 * doc/gcov.texi: Document -t option of gcov tool.
4135 2018-05-18 Martin Liska <mliska@suse.cz>
4137 PR gcov-profile/84846
4138 * gcov.c (print_usage): Add new -t option.
4139 (process_args): Handle the option.
4140 (generate_results): Use stdout as output when requested by
4143 2018-05-18 Martin Liska <mliska@suse.cz>
4145 PR gcov-profile/84846
4146 * coverage.c (coverage_init): Write PWD to .gcno file.
4147 * doc/gcov.texi: Document how working directory is printed.
4148 * gcov-dump.c (dump_gcov_file): Print PWD.
4149 * gcov.c (output_intermediate_file): Likewise.
4150 (read_graph_file): Read PWD string.
4151 (output_lines): Print PWD.
4153 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4156 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4157 for retval and return false if all args to phi are zero.
4159 2018-05-18 Richard Biener <rguenther@suse.de>
4161 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4163 (evrp_dom_walker::before_dom_children): Call it.
4165 2018-05-18 Richard Biener <rguenther@suse.de>
4167 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4168 results when processing array refs with variable index.
4170 2018-05-18 Toon Moene <toon@moene.org>
4172 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4173 directly after that of -floop-interchange. Indicate that both
4174 options are enabled by default when specifying -O3.
4176 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4178 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4179 iterator. Delete separate integer-mode vec_set<mode> expander.
4180 (aarch64_simd_vec_setv2di): Delete.
4181 (vec_setv2di): Delete.
4182 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4183 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4184 the "w, r" alternative.
4186 2018-05-18 Martin Liska <mliska@suse.cz>
4188 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4189 * tree-pass.h (make_pass_lower_switch_O0): New function.
4190 * tree-switch-conversion.c (node_has_low_bound): Remove.
4191 (node_has_high_bound): Likewise.
4192 (node_is_bounded): Likewise.
4193 (class pass_lower_switch): Make it a template type and create
4195 (pass_lower_switch::execute): Add template argument.
4196 (make_pass_lower_switch): New function.
4197 (make_pass_lower_switch_O0): New function.
4198 (do_jump_if_equal): Remove.
4199 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4200 all the hard work to tree optimization passes.
4202 2018-05-18 Martin Liska <mliska@suse.cz>
4204 * dbgcnt.c (limit_low): Renamed from limit.
4205 (limit_high): New variable.
4206 (dbg_cnt_is_enabled): Check for upper limit.
4207 (dbg_cnt): Adjust dumping.
4208 (dbg_cnt_set_limit_by_index): Add new argument for high
4210 (dbg_cnt_set_limit_by_name): Likewise.
4211 (dbg_cnt_process_single_pair): Parse new format.
4212 (dbg_cnt_process_opt): Use strtok.
4213 (dbg_cnt_list_all_counters): Remove 'value' and add
4215 * doc/invoke.texi: Document changes.
4217 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4219 * doc/sourcebuild.texi (scalar_all_fma): Document.
4220 * tree.def (FMA_EXPR): Delete.
4221 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4222 * internal-fn.c (ternary_direct): New macro.
4223 (expand_ternary_optab_fn): Likewise.
4224 (direct_ternary_optab_supported_p): Likewise.
4225 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4226 * builtins.c (fold_builtin_fma): Delete.
4227 (fold_builtin_3): Don't call it.
4228 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4229 * expr.c (expand_expr_real_2): Likewise.
4230 * fold-const.c (operand_equal_p): Likewise.
4231 (fold_ternary_loc): Likewise.
4232 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4233 * gimple.c (DEFTREECODE): Likewise.
4234 * gimplify.c (gimplify_expr): Likewise.
4235 * optabs-tree.c (optab_for_tree_code): Likewise.
4236 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4237 * tree-eh.c (operation_could_trap_p): Likewise.
4238 (stmt_could_throw_1_p): Likewise.
4239 * tree-inline.c (estimate_operator_cost): Likewise.
4240 * tree-pretty-print.c (dump_generic_node): Likewise.
4241 (op_code_prio): Likewise.
4242 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4243 * tree-ssa-operands.c (get_expr_operands): Likewise.
4244 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4245 * fold-const-call.h (fold_fma): Delete.
4246 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4247 CFN_FNMA and CFN_FNMS.
4249 * genmatch.c (combined_fn): New enum.
4250 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4251 (commutative_op): New function.
4252 (commutate): Use it. Handle more than 2 operands.
4253 (dt_operand::gen_gimple_expr): Use commutative_op.
4254 (parser::parse_expr): Allow :c to be used with non-binary
4255 operators if the commutative operand is known.
4256 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4257 CFN_FMS, CFN_FNMA and CFN_FNMS.
4258 (backprop::process_assign_use): Remove FMA_EXPR handling.
4259 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4260 (gen_hsa_fma): New function.
4261 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4262 IFN_FNMA and IFN_FNMS.
4263 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4264 * gimple-fold.h (follow_all_ssa_edges): Declare.
4265 * gimple-fold.c (follow_all_ssa_edges): New function.
4266 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4267 gimple_build interface and use follow_all_ssa_edges to fold the result.
4268 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4269 instead of checking for optabs directly.
4270 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4271 rather than FMA_EXPRs.
4272 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4273 call to IFN_FMA instead of an FMA_EXPR.
4275 2018-05-17 Jim Wilson <jimw@sifive.com>
4277 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4278 sign extended promoted subreg, and the range does not have the sign bit
4279 set, then do a sign extend.
4281 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4282 test, check for sign extended subreg and/or constant operands, and
4283 do a sign extend in that case.
4285 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4287 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4288 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4290 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4291 Change logics_shift_reg to logics_shift_imm.
4292 (thunderx2t99_fp_loadpair_basic): Delete.
4293 (thunderx2t99_fp_storepair_basic): Delete.
4294 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4295 (thunderx2t99_asimd_polynomial): Delete.
4296 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4297 and neon_fp_mul_d_scalar_q.
4298 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4299 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4300 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4301 (thunderx2t99_asimd_lut): Add missing tbl types.
4302 (thunderx2t99_asimd_ext): Delete.
4303 (thunderx2t99_asimd_load1_1_mult): Delete.
4304 (thunderx2t99_asimd_load1_2_mult): Delete.
4305 (thunderx2t99_asimd_load1_ldp): New.
4306 (thunderx2t99_asimd_load1): New.
4307 (thunderx2t99_asimd_load2): Add missing *load2* types.
4308 (thunderx2t99_asimd_load3): New.
4309 (thunderx2t99_asimd_load4): New.
4310 (thunderx2t99_asimd_store1_1_mult): Delete.
4311 (thunderx2t99_asimd_store1_2_mult): Delete.
4312 (thunderx2t99_asimd_store2_mult): Delete.
4313 (thunderx2t99_asimd_store2_onelane): Delete.
4314 (thunderx2t99_asimd_store_stp): New.
4315 (thunderx2t99_asimd_store1): New.
4316 (thunderx2t99_asimd_store2): New.
4317 (thunderx2t99_asimd_store3): New.
4318 (thunderx2t99_asimd_store4): New.
4320 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4322 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4323 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4325 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4326 Segher Boessenkool <segher@kernel.crashing.org>
4329 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4332 2018-05-17 Richard Biener <rguenther@suse.de>
4334 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4335 for pruning loop and prune defs feeding only already visited PHIs.
4337 2018-05-17 Richard Biener <rguenther@suse.de>
4339 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4341 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4342 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/85793
4345 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4346 for VMAT_ELEMENTWISE.
4348 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4350 * internal-fn.h (lookup_internal_fn): Declare
4351 * internal-fn.c (lookup_internal_fn): New function.
4352 * gimple.c (gimple_build_call_from_tree): Handle calls to
4354 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4355 internal function names.
4356 * tree-pretty-print.c (dump_generic_node): Likewise.
4357 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4359 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4361 * gimple-fold.h (gimple_build): Make the function forms take
4362 combined_fn rather than built_in_function.
4363 (gimple_simplify): Likewise.
4364 * gimple-match-head.c (gimple_simplify): Likewise.
4365 * gimple-fold.c (gimple_build): Likewise.
4366 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4367 rather than gimple_build_call_internal.
4368 (get_initial_defs_for_reduction): Likewise.
4369 (vect_create_epilog_for_reduction): Likewise.
4370 (vectorizable_live_operation): Likewise.
4372 2018-05-17 Martin Liska <mliska@suse.cz>
4374 * gimple-ssa-sprintf.c (format_directive): Do not use
4375 space in between 'G_' and '('.
4377 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4380 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4381 even if the mask is not all ones.
4384 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4386 (ix86_gimple_fold_builtin): Likewise.
4389 * config/i386/i386.c: Include tree-vector-builder.h.
4390 (ix86_vector_shift_count): New function.
4391 (ix86_fold_builtin): Fold shift builtins by scalar count.
4392 (ix86_gimple_fold_builtin): Likewise.
4394 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4395 _mm512_setzero): New intrinsics.
4397 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4398 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4400 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4401 code generation for cases where splatting a value is not useful.
4402 * simplify-rtx.c (simplify_ternary_operation): Simplify
4403 vec_merge across a vec_duplicate and a paradoxical subreg forming
4404 a vector mode to a vec_concat.
4406 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4408 * config.gcc: Support "goldmont-plus".
4409 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4411 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4412 PROCESSOR_GOLDMONT_PLUS.
4413 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4414 (processor_target_table): Add "goldmont-plus".
4415 (PTA_GOLDMONT_PLUS): Define.
4416 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4417 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4418 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4419 (fold_builtin_cpu): Add "goldmont-plus".
4420 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4421 (ix86_option_override_internal): Add "goldmont-plus".
4422 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4423 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4424 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4425 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4427 2018-05-17 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/85757
4430 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4431 remove defs that only feed that PHI from further processing.
4433 2018-05-16 Jim Wilson <jimw@sifive.com>
4435 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4437 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4439 2018-05-16 Mark Wielaard <mark@klomp.org>
4441 * dwarf2out.c (count_index_strings): New function.
4442 (output_indirect_strings): Call count_index_strings and generate
4443 header for dwarf_version >= 5.
4445 2018-05-16 Mark Wielaard <mark@klomp.org>
4447 * dwarf2out.c (dwarf_FORM): New function.
4448 (set_indirect_string): Use dwarf_FORM.
4449 (reset_indirect_string): Likewise.
4450 (size_of_die): Likewise.
4451 (value_format): Likewise.
4452 (output_die): Likewise.
4453 (add_skeleton_AT_string): Likewise.
4454 (output_macinfo_op): Likewise.
4455 (index_string): Likewise.
4456 (output_index_string_offset): Likewise.
4457 (output_index_string): Likewise.
4458 (count_index_strings): Likewise.
4460 2018-05-16 Carl Love <cel@us.ibm.com>
4462 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4463 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4465 2018-05-16 Martin Jambor <mjambor@suse.cz>
4467 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4468 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4470 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4472 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4473 (fnma<mode>4): Likewise.
4474 (fms<mode>4): Likewise.
4475 (fnms<mode>4): Likewise.
4476 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4477 (aarch64_fnma<mode>4): Likewise.
4478 (aarch64_fms<mode>4): Likewise.
4479 (aarch64_fnms<mode>4): Likewise.
4480 (aarch64_fnmadd<mode>4): Likewise.
4482 2018-05-16 Jason Merrill <jason@redhat.com>
4484 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4486 2018-05-16 Richard Biener <rguenther@suse.de>
4488 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4489 (dump_stmt_cost): Declare.
4490 (add_stmt_cost): Dump cost we add.
4491 (add_stmt_costs): New function.
4492 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4494 (vect_analyze_stmt): Adjust prototype.
4495 (vectorizable_condition): Likewise.
4496 (vectorizable_live_operation): Likewise.
4497 (vectorizable_reduction): Likewise.
4498 (vectorizable_induction): Likewise.
4499 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4500 cost vector to pass to vectorizable_ and record afterwards.
4501 (vect_model_reduction_cost): Take cost vector argument and adjust.
4502 (vect_model_induction_cost): Likewise.
4503 (vectorizable_reduction): Likewise.
4504 (vectorizable_induction): Likewise.
4505 (vectorizable_live_operation): Likewise.
4506 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4507 SLP_TREE_NUMBER_OF_VEC_STMTS.
4508 (vect_analyze_slp_cost_1): Remove.
4509 (vect_analyze_slp_cost): Likewise.
4510 (vect_slp_analyze_node_operations): Take visited args and
4511 a target cost vector. Avoid processing already visited stmt sets.
4512 (vect_slp_analyze_operations): Use a local cost vector to gather
4513 costs and register those of non-discarded instances.
4514 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4515 (vect_schedule_slp_instance): Remove copying of
4516 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4518 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4519 adding cost. Record cost entry location.
4520 (vect_prologue_cost_for_slp_op): Function to compute cost of
4521 a constant or invariant generated for SLP vect in the prologue,
4522 split out from vect_analyze_slp_cost_1.
4523 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4524 (vect_model_promotion_demotion_cost): Likewise.
4525 (vect_model_store_cost): Likewise, make static.
4526 (vect_model_load_cost): Likewise.
4527 (vectorizable_bswap): Add cost vector arg and adjust.
4528 (vectorizable_call): Likewise.
4529 (vectorizable_simd_clone_call): Likewise.
4530 (vectorizable_conversion): Likewise.
4531 (vectorizable_assignment): Likewise.
4532 (vectorizable_shift): Likewise.
4533 (vectorizable_operation): Likewise.
4534 (vectorizable_store): Likewise.
4535 (vectorizable_load): Likewise.
4536 (vectorizable_condition): Likewise.
4537 (vectorizable_comparison): Likewise.
4538 (can_vectorize_live_stmts): Likewise.
4539 (vect_analyze_stmt): Likewise.
4540 (vect_transform_stmt): Adjust calls to vectorizable_*.
4541 * tree-vectorizer.c: Include gimple-pretty-print.h.
4542 (dump_stmt_cost): New function.
4544 2018-05-16 Richard Biener <rguenther@suse.de>
4546 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4547 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4548 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4549 (check_name): New callback.
4550 (dse_classify_store): Track cycles via a visited bitmap of PHI
4551 defs and simplify handling of in-loop and across loop dead stores
4552 and properly fail for loop-variant refs. Handle byte-tracking with
4553 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4556 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4558 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4559 (vect_get_mask_type_for_stmt): Likewise.
4560 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4562 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4563 to determine the statement's vector type and the vector type that
4564 should be used for calculating nunits. Deal with cases in which
4565 the type has to be deferred.
4566 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4567 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4568 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4569 (vect_determine_vf_for_stmt): New functions, split out from...
4570 (vect_determine_vectorization_factor): ...here.
4571 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4572 (vect_get_mask_type_for_stmt): New functions, split out from
4573 vect_determine_vectorization_factor.
4575 2018-05-16 Richard Biener <rguenther@suse.de>
4577 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4578 verify the [VEC_]COND_EXPR embedded comparison.
4580 2018-05-15 Martin Sebor <msebor@redhat.com>
4582 PR tree-optimization/85753
4583 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4584 RECORD_TYPE in addition to ARRAY_TYPE.
4586 2018-05-15 Martin Sebor <msebor@redhat.com>
4589 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4591 2018-05-15 Richard Biener <rguenther@suse.de>
4593 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4594 add by_clobber_p one. Change algorithm to collect all defs
4595 representing uses we need to walk and try reducing them to
4596 a single one before failing.
4597 (dse_dom_walker::dse_optimize_stmt): Adjust.
4599 2018-05-13 Mark Wielaard <mark@klomp.org>
4601 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4602 (size_of_loc_descr): Likewise.
4603 (output_loc_operands): Likewise.
4604 (output_loc_operands_raw): Likewise.
4605 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4606 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4607 (hash_loc_operands): Likewise.
4608 (compare_loc_operands): Likewise.
4610 2018-05-14 Mark Wielaard <mark@klomp.org>
4612 * dwarf2out.c (count_index_addrs): New function.
4613 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4615 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4617 PR tree-optimization/83648
4618 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4619 return value as malloc candidate.
4621 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4624 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4625 param as true in call to suggest_attribute.
4627 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4629 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4630 -mreadonly-in-sdata.
4632 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4634 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4636 (aarch64_crypto_aesd_fused): Likewise.
4638 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4640 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4641 (movsi_aarch64): Likewise.
4642 (load_pairsi): Likewise.
4643 (load_pairdi): Likewise.
4644 (store_pairsi): Likewise.
4645 (store_pairdi): Likewise.
4646 (load_pairsf): Likewise.
4647 (load_pairdf): Likewise.
4648 (store_pairsf): Likewise.
4649 (store_pairdf): Likewise.
4650 (zero_extend): Likewise.
4651 (trunc): Swap alternatives.
4652 (fcvt_target): Add '?' to prefer w over r.
4654 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4657 * config/i386/i386.md: Disallow non-commutative arithmetics in
4658 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4659 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4660 in the peephole2 before it.
4662 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4664 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4665 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4666 (ix86_handle_option): Handle -mcldemote.
4667 * config.gcc: New header.
4668 * config/i386/cldemoteintrin.h: New file.
4669 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4670 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4672 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4673 OPTION_MASK_ISA_CLDEMOTE.
4674 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4675 (ix86_valid_target_attribute_inner_p): Ditto.
4676 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4677 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4678 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4679 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4680 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4682 * config/i386/i386.opt: Add -mcldemote.
4683 * config/i386/x86intrin.h: New header.
4684 * doc/invoke.texi: Add -mcldemote.
4686 2018-05-14 Richard Biener <rguenther@suse.de>
4688 * doc/match-and-simplify.texi: Adjust :s documentation.
4690 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4692 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4693 intended memcpy size.
4694 (REORDER_45): Likewise.
4696 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4698 * sort.cc: New file.
4699 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4700 * vec.c (qsort_chk): Use gcc_qsort.
4701 * Makefile.in (OBJS-libcommon): Add sort.o.
4702 (build/sort.o): New target. Use it...
4703 (BUILD_RTL): ... here, and...
4704 (build/gencfn-macros): ... here, and...
4705 (build/genmatch): ... here.
4707 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4708 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4711 * config/nds32/nds32-graywolf.md: New file.
4712 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4713 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4715 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4716 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4717 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4718 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4719 * config/nds32/pipelines.md: Include n15 settings.
4721 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4722 Chung-Ju Wu <jasonwucj@gmail.com>
4724 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4725 * config/nds32/nds32-n13.md: New file.
4726 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4727 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4729 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4730 * config/nds32/nds32.md (pipeline_model): Add n13.
4731 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4732 * config/nds32/pipelines.md: Include n13 settings.
4734 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4735 Chung-Ju Wu <jasonwucj@gmail.com>
4737 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4738 * config/nds32/nds32-n10.md: New file.
4739 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4740 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4742 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4743 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4744 * config/nds32/nds32.md (pipeline_model): Add n10.
4745 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4746 * config/nds32/pipelines.md: Include n10 settings.
4748 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4749 Kito Cheng <kito.cheng@gmail.com>
4750 Chung-Ju Wu <jasonwucj@gmail.com>
4752 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4753 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4754 Add enum values for DSP extension instructions.
4755 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4757 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4758 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4760 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4761 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4762 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4763 * config/nds32/nds32-intrinsic.md: Likewise.
4764 * config/nds32/nds32_intrinsic.h: Likewise.
4765 * config/nds32/nds32-md-auxiliary.c: Likewise.
4766 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4767 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4768 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4769 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4770 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4771 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4772 TYPE_DMAC in switch statement.
4773 * config/nds32/nds32.c: New checking and implementation for DSP
4774 extension instructions.
4775 * config/nds32/nds32.h: Likewise.
4776 * config/nds32/nds32.md: Likewise.
4777 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4778 * config/nds32/predicates.md: Implement new predicates for DSP
4781 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4783 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4784 Reformat alternatives and attributes so it is easier to identify
4785 which constraints/attributes go with which instruction.
4786 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4787 (mov<mode>_softfloat32, FMOVE64): Likewise.
4788 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4789 (mov<mode>_softfloat64, FMOVE64): Likewise.
4791 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4793 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4795 (Basic PowerPC Built-in Functions): The new name of the
4796 subsection previously known as "PowerPC Built-in Functions".
4797 (Basic PowerPC Built-in Functions Available on all Configurations):
4799 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4800 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4801 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4802 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4804 2018-05-11 Martin Jambor <mjambor@suse.cz>
4807 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4810 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4813 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4815 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4817 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4818 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4819 (ix86_handle_option): Handle -mwaitpkg.
4820 * config.gcc: New header.
4821 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4822 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4823 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4825 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4826 OPTION_MASK_ISA_WAITPKG.
4827 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4828 (ix86_option_override_internal): Add PTA_WAITPKG.
4829 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4830 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4831 IX86_BUILTIN_TPAUSE.
4832 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4833 __builtin_ia32_umwait and __builtin_ia32_tpause.
4834 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4835 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4836 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4837 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4838 UNSPECV_TPAUSE): New.
4839 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4840 * config/i386/i386.opt: Add -mwaitpkg.
4841 * config/i386/waitpkgintrin.h: New file.
4842 * config/i386/x86intrin.h: New header.
4843 * doc/invoke.texi: Add -mwaitpkg.
4845 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4848 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4850 (cortex-m0): Use armv6s-m isa.
4851 (cortex-m0plus): Likewise.
4852 (cortex-m1): Likewise.
4853 (cortex-m0.small-multiply): Likewise.
4854 (cortex-m0plus.small-multiply): Likewise.
4855 (cortex-m1.small-multiply): Likewise.
4857 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4858 Jakub Jelinek <jakub@redhat.com>
4860 PR tree-optimization/85692
4861 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4862 source permute as well.
4864 2018-05-11 Martin Liska <mliska@suse.cz>
4867 * doc/extend.texi: Document LLVM style format for no_sanitize
4870 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4872 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4873 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4874 (mode_supports_vsx_dform_quad): Likewise.
4875 (mode_supports_vmx_dform): Move these functions to be next to the
4876 other mode_supports functions.
4877 (mode_supports_dq_form): Likewise.
4878 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4879 mode_supports_dq_form.
4880 (reg_offset_addressing_ok_p): Likewise.
4881 (offsettable_ok_by_alignment): Likewise.
4882 (rs6000_legitimate_offset_address_p): Likewise.
4883 (legitimate_lo_sum_address_p): Likewise.
4884 (rs6000_legitimize_address): Likewise.
4885 (rs6000_legitimize_reload_address): Likewise.
4886 (rs6000_secondary_reload_inner): Likewise.
4887 (rs6000_preferred_reload_class): Likewise.
4888 (rs6000_output_move_128bit): Likewise.
4890 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4892 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4893 Generate SImode target register for null target.
4894 <case IX86_BUILTIN_XGETBV>: Ditto.
4895 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4896 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4898 2018-05-10 Carl Love <cel@us.ibm.com>
4900 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4901 dcbtt and dcbtstt if operands[2] is 0.
4903 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4906 * config/i386/sse.md (usadv64qi): New expander.
4908 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4910 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4911 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4912 -maltivec=be support.
4913 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4914 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4915 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4916 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4917 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4918 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4919 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4920 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4921 altivec_vsumsws): Adjust.
4922 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4923 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4924 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4926 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4927 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4928 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4929 (altivec_lve<VI_char>x): Delete expand.
4930 (*altivec_lve<VI_char>x_internal): Rename to...
4931 (altivec_lve<VI_char>x): ... this.
4932 (altivec_lvxl_<mode>): Delete expand.
4933 (*altivec_lvxl_<mode>_internal): Rename to ...
4934 (altivec_lvxl_<mode>): ... this.
4935 (altivec_stvxl_<mode>): Delete expand.
4936 (*altivec_stvxl_<mode>_internal): Rename to ...
4937 (altivec_stvxl_<mode>): ... this.
4938 (altivec_stve<VI_char>x): Delete expand.
4939 (*altivec_stve<VI_char>x_internal): Rename to ...
4940 (altivec_stve<VI_char>x): ... this.
4941 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4942 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4943 reduc_plus_scal_<mode>): Adjust.
4944 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4946 (rs6000_cpu_cpp_builtins): Adjust.
4947 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4948 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4949 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4950 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4951 -maltivec=be support.
4952 (rs6000_split_vec_extract_var): Adjust.
4953 (rs6000_split_v4si_init): Adjust.
4954 (swap_selector_for_mode): Delete.
4955 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4956 altivec_expand_stvex_be): Delete.
4957 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4958 -maltivec=be support.
4959 (rs6000_gimple_fold_builtin): Ditto.
4960 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4962 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4963 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4964 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4965 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4966 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4967 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4968 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4969 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4970 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4971 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4972 anonymous split): Adjust.
4973 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4974 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4976 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4978 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4979 when --with-gxx-include-dir is also specified.
4980 * configure: Regenerate.
4982 2018-05-09 Jim Wilson <jimw@sifive.com>
4985 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4986 * config/riscv/t-withmultilib: New.
4987 * config/riscv/withmultilib.h: New.
4988 * doc/install.texi: Document RISC-V --with-multilib-list support.
4990 2018-05-09 Richard Biener <rguenther@suse.de>
4992 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4994 (vect_bb_vectorization_profitable_p): Adjust. Compute
4995 actual scalar cost using the cost vector and the add_stmt_cost
4998 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5000 PR rtl-optimization/85645
5001 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5002 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5004 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5006 PR rtl-optimization/85645
5007 * shrink-wrap.c (spread_components): Return a boolean saying if
5008 anything was changed.
5009 (try_shrink_wrapping_separate): Iterate spread_components until
5010 nothing changes anymore.
5012 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5014 PR rtl-optimization/85645
5015 * regrename.c (build_def_use): Also kill the chains that include the
5016 destination of a REG_CFA_REGISTER note.
5018 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5020 PR rtl-optimization/85645
5021 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5022 insn that has a REG_CFA_REGISTER note.
5024 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5026 * cfgexpand.c (expand_clobber): New function.
5027 (expand_gimple_stmt_1): Use it.
5028 * tree-vect-stmts.c (vect_clobber_variable): New function,
5030 (vectorizable_simd_clone_call): ...here.
5031 (vectorizable_store): Emit a clobber either side of an
5032 IFN_STORE_LANES sequence.
5033 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5035 2018-05-09 Tom de Vries <tom@codesourcery.com>
5038 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5039 (define_insn "trap_if_false"): Add exit after trap.
5041 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5043 PR rtl-optimization/85638
5044 * bb-reorder.c: Include common/common-target.h.
5045 (create_forwarder_block): New function extracted from...
5046 (fix_up_crossing_landing_pad): ...here. Rename into...
5047 (dw2_fix_up_crossing_landing_pad): ...this.
5048 (sjlj_fix_up_crossing_landing_pad): New function.
5049 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5050 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5051 from both partitions and exit the loop after one iteration.
5053 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5056 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5058 (Basic PowerPC Built-in Functions): The new name of the
5059 subsection previously known as "PowerPC Built-in Functions".
5060 (Basic PowerPC Built-in Functions Available on all Configurations):
5062 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5064 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5066 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5068 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5071 2018-05-08 Jim Wilson <jimw@sifive.com>
5073 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5074 (LD_EMUL_SUFFIX): New.
5075 (LINK_SPEC): Use it.
5077 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5079 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5081 (Basic PowerPC Built-in Functions): The new name of the
5082 subsection previously known as "PowerPC Built-in Functions".
5083 (Basic PowerPC Built-in Functions Available on all Configurations):
5085 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5087 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5089 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5091 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5094 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5097 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5098 after cmpelim optimization.
5100 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5102 * config.gcc: Support "goldmont".
5103 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5104 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5106 * config/i386/i386.c (m_GOLDMONT): Define.
5107 (processor_target_table): Add "goldmont".
5108 (PTA_GOLDMONT): Define.
5109 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5110 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5111 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5112 (fold_builtin_cpu): Add "goldmont".
5113 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5114 (ix86_option_override_internal): Add "goldmont".
5115 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5116 (processor_type): Add PROCESSOR_GOLDMONT.
5117 * config/i386/i386.md: Add CPU "glm".
5118 * config/i386/glm.md: New file.
5119 * config/i386/x86-tune.def: Add m_GOLDMONT.
5120 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5122 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5125 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5127 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5128 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5129 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5132 * config/i386/i386.c (ix86_fold_builtin): Handle
5133 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5136 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5137 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5139 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5142 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5143 (check_arch): Likewise.
5144 (check_fpu): Return the result rather than printing it.
5145 (end arch): Fix operator precedence.
5146 (end cpu): Likewise.
5147 (END): Print the result from check_fpu.
5149 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5150 Alan Hayward <alan.hayward@arm.com>
5151 David Sherwood <david.sherwood@arm.com>
5153 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5154 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5155 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5156 (*fcmuo<mode>_and): New patterns.
5158 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5160 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5161 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5162 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5163 (cmp_op, sve_imm_con): New code attributes.
5164 (SVE_COND_INT_CMP, imm_con): Delete.
5165 (cmp_op): Remove above unspecs from int attribute.
5166 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5168 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5169 comparison-specific unspecs.
5170 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5171 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5172 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5173 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5174 (*vec_fcm<cmp_op><mode>): Rename to...
5175 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5176 (*vec_fcmuo<mode>): Rename to...
5177 (*fcmuo<mode>): ...this and adjust likewise.
5178 (*pred_fcm<cmp_op><mode>): New pattern.
5179 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5180 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5182 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5184 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5185 (aarch64_emit_sve_predicated_cond): New function.
5186 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5187 (aarch64_emit_unspec_cond_or): Replace with...
5188 (aarch64_emit_sve_or_conds): ...this new function. Use
5189 aarch64_emit_sve_ptrue_op for the individual comparisons and
5190 aarch64_emit_binop to OR them together.
5191 (aarch64_emit_inverted_unspec_cond): Replace with...
5192 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5193 aarch64_emit_sve_ptrue_op for the comparison and
5194 aarch64_emit_unop to invert the result.
5195 (aarch64_expand_sve_vec_cmp_float): Update after the above
5196 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5198 2018-05-07 Nathan Sidwell <nathan@acm.org>
5200 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5201 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5202 (Backwards Compatibility): Likewise.
5204 2018-05-07 Luis Machado <luis.machado@linaro.org>
5208 2018-05-07 Luis Machado <luis.machado@linaro.org>
5210 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5211 <prefetch_dynamic_strides>: New const bool field.
5212 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5213 prefetch_dynamic_strides.
5214 (exynosm1_prefetch_tune): Likewise.
5215 (thunderxt88_prefetch_tune): Likewise.
5216 (thunderx_prefetch_tune): Likewise.
5217 (thunderx2t99_prefetch_tune): Likewise.
5218 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5220 (aarch64_override_options_internal): Update to set
5221 PARAM_PREFETCH_DYNAMIC_STRIDES.
5222 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5223 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5224 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5225 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5226 prefetch-dynamic-strides setting.
5228 2018-05-07 Luis Machado <luis.machado@linaro.org>
5230 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5231 <minimum_stride>: New const int field.
5232 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5233 minimum_stride field.
5234 (exynosm1_prefetch_tune): Likewise.
5235 (thunderxt88_prefetch_tune): Likewise.
5236 (thunderx_prefetch_tune): Likewise.
5237 (thunderx2t99_prefetch_tune): Likewise.
5238 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5239 (aarch64_override_options_internal): Update to set
5240 PARAM_PREFETCH_MINIMUM_STRIDE.
5241 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5242 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5243 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5244 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5245 stride is constant and is below the minimum stride threshold.
5247 2018-05-07 Luis Machado <luis.machado@linaro.org>
5249 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5252 2018-05-07 Luis Machado <luis.machado@linaro.org>
5254 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5255 <prefetch_dynamic_strides>: New const bool field.
5256 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5257 prefetch_dynamic_strides.
5258 (exynosm1_prefetch_tune): Likewise.
5259 (thunderxt88_prefetch_tune): Likewise.
5260 (thunderx_prefetch_tune): Likewise.
5261 (thunderx2t99_prefetch_tune): Likewise.
5262 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5264 (aarch64_override_options_internal): Update to set
5265 PARAM_PREFETCH_DYNAMIC_STRIDES.
5266 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5267 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5268 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5269 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5270 prefetch-dynamic-strides setting.
5272 2018-05-07 Luis Machado <luis.machado@linaro.org>
5274 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5275 <minimum_stride>: New const int field.
5276 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5277 minimum_stride field.
5278 (exynosm1_prefetch_tune): Likewise.
5279 (thunderxt88_prefetch_tune): Likewise.
5280 (thunderx_prefetch_tune): Likewise.
5281 (thunderx2t99_prefetch_tune): Likewise.
5282 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5283 (aarch64_override_options_internal): Update to set
5284 PARAM_PREFETCH_MINIMUM_STRIDE.
5285 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5286 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5287 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5288 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5289 stride is constant and is below the minimum stride threshold.
5291 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5294 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5295 the type is addressable. Don't force op into register if it has
5298 2018-05-05 Roland McGrath <mcgrathr@google.com>
5301 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5302 any section for which we don't know a specific type it should have,
5303 regardless of name. Previously this was done only for the exact
5304 names ".init_array", ".fini_array", and ".preinit_array".
5305 (default_elf_asm_named_section): Add comment about
5306 relationship with default_section_type_flags and SECTION_NOTYPE.
5307 (get_section): Don't consider it a type conflict if one side has
5308 SECTION_NOTYPE and the other doesn't, as long as neither has the
5309 SECTION_BSS et al used in the default_section_type_flags logic.
5311 2018-05-05 Tom de Vries <tom@codesourcery.com>
5314 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5315 (workaround_barsyncs): New function.
5316 (nvptx_reorg): Use workaround_barsyncs.
5317 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5318 (define_expand "nvptx_membar_cta"): New define_expand.
5319 (define_insn "*nvptx_membar_cta"): New insn.
5321 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5323 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5324 To improve optimization opportunities.
5325 * builtin-types.def: The new needed builtin types for the above.
5327 2018-05-04 Richard Biener <rguenther@suse.de>
5329 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5330 * gimple-ssa-store-merging.c
5331 (imm_store_chain_info::output_merged_store): Remove redundant create,
5332 release split_store vector contents on failure.
5333 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5334 scalar stmt vector on cache hit.
5336 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5338 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5340 * config.gcc (powerpc-xilinx-eabi*): Remove.
5341 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5343 (fusion_addis_mem_combo_load): Ditto.
5344 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5346 (rs6000_cpu_cpp_builtins): Ditto.
5347 * config/rs6000/rs6000-linux.c
5348 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5349 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5350 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5352 (rs6000_setup_reg_addr_masks): Ditto.
5353 (rs6000_init_hard_regno_mode_ok): Ditto.
5354 (rs6000_option_override_internal): Ditto.
5355 (legitimate_lo_sum_address_p): Ditto.
5356 (rs6000_legitimize_address): Ditto.
5357 (rs6000_legitimize_reload_address): Ditto.
5358 (rs6000_legitimate_address_p): Ditto.
5359 (abi_v4_pass_in_fpr): Ditto.
5360 (setup_incoming_varargs): Ditto.
5361 (rs6000_gimplify_va_arg): Ditto.
5362 (rs6000_split_multireg_move): Ditto.
5363 (rs6000_savres_strategy): Ditto.
5364 (rs6000_emit_prologue_components): Ditto.
5365 (rs6000_emit_epilogue_components): Ditto.
5366 (rs6000_emit_prologue): Ditto.
5367 (rs6000_emit_epilogue): Ditto.
5368 (rs6000_elf_file_end): Ditto.
5369 (rs6000_function_value): Ditto.
5370 (rs6000_libcall_value): Ditto.
5371 * config/rs6000/rs6000.h: Ditto.
5372 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5373 (TARGET_MINMAX): ... this. New.
5374 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5375 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5376 (*movsi_internal1_single): Delete.
5377 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5378 mfpu=, mxilinx-fpu): Delete.
5379 * config/rs6000/singlefp.h: Delete.
5380 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5381 * config/rs6000/t-rs6000: Ditto.
5382 * config/rs6000/t-xilinx: Delete.
5383 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5384 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5385 (VStype_simple): Delete.
5386 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5387 * config/rs6000/xfpu.h: Delete.
5388 * config/rs6000/xfpu.md: Delete.
5389 * config/rs6000/xilinx.h: Delete.
5390 * config/rs6000/xilinx.opt: Delete.
5391 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5392 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5394 2018-05-04 Tom de Vries <tom@codesourcery.com>
5397 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5400 2018-05-04 Richard Biener <rguenther@suse.de>
5403 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5404 (expand_complex_div_wide): Likewise.
5405 (expand_complex_operations_1): Likewise.
5406 (expand_complex_libcall): Preserve EH info of the original stmt.
5407 (tree_lower_complex): Handle removed blocks.
5408 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5409 on complex multiplication and division libcall builtins.
5411 2018-05-04 Richard Biener <rguenther@suse.de>
5414 * fold-const.c (negate_expr_p): Restrict negation of operand
5415 zero of a division to when we know that can happen without
5417 (fold_negate_expr_1): Likewise.
5419 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5422 * real.h (real_nextafter): Declare.
5423 * real.c (real_nextafter): New function.
5424 * fold-const-call.c (fold_const_nextafter): New function.
5425 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5426 CASE_CFN_NEXTTOWARD.
5427 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5428 even when arg1_mode is different from arg0_mode.
5430 2018-05-03 Nathan Sidwell <nathan@acm.org>
5432 * doc/extend.texi (Deprecated Features): Remove
5434 (Backwards Compatibility): Likewise.
5435 * doc/invoke.texi (C++ Language Options): Likewise.
5436 (C++ Dialect Options): Likewise.
5438 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5441 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5442 _mm512_mask_mullox_epi64): New intrinsics.
5444 2018-05-03 Tom de Vries <tom@codesourcery.com>
5447 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5448 dump files): Add offload-tree.
5450 2018-05-03 Richard Biener <rguenther@suse.de>
5452 PR tree-optimization/85615
5453 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5454 to loops not nested in BBs loop father to avoid creating multi-entry
5457 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5459 PR tree-optimization/70291
5460 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5461 arguments. Change return type to tree. Emit libcall as a new
5462 statement rather than replacing existing one when inplace_p is true.
5463 (expand_complex_multiplication_components): New function.
5464 (expand_complex_multiplication): Expand floating-point complex
5465 multiplication using the above.
5466 (expand_complex_division): Rename inner_type parameter to type.
5467 Update expand_complex_libcall call-site.
5468 (expand_complex_operations_1): Update expand_complex_multiplication
5469 and expand_complex_division call-sites.
5471 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5474 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5475 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5476 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5477 the highest significant bit of the shift count mask is clear. In
5478 check whether and[sq]i3 is needed verify that all significant bits
5479 of the shift count other than the highest are set.
5481 2018-05-02 Tom de Vries <tom@codesourcery.com>
5484 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5485 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5486 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5487 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5488 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5489 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5490 __builtin_goacc_parlevel_size.
5492 2018-05-02 Richard Biener <rguenther@suse.de>
5494 PR tree-optimization/85597
5495 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5496 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5499 2018-05-02 Tom de Vries <tom@codesourcery.com>
5502 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5503 dump files): Add ltrans-tree.
5505 2018-05-02 Tom de Vries <tom@codesourcery.com>
5508 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5509 dump files): Add wpa-ipa.
5511 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5513 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5514 powerpc*-*-linux*paired* target.
5515 * config/rs6000/750cl.h: Delete.
5516 * config/rs6000/paired.h: Delete.
5517 * config/rs6000/paired.md: Delete.
5518 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5520 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5521 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5522 comment. Remove paired float support.
5523 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5524 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5526 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5527 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5529 * config/rs6000/rs6000.c: Remove paired float support.
5530 (paired_expand_vector_init, paired_expand_vector_move,
5531 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5532 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5533 paired_expand_builtin, paired_expand_predicate_builtin,
5534 paired_init_builtins): Delete.
5535 * config/rs6000/rs6000.h: Remove paired float support.
5536 * config/rs6000/rs6000.md: Remove paired float support.
5537 (move_from_CR_ov_bit): Delete.
5538 * config/rs6000/rs6000.opt (mpaired): Delete.
5539 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5540 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5542 2018-05-02 Richard Biener <rguenther@suse.de>
5545 * gimplify.c (gimplify_save_expr): When in SSA form allow
5546 SAVE_EXPRs to compute to SSA vars.
5548 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5551 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5552 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5553 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5554 clobber operands[2], instead use a new pseudo. Formatting fixes.
5556 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5558 PR tree-optimization/85586
5559 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5560 exit early for statements in the same group if the accesses are
5563 2018-05-02 Tom de Vries <tom@codesourcery.com>
5566 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5569 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5571 PR tree-optimization/85143
5572 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5574 2018-05-01 Tom de Vries <tom@codesourcery.com>
5577 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5578 not found" error message.
5580 2018-05-01 Tom de Vries <tom@codesourcery.com>
5583 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5584 * vec.c (test_ordered_remove_if): New function.
5585 (vec_c_tests): Call test_ordered_remove_if.
5586 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5587 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5588 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5589 VEC_ORDERED_REMOVE_IF.
5591 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5593 PR tree-optimization/82665
5594 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5595 pointer subtraction where arguments come from a memchr call.
5597 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5599 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5600 --push-state --as-needed and --pop-state instead of --as-needed and
5601 --no-as-needed if ld supports it.
5602 * configure: Regenerated.
5605 * doc/install.texi2html: Replace _002d with - and _002a with * in
5606 generated html files using sed.
5608 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5611 * gcc-rich-location.c (blank_line_before_p): New function.
5612 (use_new_line): New function.
5613 (gcc_rich_location::add_fixit_insert_formatted): New function.
5614 * gcc-rich-location.h
5615 (gcc_rich_location::add_fixit_insert_formatted): New function.
5617 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5619 * selftest.c (assert_streq): Rename "expected" and "actual" to
5620 "val1" and "val2". Extend NULL-handling to cover both inputs
5621 symmetrically, while still requiring both to be non-NULL for a pass.
5622 * selftest.h (assert_streq): Rename "expected" and "actual" to
5624 (ASSERT_EQ): Likewise.
5625 (ASSERT_EQ_AT): Likewise.
5626 (ASSERT_KNOWN_EQ): Likewise.
5627 (ASSERT_KNOWN_EQ_AT): Likewise.
5628 (ASSERT_NE): Likewise.
5629 (ASSERT_MAYBE_NE): Likewise.
5630 (ASSERT_MAYBE_NE_AT): Likewise.
5631 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5632 the assertion to pass.
5633 (ASSERT_STREQ_AT): Likewise.
5635 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5637 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5638 interaction with -pie.
5640 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5642 * selftest.h: Fix alphabetization of per-source-file selftest
5645 2018-04-30 Jason Merrill <jason@redhat.com>
5647 PR c++/61982 - dead stores to destroyed objects.
5648 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5651 2018-04-30 Jason Merrill <jason@redhat.com>
5653 * tree.c (build_clobber): New.
5654 * tree.h: Declare it.
5655 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5657 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5659 * diagnostic-show-locus.c (layout::layout): Update for
5660 location_get_source_line returning a char_span.
5661 (struct char_span): Move to input.h.
5662 (struct correction): Update for fields in char_span becoming
5664 (struct source_line): Update for location_get_source_line
5665 returning a char_span.
5666 (layout::print_line): Likewise.
5667 * edit-context.c (edited_file::print_content): Likewise.
5668 (edited_file::print_diff_hunk): Likewise.
5669 (edited_file::print_run_of_changed_lines): Likewise.
5670 (edited_file::get_num_lines): Likewise.
5671 (edited_line::edited_line): Likewise.
5672 * final.c (asm_show_source): Likewise.
5673 * input.c (location_get_source_line): Convert return type
5674 from const char * to char_span, losing the final "line_len"
5676 (dump_location_info): Update for the above.
5677 (get_substring_ranges_for_loc): Likewise. Use a char_span
5678 when handling the literal within the line.
5679 (test_reading_source_line): Update for location_get_source_line
5680 returning a char_span.
5681 * input.h (class char_span): Move here from
5682 diagnostic-show-locus.c, converting from a struct to a class.
5683 Make data members private.
5684 (char_span::operator bool): New.
5685 (char_span::length): New.
5686 (char_span::get_buffer): New.
5687 (char_span::operator[]): New.
5688 (char_span::subspan): Make const.
5689 (char_span::xstrdup): New.
5690 (location_get_source_line): Convert return type from const char *
5691 to char_span, losing the final "line_size" param.
5693 2018-04-30 Jan Hubicka <jh@suse.cz>
5695 * lto-wrapper.c (ltrans_priorities): New static var.
5696 (cmp_priority): New.
5697 (run_gcc): Read priorities and if doing parallel build order
5698 the Makefile by them.
5700 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5702 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5704 2018-04-30 Richard Biener <rguenther@suse.de>
5706 * tree-cfg.c (verify_address): Remove base argument, add
5707 flag whether to check TREE_ADDRESSABLE and do that.
5708 (verify_expr): Remove.
5709 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5710 (verify_gimple_assign_single): Likewise.
5711 (verify_gimple_switch): Likewise.
5712 (verify_expr_location_1): Dereference tp once. Add (disabled)
5713 piece from verify_expr.
5714 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5716 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5718 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5720 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5722 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5723 (small_data_pattern): Likewise.
5724 (arc_rewrite_small_data): Likewise.
5725 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5726 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5727 (get_symbol_alignment): New function.
5728 (legitimate_small_data_address_p): Likewise.
5729 (legitimate_scaled_address): Update, call
5730 legitimate_small_data_address_p.
5731 (output_sdata): New static variable.
5732 (arc_print_operand): Update how we handle small data operands.
5733 (arc_print_operand_address): Likewise.
5734 (arc_legitimate_address_p): Update, use
5735 legitimate_small_data_address_p.
5736 (arc_rewrite_small_data_p): Remove.
5737 (arc_rewrite_small_data_1): Likewise.
5738 (arc_rewrite_small_data): Likewise.
5739 (small_data_pattern): Likewise.
5740 (compact_sda_memory_operand): Update to use
5741 legitimate_small_data_address_p and get_symbol_alignment.
5742 (prepare_move_operands): Don't rewite sdata pattern.
5743 (prepare_extend_operands): Remove.
5744 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5746 (zero_extendqisi2): Likewise.
5747 (zero_extendhisi2): Likewise.
5748 (extendqihi2): Likewise.
5749 (extendqisi2): Likewise.
5750 (extendhisi2): Likewise.
5754 * config/arc/constraints.md (Usd): Change it to memory constraint.
5756 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5758 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5759 as source of std instructions.
5760 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5761 allow 6-bit constants as source for store instructions.
5762 (movdi_insn): Update instruction pattern to allow 6-bit constants
5763 as source for store instructions.
5765 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5767 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5769 2018-04-30 Nathan Sidwell <nathan@acm.org>
5770 Sandra Loosemore <sandra@codesourcery.com>
5772 * dumpfile.c (dump_open): Allow '-' for stdout.
5773 * doc/invoke.texi (Developer Options): Document dump filename
5774 determination early. Document stdin/stdout selection.
5776 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5778 Microblaze Target: PIC data text relative
5780 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5781 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5783 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5784 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5785 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5786 New addressing mode for data-text relative position indepenedent code.
5787 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5788 'ADDRESS_SYMBOLIC_TXT_REL'.
5789 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5790 (microblaze_legitimate_pic_operand): Exclude function calls from
5791 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5792 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5794 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5795 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5796 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5797 for 'address + offset'.
5798 (microblaze_expand_prologue): Add new function prologue call for
5800 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5801 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5802 table in case of TARGET_PIC_DATA_TEXT_REL.
5803 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5804 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5805 Add new macros 'UNSPEC_TEXT',
5806 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5807 + exclude function calls from 'UNSPEC_PLT' in case of data text
5809 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5810 new target hook for generating address diff vector tables in case of
5812 * doc/tm.texi : Regenerate.
5813 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5814 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5815 of addr diff vector generation.
5816 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5817 target hook definition.
5818 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5819 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5820 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5821 option for data text relative.
5823 2018-04-30 Richard Biener <rguenther@suse.de>
5825 * tree-chrec.h (evolution_function_is_constant_p): Remove
5827 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5829 2018-04-30 Richard Biener <rguenther@suse.de>
5832 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5834 2018-04-30 Richard Biener <rguenther@suse.de>
5836 PR tree-optimization/28364
5837 PR tree-optimization/85275
5838 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5839 copying first exit test.
5841 2018-04-28 Mark Wielaard <mark@klomp.org>
5843 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5845 (dwarf_AT): Handle DW_AT_addr_base.
5846 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5848 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5851 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5852 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5853 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5854 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5856 2018-04-28 Richard Biener <rguenther@suse.de>
5858 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5859 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5860 to reflect use. Only add interesting stmts.
5862 2018-04-27 Martin Jambor <mjambor@suse.cz>
5865 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5866 the jump function allows for passing through aggregate values.
5868 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5870 * input.h (in_system_header_at): Convert from macro to inline
5872 (from_macro_expansion_at): Likewise.
5873 (from_macro_definition_at): Likewise.
5875 2018-04-27 Jeff Law <law@redhat.com>
5877 * config.gcc: Mark tile* targets as deprecated/obsolete.
5879 2018-04-27 Richard Biener <rguenther@suse.de>
5881 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5884 2018-04-27 Richard Biener <rguenther@suse.de>
5886 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5888 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5890 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5891 with Yd constraint. Set "preferred_for_speed" attribute from
5892 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5894 (*movdi_internal): Ditto.
5895 (movti_interunit splitters): Remove
5896 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5897 (movdi_interunit splitters): Ditto.
5898 * config/i386/constraints.md (Ye): Remove.
5899 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5901 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5904 * config/aarch64/constraints.md (Usg): Limit to 31.
5907 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5909 PR tree-optimization/85529
5910 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5911 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5912 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5913 zero extension or masking of the MSB bit.
5914 (optimize_range_tests): Add FIRST_BB argument, pass it through
5915 to optimize_range_tests_var_bound.
5916 (maybe_optimize_range_tests, reassociate_bb): Adjust
5917 optimize_range_tests callers.
5919 2018-04-26 Richard Biener <rguenther@suse.de>
5920 Jakub Jelinek <jakub@redhat.com>
5922 * cgraph.h (symbol_table): Just declare debug method here.
5923 * symtab.c (symbol_table::debug): Define.
5925 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5927 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5929 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5931 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5932 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5933 (*movdi_internal): Substitute Yi and Yj constraint with x
5934 and Ym and Yn constraint with y constraint. Update "isa"
5935 attribute and set "preferred_for_speed" attribute from
5936 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5937 (*movsi_internal): Ditto.
5938 (*movdf_internal): Ditto.
5939 (*movsf_internal): Ditto.
5940 (*zero_extendsidi2): Ditto.
5941 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5942 (sse2_loadld): Ditto.
5943 (*vec_extract<ssevecmodelower>_0): Ditto.
5944 (*vec_extractv4si_0_zext_sse4): Ditto.
5945 (vec_concatv2di): Ditto.
5946 (*vec_dup<mode>): Ditto.
5947 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5948 * config/i386/constraints.md (Yi): Remove.
5953 2018-04-26 Nathan Sidwell <nathan@acm.org>
5955 * dumpfile.c (dump_open): New.
5956 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5957 (dump_finish): Detect stdio/stderr by value not name.
5959 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5961 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5963 2018-04-26 Tom de Vries <tom@codesourcery.com>
5966 * config/nvptx/nvptx.c (verify_neutering_jumps)
5967 (verify_neutering_labels): New function
5968 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5970 2018-04-26 Tom de Vries <tom@codesourcery.com>
5973 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5974 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5977 2018-04-26 Richard Biener <rguenther@suse.de>
5978 Tom de Vries <tom@codesourcery.com>
5981 * lto-streamer-out.c (output_function): Fixup loops if required to match
5982 discovery done in the reader.
5984 2018-04-26 Richard Biener <rguenther@suse.de>
5986 PR tree-optimization/85116
5987 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5988 have a loop exit from the single latch predecessor. Remove
5989 case of header with just condition.
5990 (ch_base::copy_headers): Exclude infinite loops from any
5992 (pass_ch::execute): Record exits.
5994 2018-04-26 Richard Biener <rguenther@suse.de>
5996 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5997 prologue cost vector and pass it to vect_get_load_cost.
5998 (vect_get_peeling_costs_all_drs): Likewise.
5999 (vect_peeling_hash_get_lowest_cost): Likewise.
6000 (vect_enhance_data_refs_alignment): Likewise.
6002 2018-04-26 Richard Biener <rguenther@suse.de>
6005 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6006 checking of integer<->pointer conversions.
6007 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6008 sign-/zero-extending pointer types.
6009 (expand_omp_for_static_chunk): Likewise.
6011 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6012 Jean Lee <xiaoyur347@gmail.com>
6014 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6015 (TARGET_ASAN_SHADOW_OFFSET): Define.
6016 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6017 true for -fsanitize=address.
6019 2018-04-25 Mark Wielaard <mark@klomp.org>
6021 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6024 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6026 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6027 than "alu", remove explicit "memory" and "imm_disp" attributes.
6028 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6031 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6032 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6033 gen_lowpart_no_emit.
6035 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6038 * config/i386/i386.c (ix86_expand_builtin): Change memory
6039 operand to XI, extend p0 to Pmode.
6040 * config/i386/i386.md: Change unspec volatile and operand
6041 1 mode to XI, change operand 0 mode to P.
6043 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6045 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6046 GET_MODE_MASK before any checking.
6047 (nds32_can_use_bset_p): Likewise.
6048 (nds32_can_use_btgl_p): Likewise.
6050 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6052 * config/nds32/nds32-doubleword.md: New define_split pattern for
6053 illegal register number.
6055 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6057 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6059 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6061 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6063 2018-04-25 Richard Biener <rguenther@suse.de>
6065 * lto-streamer.h (LTO_major_version): Bump to 8.
6067 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6069 * BASE-VER: Set to 9.0.0.
6071 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6073 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6074 in __abskf2 and __powikf2.
6076 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6079 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6080 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6081 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6082 Use the above on operand 2. Reindent.
6083 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6085 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6088 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6090 * config/i386/i386.opt (mcet): Removed.
6091 * doc/install.texi: Remove -mcet documentation.
6092 * doc/invoke.texi: Likewise.
6094 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6097 * doc/install.texi: Remove -mcet from bootstrap-cet.
6099 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6102 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6103 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6107 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6108 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6109 containing a CONST_VECTOR.
6111 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6113 * doc/install.texi: Update newlib dependency for nvptx.
6115 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6118 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6119 instead of INTVAL when shifting x left.
6121 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6123 PR tree-optimization/85478
6124 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6125 vect_grouped_store_supported for single element vectors.
6127 2018-04-24 Richard Biener <rguenther@suse.de>
6130 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6131 load cost increase to the case of non-constant step.
6133 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6136 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6137 destination if any_malformed_asm.
6139 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6142 * expr.c (store_field): In the bitfield case, if the value comes from
6143 a function call and is returned in registers by means of a PARALLEL,
6144 do not change the mode of the temporary unless BLKmode and VOIDmode.
6146 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6148 PR rtl-optimization/85423
6149 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6150 dependencies to debug insns when the previous insn is non-debug.
6152 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6154 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6155 enums into a single definition.
6156 (fls): Fix predicates and printing.
6159 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6161 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6162 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6163 and short u6 immediate.
6164 (check_if_valid_sleep_operand): Remove.
6165 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6167 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6169 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6170 flag_always_save_lp condition.
6171 * config/nds32/nds32.opt (malways-save-lp): New option.
6173 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6175 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6176 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6177 * config/nds32/nds32.h
6178 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6179 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6181 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6183 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6184 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6186 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6187 Chung-Ju Wu <jasonwucj@gmail.com>
6189 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6191 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6192 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6194 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6196 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6198 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6200 * config/nds32/nds32-protos.h (nds32_data_alignment,
6201 nds32_local_alignment): Declare.
6202 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6203 nds32_local_alignment): New functions.
6204 (TARGET_CONSTANT_ALIGNMENT): Define.
6205 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6207 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6209 * config/nds32/nds32.c
6210 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6211 (TARGET_MODES_TIEABLE_P): Likewise.
6213 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6215 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6218 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6219 Chung-Ju Wu <jasonwucj@gmail.com>
6221 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6222 for unaligned access.
6223 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6225 * config/nds32/nds32-intrinsic.md: Likewise.
6226 * config/nds32/nds32_intrinsic.h: Likewise.
6227 * config/nds32/nds32.h (nds32_builtins): Likewise.
6228 * config/nds32/nds32.opt (munaligned-access): New option.
6229 * config/nds32/nds32.c (nds32_asm_file_start): Display
6230 flag_unaligned_access status.
6232 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6234 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6235 -mno-relax is present.
6236 * config/riscv/linux.h (LINK_SPEC): Ditto.
6238 2018-04-20 Martin Sebor <msebor@redhat.com>
6241 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6243 (gimple_fold_builtin_stxcpy_chk): Same.
6244 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6246 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6249 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6250 __powikf2 when long double is IEEE 128-bit.
6252 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6254 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6255 step to make sure stack always aligned.
6257 2018-04-20 Carl Love <cel@us.ibm.com>
6260 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6261 size check for arg0.
6263 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6264 Tom de Vries <tom@codesourcery.com>
6267 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6268 Emit insns for calls too.
6269 (nvptx_find_par): Always look for worker-level predecessor insn.
6270 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6272 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6273 (nvptx_process_pars): Propagate frames for calls.
6275 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6278 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6280 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6281 (ix86_handle_option): Don't handle OPT_mibt.
6282 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6284 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6286 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6287 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6288 (ix86_target_macros): Define __CET__ with flag_cf_protection
6289 for -fcf-protection.
6290 * config/i386/i386.c (isa2_opts): Remove -mibt.
6291 * config/i386/i386.h (TARGET_IBT): Removed.
6292 (TARGET_IBT_P): Likewise.
6293 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6294 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6295 * config/i386/i386.opt (mcet): Update help message.
6298 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6299 -mcet as an alias for -mshstk.
6301 2018-04-20 Richard Biener <rguenther@suse.de>
6304 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6305 complexity by forcing a single use of the multiply operand.
6307 2018-04-20 Martin Jambor <mjambor@suse.cz>
6310 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6311 recursion dependency to only apply to non-clones.
6313 2018-04-20 Martin Jambor <mjambor@suse.cz>
6316 * ipa-cp.c (create_specialized_node): Check that clones of
6317 self-recursive edges exist during IPA-CP.
6319 2018-04-19 Toon Moene <toon@moene.org>
6321 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6324 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6326 PR tree-optimization/85467
6327 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6328 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6329 VECTOR_CST element to type.
6331 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6334 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6335 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6336 (builtin_longjmp): Likewise.
6337 (save_stack_nonlocal): New pattern.
6338 (restore_stack_nonlocal): Likewise.
6340 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6343 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6344 Replace ASM_OUTPUT_LABEL with fprintf.
6346 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6349 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6350 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6351 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6352 define __IBT__ and __SHSTK__ for -fcf-protection.
6353 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6355 (ix86_trampoline_init): Likewise.
6356 (x86_output_mi_thunk): Likewise.
6357 (ix86_notrack_prefixed_insn_p): Likewise.
6358 (ix86_option_override_internal): Don't disallow -fcf-protection.
6359 * config/i386/i386.md (rdssp<mode>): Also enable for
6361 (incssp<mode>): Likewise.
6362 (nop_endbr): Likewise.
6363 * config/i386/i386.opt (mcet): Change help message to built-in
6367 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6368 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6369 enable CET built-in functions.
6371 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6373 * common/config/i386/i386-common.c
6374 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6375 OPTION_MASK_ISA_MOVDIRI_UNSET,
6376 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6377 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6378 * config.gcc (movdirintrin.h): New header.
6379 * config/i386/cpuid.h (bit_MOVDIRI,
6380 bit_MOVDIR64B): New bits.
6381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6383 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6384 (VOID, PVOID, PCVOID)): New function types.
6385 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6386 __builtin_ia32_directstoreu_u64,
6387 __builtin_ia32_movdir64b): New builtins.
6388 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6389 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6391 (ix86_valid_target_attribute_inner_p): Ditto.
6392 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6393 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6394 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6395 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6396 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6397 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6398 (movdiri<mode>, movdir64b_<mode>): New.
6399 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6400 * config/i386/immintrin.h: Include movdirintrin.h.
6401 * config/i386/movdirintrin.h: New file.
6402 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6404 2018-04-19 Richard Biener <rguenther@suse.de>
6407 * cfg.c (clear_bb_flags): When loop state says we have
6408 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6410 2018-04-19 Richard Biener <rguenther@suse.de>
6412 PR tree-optimization/84737
6413 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6414 copying restrict info.
6415 (vect_setup_realignment): Use it.
6416 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6417 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6418 the first DR to all generated stores.
6419 (vectorizable_load): Likewise for loads.
6421 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6423 PR tree-optimization/85446
6424 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6425 the integral and pointer types to have the same precision.
6427 * doc/install.texi: Document --disable-cet being the default and
6430 2018-04-18 Martin Liska <mliska@suse.cz>
6432 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6435 2018-04-18 Martin Liska <mliska@suse.cz>
6438 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6441 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6442 arguments if they are comparable.
6444 2018-04-18 Martin Liska <mliska@suse.cz>
6447 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6450 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6453 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6456 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6457 ENDBR after calling __morestack.
6459 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6462 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6463 by using gcc_base_ver to generate a gcc_driver_version, and use
6464 it when generating GCC_DRIVER_NAME.
6465 * configure: Regenerate.
6467 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6470 * config.gcc: Obsolete powerpc*-*-*spe*.
6472 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6475 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6476 (stabstr_D): Change type of unum from unsigned int to
6477 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6480 2018-04-17 Jim Wilson <jimw@sifive.com>
6483 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6484 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6485 Set arg_pointer_offset after using pretend_args_size.
6487 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6489 PR rtl-optimization/85431
6490 * dse.c (record_store): Ignore zero width stores.
6493 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6494 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6495 __builtin_stack_restore rather than after it.
6496 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6497 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6498 argument instead of virtual_dynamic_stack_rtx.
6500 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6502 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6504 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6505 Add note to error message to explain internal mapping of overloaded
6506 built-in function name to non-overloaded built-in function name.
6507 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6510 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6513 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6514 where the inputs overlap with the output.
6516 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6518 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6519 (=v, v) alternative and explicit "memory" attribute.
6520 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6521 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6523 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6524 "sselog1" type instead of "sselog".
6525 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6526 "sselog". Remove explicit "memory" attribute.
6527 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6528 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6530 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6531 "sselog1" type instead of "sselog". Remove explicit "memory"
6533 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6534 use "sselog1" type instead of "sselog". Remove explicit "memory"
6536 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6537 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6539 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6540 "sselog1" type instead of "sselog". Remove explicit "memory"
6542 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6543 use "sselog1" type instead of "sselog". Remove explicit "memory"
6547 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6550 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6553 2018-04-17 Martin Jambor <mjambor@suse.cz>
6556 * ipa-cp.c (create_specialized_node): Call
6557 expand_all_artificial_thunks if necessary.
6559 2018-04-17 Martin Liska <mliska@suse.cz>
6562 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6563 in message, remote space in between '_G' and '('.
6565 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6568 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6569 avx512f_vmcmp<mode>3<round_saeonly_name>,
6570 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6571 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6572 avx512f_rndscale<mode><round_saeonly_name>,
6573 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6574 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6575 Use %<iptr>2 instead of %2 for -masm=intel.
6576 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6577 avx512f_vcvttss2usi<round_saeonly_name>,
6578 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6580 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6581 avx512f_vcvttsd2usi<round_saeonly_name>,
6582 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6583 Use %q1 instead of %1 for -masm=intel.
6584 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6585 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6586 of %3 for -masm=intel.
6587 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6589 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6591 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6593 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6594 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6596 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6598 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6599 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6600 %g1 and one with %0 and %1.
6601 (avx512er_vmrcp28<mode><round_saeonly_name>,
6602 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6604 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6605 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6606 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6607 of %0 and %{%4%} for -masm=intel.
6608 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6609 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6610 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6611 order of %0 and %{%5%}%{z%} for -masm=intel.
6613 2018-04-17 Jan Hubicka <jh@suse.cz>
6616 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6618 2018-04-17 Martin Liska <mliska@suse.cz>
6621 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6622 for target_clone error message. Make default implementation
6623 clone to be a local declaration.
6624 (separate_attrs): Add new argument and check for an empty
6626 (expand_target_clones): Handle it.
6627 (ipa_target_clone): Make redirection just for target_clones
6630 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6631 Tom de Vries <tom@codesourcery.com>
6634 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6635 tiled basic blocks without omp continue statements.
6637 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6640 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6641 vec_extract expression as having side effects to make sure it gets
6644 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6647 * config/i386/i386.c (get_builtin_code_for_version): Check
6650 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6653 * gcc/config.gcc: Support "skylake".
6654 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6656 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6657 (processor_target_table): Add "skylake".
6658 (ix86_option_override_internal): Add "skylake".
6659 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6660 PROCESSOR_CANNONLAKE.
6661 (get_builtin_code_for_version): Fix priority for
6662 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6663 PROCESSOR_SKYLAKE-AVX512.
6664 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6665 (processor_type): Add PROCESSOR_SKYLAKE.
6667 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6668 Jason Merrill <jason@redhat.com>
6671 * convert.c (convert_to_integer_1): Use direct recursion for
6672 enumeral types and types with a precision less than the number
6673 of bits in their mode.
6675 2018-04-16 Julia Koval <julia.koval@intel.com>
6678 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6679 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6681 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6684 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6685 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6686 and -mno-direct-move.
6688 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6691 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6692 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6693 Ensure negative shifts result in {0}.
6695 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6697 PR rtl-optimization/79916
6698 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6699 regs (if any) to define how to gnerate SD moves when LRA is in
6702 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6704 PR rtl-optimization/85393
6705 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6706 * except.c (expand_dw2_landing_pad_for_region): Make static.
6707 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6708 a label and unconditional jump to old_bb, rather than
6709 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6712 PR rtl-optimization/85376
6713 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6714 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6715 instead of a specific value.
6717 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6718 Bin Cheng <bin.cheng@arm.com>
6720 PR tree-optimization/82965
6721 PR tree-optimization/83991
6722 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6723 by_profile_only parameter.
6724 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6725 information if the loop was predicted to iterate too many times.
6726 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6728 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6731 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6734 2018-04-13 Martin Liska <mliska@suse.cz>
6735 Jakub Jelinek <jakub@redhat.com>
6738 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6739 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6740 * builtins.c (expand_builtin_memory_copy_args): Use
6741 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6742 handle dest_addr == pc_rtx.
6744 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6747 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6748 asked to not generate direct moves.
6749 (fix_trunc<mode>si2_stfiwx): Similar.
6750 (fix_trunc<mode>si2_internal): Similar.
6752 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6755 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6756 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6757 lookup if dest in some wider mode is known to be const0_rtx and
6758 if so, record permanent equivalence for it to be ZERO_EXTEND of
6759 the narrower mode destination.
6761 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6763 * lto-streamer-out.c (output_function): Revert 259346.
6764 * omp-expand.c (expand_oacc_for): Likewise.
6766 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6768 PR rtl-optimization/85354
6769 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6770 * sel-sched.c (sel_global_init): ... here.
6772 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6775 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6776 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6777 mode for PE-COFF targets.
6778 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6779 (i386_pe_asm_lto_end): Likewise.
6780 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6781 (TARGET_ASM_LTO_END): Likewise.
6782 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6783 (i386_pe_asm_lto_start): New function.
6784 (i386_pe_asm_lto_end): Likewise.
6786 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6787 Richard Biener <rguenther@suse.de>
6790 * lto-streamer-out.c (output_function): Fix CFG loop state before
6792 * omp-expand.c (expand_oacc_for): Handle calls to internal
6793 functions like regular functions.
6795 2018-04-12 Richard Biener <rguenther@suse.de>
6798 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6799 for the early LTO debug to properly generate references to it
6800 during DIE emission. Do not re-use that for the skeleton for
6802 (dwarf2out_early_finish): Likewise.
6804 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6807 * config/i386/sse.md
6808 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6809 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6810 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6811 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6812 and output is a reg, avoid creating invalid lowpart subreg, but
6813 instead split into a 512-bit move. Don't split if not AVX512VL,
6814 input is xmm16+ reg and output is a mem.
6815 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6816 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6817 xmm16+ reg and output is a mem.
6819 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6821 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6822 also for flag_dwarf2_cfi_asm.
6824 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6826 PR rtl-optimization/85342
6827 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6828 a bool scalar var inside of the loop instead. Don't try to update
6829 recog_data.operand after failed apply_change_group.
6831 2018-04-12 Tom de Vries <tom@codesourcery.com>
6834 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6835 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6836 array with flexible array member as array without given dimension.
6837 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6838 argument for undefined param to true.
6840 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6843 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6844 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6845 from PowerPC section.
6846 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6847 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6848 help text that is too long.
6849 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6850 help text that is too long.
6851 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6852 help text that is too long.
6854 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6856 * config/alpha/alpha.md (stack_probe_internal): Rename
6857 from "probe_stack". Update all callers.
6859 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6861 PR rtl-optimization/84566
6862 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6863 sched_macro_fuse_insns.
6865 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6868 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6869 (compute_block_dependences): ... from here.
6871 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6873 PR tree-optimization/85331
6874 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6875 from int to HOST_WIDE_INT.
6877 2018-04-11 Martin Jambor <mjambor@suse.cz>
6880 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6881 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6882 not the same as the source val.
6883 (cgraph_edge_brings_value_p): New parameter.
6884 (gather_edges_for_value): Pass destination value to
6885 cgraph_edge_brings_value_p.
6886 (perhaps_add_new_callers): Likewise.
6887 (get_info_about_necessary_edges): Likewise and exclude values brought
6888 only by self-recursive edges.
6889 (create_specialized_node): Redirect only clones of self-calling edges.
6890 (+self_recursive_pass_through_p): New function.
6891 (find_more_scalar_values_for_callers_subset): Use it.
6892 (find_aggregate_values_for_callers_subset): Likewise.
6893 (known_aggs_to_agg_replacement_list): Removed.
6894 (decide_whether_version_node): Re-calculate known constants for all
6895 remaining context clones.
6897 2018-04-11 Richard Biener <rguenther@suse.de>
6900 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6901 from early DWARF output.
6902 (dwarf2out_early_finish): Output line info unconditionally into
6903 early DWARF and add reference to it.
6905 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6908 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6909 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6910 other than V2DFmode using iptr mode attribute.
6911 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6913 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6915 PR rtl-optimization/84659
6916 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6918 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6921 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6923 (output_loc_list): Pass address of a dummy size variable even in the
6924 locview handling loop.
6925 (index_location_lists): Add comment on why skip_loc_list_entry can't
6928 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6931 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6934 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6937 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6938 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6939 and -mstring-compare-inline-limit.
6941 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6944 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6945 for stack clash protection in a register whenever we need it to be in
6948 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6950 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6951 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6953 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6956 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6958 (mlong-double-): Ditto.
6959 * config/rs6000/sysv4.opt (msdata=): Ditto.
6960 (mtls-size=): Ditto.
6962 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6964 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6965 erroneous entries for
6966 "vector int vec_ldl (int, long int *)", and
6967 "vector unsigned int vec_ldl (int, unsigned long int *)".
6968 Add comments and entries for
6969 "vector bool char vec_ldl (int, bool char *)",
6970 "vector bool short vec_ldl (int, bool short *)",
6971 "vector bool int vec_ldl (int, bool int *)",
6972 "vector bool long long vec_ldl (int, bool long long *)",
6973 "vector pixel vec_ldl (int, pixel *)",
6974 "vector long long vec_ldl (int, long long *)",
6975 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6976 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6977 type tree bool_long_long_type_node and correct definition of
6978 bool_V2DI_type_node to make reference to this new type tree.
6979 (rs6000_mangle_type): Replace erroneous reference to
6980 bool_long_type_node with bool_long_long_type_node.
6981 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6982 comments to emphasize sign distinctions for char and int types and
6983 replace RS6000_BTI_bool_long constant with
6984 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6985 use of RS6000_BTI_pixel.
6986 (bool_long_type_node): Remove this macro definition.
6987 (bool_long_long_type_node): New macro definition
6989 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6991 PR rtl-optimization/85300
6992 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6993 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6994 simplify_unary_operation fails.
6996 2018-04-10 Martin Liska <mliska@suse.cz>
6998 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6999 cgraph_edge and ipa_ref.
7001 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7005 * config/i386/sse.md
7006 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7007 computation of the VEC_MERGE selector from mask.
7008 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7009 Fix decoding of the VEC_MERGE selector into mask.
7011 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7013 PR tree-optimization/85286
7014 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7016 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7018 * final.c (final_1): Set insn_last_address as well as
7019 insn_current_address.
7021 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7024 * explow.c (emit_stack_probe): Call validize_mem on memory location
7025 before passing it to gen_probe_stack. Create address operand and
7026 legitimize it for the probe_stack_address case.
7028 2018-04-09 Jan Hubicka <jh@suse.cz>
7031 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7032 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7033 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7035 (free_lang_data): Rebuild type inheritance graph.
7037 2018-04-09 Martin Sebor <msebor@redhat.com>
7039 * invoke.texi (-finline-small-functions): Mention other optimization
7041 (-findirect-inlining, -fpartial-inlining): Same.
7042 (-finline-functions-called-once): Same.
7043 (-freorder-blocks-and-partition): Same.
7045 2018-04-09 Jan Hubicka <jh@suse.cz>
7048 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7049 jumps; choose last target that matches the criteria (i.e.
7050 no partition changes for non-crossing jumps).
7051 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7052 support for redirecting crossing jumps to non-crossing.
7054 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7056 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7057 also for naked functions.
7059 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7061 * config/arc/arc.md (add_shift): New pattern.
7062 (add_shift2): Likewise.
7063 (sub_shift): Likewise.
7064 (sub_shift_cmp0_noout): Likewise.
7065 (compare_si_ashiftsi): Likewise.
7066 (xbfu_cmp0_noout): New combine pattern.
7067 (xbfu_cmp0"): Likewise.
7068 (movsi_set_cc_insn): Place the predicable variant first.
7069 (commutative_binary_cmp0_noout): Remove clobber.
7070 (commutative_binary_cmp0): New pattern.
7071 (noncommutative_binary_cmp0): Likewise.
7072 (noncommutative_binary_cmp0_noout): Likewise.
7073 (noncommutative_binary_comparison_result_used): Removed.
7074 (rsub_cmp0): New pattern.
7075 (rsub_cmp0_noout): Likewise.
7076 (extzvsi): Changed, keep only meaningful variants.
7077 (SQH, SEZ): New iterators.
7078 (SQH_postfix): New mode attribute.
7079 (SEZ_prefix): New code attribute.
7080 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7081 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7082 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7084 (noncommutative_operator): Check the availability of barrel
7087 2018-04-09 Richard Biener <rguenther@suse.de>
7089 PR tree-optimization/85284
7090 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7091 Only use the niter constraining form of simple_iv when the exit
7094 2018-04-09 Tom de Vries <tom@codesourcery.com>
7097 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7098 (define_expand "*memory_barrier"): New define_expand.
7099 (define_insn "memory_barrier"): New insn.
7101 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7103 PR rtl-optimization/80463
7104 PR rtl-optimization/83972
7105 PR rtl-optimization/83480
7107 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7108 correct producer for the insn.
7109 (tidy_control_flow): Fixup seqnos in case of debug insns.
7111 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7113 PR rtl-optimization/83913
7115 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7116 different sched-times when merging exprs.
7118 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7120 PR rtl-optimization/83962
7122 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7123 tidy_fallthru_edge and tidy_control_flow.
7125 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7127 PR rtl-optimization/83530
7129 * sel-sched.c (force_next_insn): New global variable.
7130 (remove_insn_for_debug): When force_next_insn is true, also leave only
7131 next insn in the ready list.
7132 (sel_sched_region): When the region wasn't scheduled, make another pass
7133 over it with force_next_insn set to 1.
7135 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7137 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7139 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7140 for interrupt control.
7141 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7142 functions for interrupt control.
7143 * config/nds32/nds32-intrinsic.md: Likewise.
7144 * config/nds32/nds32_intrinsic.h: Likewise.
7145 * config/nds32/nds32.h (nds32_builtins): Likewise.
7147 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7149 * config/nds32/nds32.c (nds32_init_machine_status,
7150 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7151 strict_aligned_p field.
7152 (nds32_expand_to_rtl_hook): New function.
7153 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7154 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7156 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7157 Chung-Ju Wu <jasonwucj@gmail.com>
7159 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7160 * config/nds32/nds32-n7.md: New file.
7161 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7162 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7164 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7165 * config/nds32/nds32.md (pipeline_model): Add n7.
7166 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7167 * config/nds32/pipelines.md: Include n7 settings.
7169 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7170 Chung-Ju Wu <jasonwucj@gmail.com>
7172 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7173 * config/nds32/nds32-e8.md: New file.
7174 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7175 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7177 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7178 * config/nds32/nds32.md (pipeline_model): Add e8.
7179 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7180 * config/nds32/pipelines.md: Include e8 settings.
7182 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7183 Chung-Ju Wu <jasonwucj@gmail.com>
7185 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7186 * config/nds32/nds32-n8.md: New file.
7187 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7188 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7190 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7191 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7192 * config/nds32/nds32.md (pipeline_model): Add n8.
7193 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7194 * config/nds32/pipelines.md: Include n8 settings.
7196 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7197 Chung-Ju Wu <jasonwucj@gmail.com>
7199 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7200 * config/nds32/nds32-n9-2r1w.md: New file.
7201 * config/nds32/nds32-n9-3r2w.md: New file.
7202 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7203 nds32_register_ports): New or modify for cpu n9.
7204 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7206 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7207 * config/nds32/nds32-utils.c: New file.
7208 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7209 TARGET_MUL_SLOW): Define.
7210 * config/nds32/nds32.md (pipeline_model): New attribute.
7211 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7212 New options that support cpu n9.
7213 * config/nds32/pipelines.md: Include n9 settings.
7214 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7216 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7218 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7219 information if necessary.
7220 (output_cond_branch_compare_zero): Likewise.
7221 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7222 (nds32_target_alignment): Refine for alignment.
7223 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7224 (FUNCTION_BOUNDARY): Modify.
7225 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7227 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7229 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7231 * config/nds32/constants.md (unspec_volatile_element): Add values for
7232 TLB operation and data prefetch.
7233 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7234 functions for TLB operation and data prefetch.
7235 * config/nds32/nds32-intrinsic.md: Likewise.
7236 * config/nds32/nds32_intrinsic.h: Likewise.
7237 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7238 (nds32_print_operand): Likewise.
7239 * config/nds32/nds32.h (nds32_builtins): Likewise.
7241 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7242 Andrew Pinski <pinsika@gcc.gnu.org>
7245 * match.pd: Use constant_boolean_node of correct type instead of
7246 boolean_true_node or boolean_false_node for simplifying
7247 pointer comparisons to zero.
7249 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7251 PR tree-optimization/80021
7252 * tree.c (verify_type_variant): Make error call in verify_variant_match
7253 translatable and remove final full stop.
7255 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7257 * config/nds32/constants.md (unspec_volatile_element): Add
7258 UNSPEC_VOLATILE_EH_RETURN.
7259 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7260 nds32_output_stack_pop): Support dwarf exception handling process.
7261 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7262 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7263 exception handling process.
7264 (nds32_compute_stack_frame): Likewise.
7265 (nds32_return_addr_rtx): Likewise.
7266 (nds32_initial_elimination_offset): Likewise.
7267 (nds32_expand_prologue): Likewise.
7268 (nds32_expand_epilogue): Likewise.
7269 (nds32_dynamic_chain_address): New function.
7270 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7272 (DYNAMIC_CHAIN_ADDRESS): Define.
7273 (EH_RETURN_DATA_REGNO): Define.
7274 (EH_RETURN_STACKADJ_RTX): Define.
7275 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7276 patterns for dwarf exception handling.
7278 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7280 * config/nds32/nds32.h: Clean up obsolete macros.
7282 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7284 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7285 Add enum values for particular instructions.
7286 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7287 particular intrinsic functions.
7288 * config/nds32/nds32-intrinsic.md: Likewise.
7289 * config/nds32/nds32_intrinsic.h: Likewise.
7290 * config/nds32/nds32.h (nds32_builtins): Likewise.
7291 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7292 (btst, ave): New patterns for particular instructions.
7294 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7296 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7297 Add enum values for atomic load/store and memory sync.
7298 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7300 * config/nds32/nds32-intrinsic.md: Likewise.
7301 * config/nds32/nds32_intrinsic.h: Likewise.
7302 * config/nds32/nds32.h (nds32_builtins): Likewise.
7304 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7306 PR tree-optimization/85257
7307 * fold-const.c (native_encode_vector): If not all elts could fit
7308 and off is -1, return 0 rather than offset.
7309 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7310 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7311 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7312 adjust buffer in native_interpret_expr call.
7314 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7316 * config/nds32/constants.md (unspec_volatile_element): Add cache
7317 control enum values.
7318 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7319 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7320 * config/nds32/nds32.c (nds32_cctl_names): New.
7321 (nds32_print_operand): Handle cache control register names.
7322 * config/nds32/nds32.h (nds32_builtins): New enum values.
7323 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7325 * config/nds32/nds32.md (type): Add mmu.
7326 * config/nds32/pipelines.md (simple_insn): Add mmu.
7328 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7330 * config/nds32/nds32.md (type): Remove call.
7331 * config/nds32/pipelines.md (simple_insn): Likewise.
7333 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7335 * config/nds32/constants.md (unspec_volatile_element): Add
7336 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7337 UNSPEC_VOLATILE_FMFCFG.
7338 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7339 description for fmfcfg and fmfcsr.
7340 (bdesc_1arg): Add fmtcsr.
7341 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7342 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7343 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7344 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7345 unspec_fmfcfg): New patterns.
7346 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7347 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7348 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7349 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7350 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7351 __nds32__fmfcfg): Define.
7353 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7355 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7356 intrinsic register names.
7357 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7358 intrinsic register enum values and macros.
7360 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7362 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7363 for load/store addressing form.
7364 (nds32_print_operand_address): Likewise.
7366 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7369 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7370 based on LABEL_REF. Remove useless assertion.
7371 (pic_address_needs_scratch): Fix formatting.
7372 (sparc_legitimize_pic_address): Minor tweaks.
7373 (sparc_delegitimize_address): Adjust assertion accordingly.
7374 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7375 into symbolic_operand.
7376 (movsi_high_pic_label_ref): Likewise.
7377 (movsi_lo_sum_pic_label_ref): Likewise.
7378 (movdi_pic_label_ref): Likewise.
7379 (movdi_high_pic_label_ref): Likewise.
7380 (movdi_lo_sum_pic_label_ref): Likewise.
7382 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7384 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7385 custom LIB_SPEC setup.
7387 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7388 Kito Cheng <kito.cheng@gmail.com>
7390 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7391 * config/riscv/freebsd.h: New.
7393 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7395 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7396 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7399 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7400 Kito Cheng <kito.cheng@gmail.com>
7402 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7403 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7404 * config/nds32/nds32-protos.h (nds32_output_call,
7405 nds32_output_return): Declare.
7406 * config/nds32/nds32.md: Refine all the call and return patterns.
7408 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7411 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7412 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7414 PR rtl-optimization/84872
7415 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7416 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7419 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7421 * expr.c (copy_blkmode_to_reg): Revert 254862.
7422 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7424 2018-04-06 Richard Biener <rguenther@suse.de>
7427 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7428 after seeing a component reference with an adjacent field. Treat
7429 refs to arrays at struct end of external decls similar to
7430 refs to unconstrained commons.
7432 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7435 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7436 look through SAVE_EXPRs with non-side-effects argument. Adjust
7438 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7441 2018-04-06 Richard Biener <rguenther@suse.de>
7444 * alias.c (find_base_term): New wrapper around find_base_term
7445 unwinding CSELIB_VAL_PTR changes.
7446 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7449 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7451 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7453 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7454 constant definitions.
7455 ("nop"): lr 0,0 -> nopr r0
7456 ("nop_lr0", "nop_lr1"): New insn definitions.
7458 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7460 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7461 NDS32_V3PUSH_AVAILABLE_P macro.
7463 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7464 Chung-Ju Wu <jasonwucj@gmail.com>
7466 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7467 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7468 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7469 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7470 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7471 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7472 * config/nds32/constraints.md: New constraints and checking for hard
7473 float configuration.
7474 * config/nds32/iterators.md: New mode iterator and attribute for hard
7475 float configuration.
7476 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7478 * config/nds32/nds32-fpu.md: New file.
7479 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7480 deal with hard float code generation.
7481 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7483 (abi_type, float_reg_number): New enum type.
7484 * config/nds32/nds32-predicates.c: New predicates for hard float.
7485 * config/nds32/nds32-protos.h: Declare functions for hard float.
7486 * config/nds32/nds32.c: Implementation for hard float configuration.
7487 * config/nds32/nds32.h: Definitions for hard float configuration.
7488 * config/nds32/nds32.md: Include hard float machine description and
7489 modify patterns for hard float configuration.
7490 * config/nds32/nds32.opt: New options for hard float configuration.
7491 * config/nds32/predicates.md: New predicates for hard float
7494 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7496 * common/config/nds32/nds32-common.c
7497 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7499 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7502 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7503 CONSTRUCTOR_ELT (ctor, ...)->value.
7505 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7508 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7510 2018-04-05 Tom de Vries <tom@codesourcery.com>
7513 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7516 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7517 Kito Cheng <kito.cheng@gmail.com>
7519 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7520 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7521 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7522 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7524 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7525 Kito Cheng <kito.cheng@gmail.com>
7527 * config/nds32/constraints.md (Ufe): New memory constraint.
7528 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7529 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7530 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7532 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7533 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7535 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7537 * config/nds32/nds32.md: Use optimize_size in the condition for
7538 alu-shift instructions.
7540 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7542 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7546 * config/nds32/nds32.md (negsi2): Refine pattern.
7548 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7549 Chung-Ju Wu <jasonwucj@gmail.com>
7551 * config/nds32/iterators.md (shift_rotate): New code iterator.
7552 (shift): New code attribute.
7553 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7554 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7555 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7556 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7557 bit-wise operations.
7558 (andsi3, *andsi3): Ditto.
7559 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7560 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7561 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7562 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7563 nds32_ior_operand, nds32_xor_operand): New predicates.
7565 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7567 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7568 (addsi3, subsi3): ... this.
7570 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7572 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7574 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7576 * config/nds32/nds32.md: Adjust indention.
7578 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7580 * config/nds32/nds32.md (feature): New attribute.
7582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7584 * config/nds32/nds32.md (subtype): New attribute.
7586 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7589 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7590 expansion to perform a bitwise AND of the argument followed by a
7591 boolean negation of the result.
7593 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7595 PR rtl-optimization/84878
7596 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7597 the basic block. Assert the use reference is not artificial and that
7598 it has an associated insn.
7600 2018-04-04 Michael Matz <matz@suse.de>
7602 * builtins.c (compute_objsize): Pass correct operand
7603 to array_at_struct_end_p.
7605 2018-04-04 Richard Biener <rguenther@suse.de>
7608 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7609 from contexts for DINFO_LEVEL_TERSE and below.
7611 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7613 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7614 resiter_operand condition.
7615 * config/nds32/nds32.md (*move<mode>): Ditto.
7617 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7618 Monk Chiang <sh.chiang04@gmail.com>
7620 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7622 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7624 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7626 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7627 Kito Cheng <kito.cheng@gmail.com>
7629 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7630 nds32_cond_code_str, output_cond_branch,
7631 output_cond_branch_compare_zero, nds32_expand_cbranch,
7632 nds32_expand_cstore, nds32_expand_movcc,
7633 nds32_output_cbranchsi4_equality_zero,
7634 nds32_output_cbranchsi4_equality_reg,
7635 nds32_output_cbranchsi4_equality_reg_or_const_int,
7636 nds32_output_cbranchsi4_greater_less_zero: New functions.
7637 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7638 nds32_expand_cstore, nds32_expand_movcc,
7639 nds32_output_cbranchsi4_equality_zero,
7640 nds32_output_cbranchsi4_equality_reg,
7641 nds32_output_cbranchsi4_equality_reg_or_const_int,
7642 nds32_output_cbranchsi4_greater_less_zero): Declare.
7643 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7644 nds32_rimm11s_operand): New predicates.
7645 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7646 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7649 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7651 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7652 * config/nds32/nds32.md: Ditto.
7653 * config/nds32/pipelines.md: Ditto.
7655 2018-04-04 Richard Biener <rguenther@suse.de>
7657 PR tree-optimization/85168
7658 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7659 propagating abnormals.
7661 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7663 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7665 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7666 Kito Cheng <kito.cheng@gmail.com>
7668 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7669 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7670 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7671 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7672 * config/nds32/nds32.md (sibcall_internal): New.
7673 (sibcall_register): Remove.
7674 (sibcall_immediate): Remove.
7675 (sibcall_value_internal): New.
7676 (sibcall_value_register): Remove.
7677 (sibcall_value_immediate): Remove.
7678 * config/nds32/predicates.md (nds32_general_register_operand): New.
7679 (nds32_call_address_operand): New.
7681 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7683 PR rtl-optimization/85167
7684 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7685 bb_defs if *split_p, instead preinitialize it to NULL.
7687 PR tree-optimization/85156
7688 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7689 evaluating the argument multiple times.
7691 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7693 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7695 (_mm_cvtpd_ps): Likewise.
7696 (_mm_cvttpd_epi32): Likewise.
7697 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7698 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7699 vector, pixel, and bool following altivec.h include.
7701 2018-04-03 Martin Sebor <msebor@redhat.com>
7703 * doc/extend.texi (Common Function Attributes): Clarify.
7704 (const attribute): Likewise.
7705 (pure attribute): Likewise.
7707 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7710 * config/i386/i386.c (ix86_expand_vector_set): Use
7711 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7713 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7715 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7716 instructions when changing rounding bits to preserve precision bits
7717 in the x87 control word.
7719 2018-04-03 Martin Liska <mliska@suse.cz>
7721 PR tree-optimization/82491
7722 * rtl.h (strip_offset_and_add): Replace += suboffset with
7723 poly_uint64 () + suboffset.
7725 2018-03-29 Martin Liska <mliska@suse.cz>
7726 Martin Jambor <mjambor@suse.cz>
7729 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7730 param_type is not an integral or pointer type.
7732 2018-04-03 Richard Biener <rguenther@suse.de>
7734 * sese.h (recompute_all_dominators): Remove.
7736 2018-04-02 Martin Sebor <msebor@redhat.com>
7738 * doc/invoke.texi (-Wrestrict): Fix typos.
7740 2018-04-02 Jim Wilson <jimw@sifive.com>
7742 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7743 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7744 (<optab>di3, <optab>si3_extend): Likewise.
7745 (<optab>si3_mask, <optab>si3_mask_1): New.
7746 (<optab>di3_mask, <optab>di3_mask_1): New.
7747 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7748 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7749 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7751 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7753 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7756 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7758 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7759 (nds32_canonicalize_comparison): New function.
7761 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7762 Kito Cheng <kito.cheng@gmail.com>
7763 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7765 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7766 * config/nds32/constants.md (unspec_volatile_element): Add
7767 UNSPEC_VOLATILE_RELAX_GROUP.
7768 * config/nds32/nds32-relax-opt.c: New file.
7769 * config/nds32/nds32-predicates.c
7770 (nds32_symbol_load_store_p): New function.
7771 * config/nds32/nds32-protos.h
7772 (nds32_symbol_load_store_p): Declare function.
7773 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7774 * config/nds32/nds32.c
7775 (nds32_register_pass): New function to register pass.
7776 (nds32_register_passes): New function to register passes.
7777 * config/nds32/nds32.md (relax_group): New pattern.
7778 * config/nds32/nds32.opt (mrelax-hint): New option.
7779 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7781 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7783 * config/nds32/t-nds32: Modify files dependency.
7785 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7787 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7788 (PROFILE_HOOK): Define its implementation.
7790 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7792 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7793 type and 32-bit size.
7795 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7798 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7799 (V_128_256): New mode iterator.
7800 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7801 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7802 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7804 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7805 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7807 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7810 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7811 NaN inputs correctly.
7813 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7816 * config/rs6000/vsx.md (??r): New mode attribute.
7817 (*vsx_mov<mode>_64bit): Use it.
7818 (*vsx_mov<mode>_32bit): Likewise.
7820 2018-03-30 Martin Sebor <msebor@redhat.com>
7822 PR tree-optimization/84818
7823 * builtins.c (check_access): Use warning_n.
7825 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7828 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7830 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7833 2018-03-30 Julia Koval <julia.koval@intel.com>
7836 * x86-tune.def (movx, partial_reg_dependency): Enable for
7839 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7842 * lra-constraints.c (process_alt_operands): Move setting
7843 this_alternative_matches below.
7845 2018-03-29 Martin Liska <mliska@suse.cz>
7848 * doc/invoke.texi: Document how LTO works with debug info.
7849 Describe auto-load support of binutils. Mention 'x86-64'
7850 as valid option value of -march option.
7852 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7854 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7857 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7858 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7859 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7862 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7865 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7866 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7867 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7868 for RS6000_BTM_POWERPC64.
7869 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7870 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7871 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7876 2018-03-28 Carl Love <cel@us.ibm.com>
7879 2017-09-27 Carl Love <cel@us.ibm.com>
7881 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7882 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7883 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7886 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7888 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7889 instead of __vector bool.
7890 (_mm_max_pu8): Likewise.
7891 (_mm_min_pi16): Likewise.
7893 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7896 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7897 (DIVWEUO): Likewise.
7899 (DIVDEUO): Likewise.
7900 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7901 DIVWEUO and DIVDEUO.
7902 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7903 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7904 (div_extend): Likewise.
7905 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7907 (__builtin_divweuo): Likewise.
7908 (__builtin_divdeo): Likewise.
7909 (__builtin_divdeuo): Likewise.
7911 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7914 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7915 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7917 PR tree-optimization/82004
7918 * gimple-match-head.c (optimize_pow_to_exp): New function.
7919 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7920 Don't fold to exp if optimize_pow_to_exp is false.
7922 2018-03-28 Martin Liska <mliska@suse.cz>
7925 * calls.c (initialize_argument_information): Fix trailing space.
7926 * common.opt: Fix typo and provide better explanation for
7927 -fsanitize-coverage option.
7928 * config/i386/i386.opt: Fix typo.
7930 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7931 Martin Liska <mliska@suse.cz>
7934 * gimplify.c (asan_poison_variable): Don't do the check for
7935 gimplify_omp_ctxp here.
7936 (gimplify_decl_expr): Do it here.
7937 (gimplify_target_expr): Likewise.
7939 2018-03-28 Martin Liska <mliska@suse.cz>
7942 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7943 chkp_type_bounds_count if MPX is not enabled.
7945 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7947 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7949 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7952 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7953 function to create the function decl for complex long double
7954 multiply and divide for -mabi=ieeelongdouble.
7955 (init_float128_ieee): Call it.
7957 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7960 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7961 -fcf-protection=branch -mibt.
7962 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7964 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7967 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7968 arm_word_relocations.
7970 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7973 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7974 extern array declarations.
7976 2018-03-27 Richard Biener <rguenther@suse.de>
7979 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7980 explicit single_use checks.
7982 2018-03-27 Richard Biener <rguenther@suse.de>
7984 PR tree-optimization/85082
7985 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7988 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7990 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7991 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7992 Turn on fasynchronous-unwind-tables and funwind-tables.
7994 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7997 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7998 (*bmi_blsr_<mode>_ccz): Ditto.
8000 2018-03-26 Tom de Vries <tom@codesourcery.com>
8002 PR tree-optimization/85063
8003 * omp-general.c (offloading_function_p): New function. Factor out
8005 * omp-offload.c (pass_omp_target_link::gate): ... here.
8006 * omp-general.h (offloading_function_p): Declare.
8007 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8008 with attribute omp declare target for offloading functions.
8010 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8012 PR tree-optimization/84005
8013 * tree-data-ref.h (get_base_for_alignment): Declare.
8014 * tree-data-ref.c (get_base_for_alignment_1): New function.
8015 (get_base_for_alignment): Likewise.
8016 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8017 get_base_for_alignment to find a suitable base object, instead
8018 of always using drb->base_address.
8020 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8023 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8024 known size by default.
8026 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8029 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8030 and non BLKmode operands.
8032 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8035 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8036 Clean up attributes.
8038 2018-03-23 Richard Biener <rguenther@suse.de>
8041 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8042 we are going to emit early debug for LTO.
8044 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8047 * function.c (match_asm_constraints_1): Don't optimize if input
8048 doesn't satisfy general_operand predicate for output's mode.
8051 * alias.c (write_dependence_p): Don't require for x_canonicalized
8052 non-VOIDmode if x has VOIDmode.
8055 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8056 just don't try to optimize it rather than assert it never happens.
8058 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8060 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8061 macro expansions for definition of ST_INTERNAL_<mode> and
8062 LD_INTERNAL_<mode> builtins.
8063 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8065 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8067 (altivec_expand_st_builtin): Likewise.
8068 (altivec_expand_builtin): Remove calls to deleted functions.
8069 (rs6000_address_for_altivec): Delete this function.
8070 * config/rs6000/vector.md: Remove expands for
8071 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8073 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8076 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8077 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8078 re-computing once computed.
8079 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8080 (arm_init_machine_status): Initialize
8081 machine->static_chain_stack_bytes.
8083 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8086 * doc/extend.texi: Add four new prototypes for vec_ld.
8087 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8088 definitions for more logical presentation.
8089 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8090 entries for V1TI variants of __builtin_altivec_ld builtin.
8091 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8092 handling of V1TI variant of LVX icode pattern.
8093 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8094 (rs6000_gimple_fold_builtin): Likewise.
8095 (altivec_init_builtins): Add code to define
8096 __builtin_altivec_lvx_v1ti function.
8098 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8101 * function.c (match_asm_constraints_1): Don't do the optimization
8102 if input isn't a REG, SUBREG, MEM or constant.
8104 2018-03-22 Tom de Vries <tom@codesourcery.com>
8106 PR tree-optimization/84956
8107 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8108 bb_has_abnormal_pred.
8110 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8113 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8114 DECL_INITIAL (decl) to decl at the end.
8115 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8118 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8120 * doc/extend.texi (__builtin_tgmath): Document when complex
8121 integer types are treated as _Complex _Float64.
8123 2018-03-21 Tom de Vries <tom@codesourcery.com>
8125 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8127 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8129 PR tree-optimization/84960
8130 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8131 if it is ENTRY block, move them into single succ of ENTRY in that case.
8133 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8135 PR tree-optimization/84811
8136 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8137 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8138 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8139 (known_subrange_p): Update accordingly. Cast each value involved
8140 in the size comparison, rather than casting the result of the
8143 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8145 PR tree-optimization/84982
8146 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8147 by flipping the least significant bit rather than all bits from
8148 bitpos to bitpos + bitsize - 1.
8150 2018-03-21 Nathan Sidwell <nathan@acm.org>
8152 * doc/extend.texi (Deprecated Features): Remove mention of
8153 long-deleted deprecations.
8155 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8158 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8159 * configure: Regenerate.
8161 2018-03-21 Tom de Vries <tom@codesourcery.com>
8163 PR tree-optimization/83126
8164 * tree-parloops.c (num_phis): New function.
8165 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8167 2018-03-21 Nathan Sidwell <nathan@acm.org>
8169 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8170 mention anon-struct/union members and trailing attributes.
8172 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8174 PR tree-optimization/84969
8175 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8176 builtin memset partitions if they set different rhs values.
8178 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8180 PR rtl-optimization/84989
8181 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8182 VEC_DUPLICATE with scalar result mode.
8184 2018-03-21 Martin Liska <mliska@suse.cz>
8187 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8188 not intended return statement.
8190 2018-03-21 Martin Liska <mliska@suse.cz>
8193 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8194 (chkp_find_bound_slots_1): Limit number of iterations.
8196 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8199 * Minor grammar fixes for x86 options.
8201 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8204 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8205 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8207 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8210 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8211 (altivec_lvx_<mode>_1op): Likewise.
8212 (altivec_stvx_<mode>_2op): Likewise.
8213 (altivec_stvx_<mode>_1op): Likewise.
8214 (altivec_lvx_<VM2:mode>): New define_expand.
8215 (altivec_stvx_<VM2:mode>): Likewise.
8216 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8217 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8218 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8219 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8220 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8221 (rs6000_gen_lvx): Likewise.
8222 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8223 (altivec_expand_stv_builtin): Likewise.
8224 (altivec_expand_builtin): Likewise.
8225 * config/rs6000/vector.md: Likewise.
8227 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8230 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8233 2018-03-20 Richard Biener <rguenther@suse.de>
8236 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8237 sign-conversions as zero, fall back to standard scalar_stmt
8240 2018-03-20 Martin Liska <mliska@suse.cz>
8243 * predict.c (rebuild_frequencies): Handle case when we have
8244 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8246 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8249 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8250 flag_section_anchors.
8251 * varasm.c (use_blocks_for_decl_p): Remove hack for
8252 dw2_force_const_mem.
8255 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8257 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8258 be created, use lowpart_subreg of operands[0] rather than operands[0]
8260 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8261 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8262 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8263 and n constraint instead of aarch64_shift_imm_di and Usd.
8264 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8265 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8267 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8270 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8271 to favor GPR over NEON registers.
8272 (<shift>di3_neon): Likewise.
8274 2018-03-20 Tom de Vries <tom@codesourcery.com>
8277 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8278 (nvptx_process_pars): Emit bar.sync asap and alap.
8280 2018-03-20 Tom de Vries <tom@codesourcery.com>
8283 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8284 seen_label if seen_label is already set.
8286 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8289 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8290 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8291 Use 1U instead of 1. Formatting fixes.
8294 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8295 instead of TREE_TYPE (s1) for the return value.
8297 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8299 PR tree-optimization/84946
8300 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8301 bitsize + bitsize in poly_uint64 rather than poly_int64.
8304 * dwarf2asm.c: Include fold-const.c.
8305 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8306 of decl rather than decl itself.
8308 PR rtl-optimization/84643
8309 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8311 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8314 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8315 calling assemble_variable.
8317 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8320 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8321 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8323 2018-03-19 Jim Wilson <jimw@sifive.com>
8326 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8327 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8328 (riscv_first_stack_step): Likewise.
8329 (riscv_option_override): Use STACK_BOUNDARY instead of
8331 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8333 (BIGGEST_ALIGNMENT): Set to 128.
8334 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8335 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8338 2018-03-19 Richard Biener <rguenther@suse.de>
8340 PR tree-optimization/84933
8341 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8342 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8344 2018-03-19 Richard Biener <rguenther@suse.de>
8346 PR tree-optimization/84859
8347 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8348 (cond_if_else_store_replacement): Perform sinking operation on
8349 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8350 Generalize what a BB with a single eligible store is.
8352 2018-03-19 Richard Biener <rguenther@suse.de>
8354 PR tree-optimization/84929
8355 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8356 chrec_is_positive against non-chrec arg.
8358 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8361 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8363 2018-03-18 Martin Liska <mliska@suse.cz>
8365 PR rtl-optimization/84635
8366 * regrename.c (build_def_use): Use matches_mode only when
8369 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8371 PR tree-optimization/84913
8372 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8373 vectorize chains of COND_EXPRs.
8375 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8377 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8379 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8381 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8383 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8385 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8387 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8388 Kito Cheng <kito.cheng@gmail.com>
8390 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8391 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8392 (nds32_adjust_reg_alloc_order): New function.
8393 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8395 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8397 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8398 nds32_print_operand, nds32_print_operand_address): Use
8399 HOST_WIDE_INT_PRINT_DEC instead.
8401 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8403 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8405 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8408 * config/i386/i386.c (initial_ix86_tune_features,
8409 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8411 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8412 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8413 rather than 1u << ix86_tune. Formatting fix.
8414 (ix86_option_override_internal): Change ix86_arch_mask from
8415 unsigned int to unsigned HOST_WIDE_INT, initialize to
8416 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8417 (ix86_function_specific_restore): Likewise.
8419 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8422 * postreload.c (reload_combine_recognize_pattern): Perform
8423 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8424 truncate_int_for_mode the result for the destination's mode.
8427 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8428 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8430 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8433 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8436 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8437 regno_allocno_class_array and sorted_pseudos.
8438 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8439 insns where regno is used.
8441 2018-03-16 Martin Liska <mliska@suse.cz>
8444 * multiple_target.c (create_dispatcher_calls): Redirect
8445 reference in the symbol table.
8447 2018-03-16 Martin Liska <mliska@suse.cz>
8450 * multiple_target.c (create_dispatcher_calls): Redirect also
8453 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8457 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8458 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8459 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8461 2018-03-16 Julia Koval <julia.koval@intel.com>
8463 * doc/invoke.texi (Skylake Server): Add CLWB.
8464 Cannonlake): Remove CLWB.
8466 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8468 PR tree-optimization/84841
8469 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8471 (FLOAT_ONE_CONST_TYPE): Define.
8472 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8473 (sort_by_operand_rank): Put entries with higher constant_type last
8474 rather than first to match comments.
8476 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8478 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8481 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8484 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8485 check and use address of target always.
8487 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8490 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8491 (indirect_thunk_bnd_needed): Likewise.
8492 (indirect_thunks_used): Likewise.
8493 (indirect_thunks_bnd_used): Likewise.
8494 (indirect_return_needed): New.
8495 (indirect_return_bnd_needed): Likewise.
8496 (output_indirect_thunk_function): Add a bool argument for
8498 (output_indirect_thunk_function): Don't generate alias for
8499 function return thunk.
8500 (ix86_code_end): Call output_indirect_thunk_function to generate
8501 function return thunks.
8502 (ix86_output_function_return): Set indirect_return_bnd_needed
8503 and indirect_return_needed instead of indirect_thunk_bnd_needed
8504 and indirect_thunk_needed.
8506 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8508 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8509 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8510 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8512 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8513 Paul Hua <paul.hua.gm@gmail.com>
8516 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8518 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8520 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8521 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8524 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8527 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8528 instead of GET_MODE_SIZE when comparing Units.
8530 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8533 * varasm.c (hash_section): Return an unchangeble hash value
8534 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8535 Return !aarch64_can_use_per_function_literal_pools_p ().
8537 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8540 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8541 rather than address of cmode as last argument to prepare_cmp_insn.
8543 2018-03-15 Julia Koval <julia.koval@intel.com>
8545 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8546 F_AVX512VNNI, F_AVX512BITALG): New.
8548 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8551 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8552 insn for floating-point loads and stores.
8554 2018-03-14 Carl Love <cel@us.ibm.com>
8556 * config/rs6000/rs6000-c.c: Add macro definitions for
8557 ALTIVEC_BUILTIN_VEC_PERMXOR.
8558 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8559 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8560 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8561 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8563 * config/doc/extend.texi: Add prototypes for vec_permxor.
8565 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8568 * diagnostic-show-locus.c (class layout_point): Convert m_line
8569 from int to linenum_type.
8570 (line_span::comparator): Use linenum "compare" function when
8571 comparing line numbers.
8572 (test_line_span): New function.
8573 (layout_range::contains_point): Convert param "row" from int to
8575 (layout_range::intersects_line_p): Likewise.
8576 (layout::will_show_line_p): Likewise.
8577 (layout::print_source_line): Likewise.
8578 (layout::should_print_annotation_line_p): Likewise.
8579 (layout::print_annotation_line): Likewise.
8580 (layout::print_leading_fixits): Likewise.
8581 (layout::annotation_line_showed_range_p): Likewise.
8582 (struct line_corrections): Likewise for field m_row.
8583 (line_corrections::line_corrections): Likewise for param "row".
8584 (layout::print_trailing_fixits): Likewise.
8585 (layout::get_state_at_point): Likewise.
8586 (layout::get_x_bound_for_row): Likewise.
8587 (layout::print_line): Likewise.
8588 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8589 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8590 * input.c (selftest::test_linenum_comparisons): New function.
8591 (selftest::input_c_tests): Call it.
8592 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8593 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8594 * selftest.h (ASSERT_GT): New macro.
8595 (ASSERT_GT_AT): New macro.
8596 (ASSERT_LT): New macro.
8597 (ASSERT_LT_AT): New macro.
8599 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8601 PR rtl-optimization/84780
8602 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8604 2018-03-14 Martin Liska <mliska@suse.cz>
8606 * tree.c (record_node_allocation_statistics): Use
8607 get_stats_node_kind.
8608 (get_stats_node_kind): New function extracted from
8609 record_node_allocation_statistics.
8610 (free_node): Use get_stats_node_kind.
8612 2018-03-14 Richard Biener <rguenther@suse.de>
8614 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8615 that the value-set of ANTIC_IN doesn't grow.
8618 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8620 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8621 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8623 2018-03-14 Julia Koval <julia.koval@intel.com>
8625 * config.gcc (icelake-client, icelake-server): New.
8627 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8628 (initial_ix86_arch_features): Ditto.
8629 (PTA_SKYLAKE): Add SGX.
8630 (PTA_ICELAKE): Remove.
8631 (PTA_ICELAKE_CLIENT): New.
8632 (PTA_ICELAKE_SERVER): New.
8633 (ix86_option_override_internal): Split up icelake on icelake client and
8635 (get_builtin_code_for_version): Ditto.
8636 (fold_builtin_cpu): Ditto.
8637 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8638 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8639 * config/i386/i386.h (processor_type): Ditto.
8640 * doc/invoke.texi: Ditto.
8642 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8645 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8646 INTEGER_CST offset, add it together with bitpos / 8 and
8647 sign extend based on POINTER_SIZE.
8651 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8654 * config/i386/constraints.md (Yc): New register constraint.
8655 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8656 Use Yc constraint for alternative 2 of operand 0. Remove
8657 preferred_for_speed attribute.
8659 2018-03-14 Richard Biener <rguenther@suse.de>
8661 PR tree-optimization/84830
8662 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8663 with the old one to avoid oscillations.
8665 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8668 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8670 (assign_by_spills): Return a flag of reload assignment failure.
8671 Do not process the reload assignment failures. Do not spill other
8672 reload pseudos if they has the same reg class. Update n if
8674 (lra_assign): Add a return arg. Set up from the result of
8675 assign_by_spills call.
8676 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8677 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8678 usage_insns if it is not NULL.
8679 (spill_hard_reg_in_range): New function.
8680 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8681 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8682 function prototypes.
8683 (lra_assign): Change prototype.
8684 * lra.c (lra): Add code to deal with fails by splitting hard reg
8687 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8689 * config/riscv/riscv.opt (mrelax): New option.
8690 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8691 "norelax" when riscv_mrelax is disabled.
8692 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8694 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8697 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8698 reassociation for int modes.
8700 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8702 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8703 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8705 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8706 * config/aarch64/aarch64-sve.md
8707 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8708 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8709 (*extend<mode><Vwide>2): Rename to...
8710 (aarch64_sve_extend<mode><Vwide>2): ...this.
8711 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8712 renaming the old pattern to...
8713 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8715 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8716 define_expand, renaming the old pattern to...
8717 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8718 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8719 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8720 account when deciding which SVE instruction the optab should use.
8721 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8723 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8725 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8726 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8727 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8728 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8729 (tlsdesc_small_<mode>): Turn a define_expand and use
8730 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8731 (tlsdesc_small_advsimd_<mode>): ...this.
8732 (tlsdesc_small_sve_<mode>): New pattern.
8734 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8736 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8737 (UNSPEC_UMUL_HIGHPART): New constants.
8738 (MUL_HIGHPART): New int iteraor.
8739 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8740 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8742 (*<su>mul<mode>3_highpart): New define_insn.
8744 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8747 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8750 2018-03-13 Martin Liska <mliska@suse.cz>
8753 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8755 (sem_item_optimizer::~sem_item_optimizer): Release it.
8756 (sem_item_optimizer::merge_classes): Register variable aliases.
8757 (sem_item_optimizer::fixup_pt_set): New function.
8758 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8759 * ipa-icf.h: Declare new variables and functions.
8761 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8764 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8765 integer_pow2p@2 and test integer_pow2p in condition.
8766 (A < 0 ? C : 0): Similarly for @1.
8769 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8770 characters starting at p contain '\0' character, don't look beyond
8774 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8775 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8778 * reg-stack.c (change_stack): Change update_end var from int to
8779 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8780 also call set_block_for_insn on the newly added insns and rescan.
8783 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8784 on the last operand.
8787 * tree.c (stabilize_reference_1): Return save_expr (e) for
8788 STATEMENT_LIST even if it doesn't have side-effects.
8790 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8792 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8794 2018-03-12 Renlin Li <renlin.li@arm.com>
8796 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8797 aarch64_output_scalar_simd_mov_immediate.
8799 2018-03-12 Martin Sebor <msebor@redhat.com>
8801 PR tree-optimization/83456
8802 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8803 for perfectly overlapping calls to memcpy.
8804 (gimple_fold_builtin_memory_chk): Same.
8805 (gimple_fold_builtin_strcpy): Handle no-warning.
8806 (gimple_fold_builtin_stxcpy_chk): Same.
8807 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8809 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8811 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8812 parameter. Use it for SFmode.
8813 (rs6000_function_arg_advance_1): Adjust.
8814 (rs6000_function_arg): Adjust.
8815 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8817 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8819 PR rtl-optimization/84169
8820 PR rtl-optimization/84780
8821 * combine.c (can_combine_p): Check for a 2-insn combination whether
8822 the destination register is used between the two insns, too.
8824 2018-03-12 Richard Biener <rguenther@suse.de>
8826 PR tree-optimization/84803
8827 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8828 for refs DR analysis didn't process.
8830 2018-03-12 Richard Biener <rguenther@suse.de>
8832 PR tree-optimization/84777
8833 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8834 force-vectorize loops ignore whether we are optimizing for size.
8836 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8838 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8839 (TARGET_MD_ASM_ADJUST): Define.
8841 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8842 Kito Cheng <kito.cheng@gmail.com>
8843 Chung-Ju Wu <jasonwucj@gmail.com>
8845 * config/nds32/nds32.c (nds32_compute_stack_frame,
8846 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8847 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8848 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8849 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8850 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8851 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8852 * config/nds32/nds32.md (prologue, epilogue): Use macro
8853 NDS32_V3PUSH_AVAILABLE_P to do checking.
8855 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8858 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8859 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8860 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8861 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8862 addition of most attributes on !orig_type_die or the attribute not
8863 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8865 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8866 Chung-Ju Wu <jasonwucj@gmail.com>
8868 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8870 * config/nds32/nds32.opt (mvh): New option.
8872 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8873 Chung-Ju Wu <jasonwucj@gmail.com>
8875 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8877 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8878 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8881 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8882 Chung-Ju Wu <jasonwucj@gmail.com>
8884 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8886 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8887 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8889 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8890 Kito Cheng <kito.cheng@gmail.com>
8891 Chung-Ju Wu <jasonwucj@gmail.com>
8893 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8894 UNSPEC_FFMISM and UNSPEC_FLMISM.
8895 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8896 for ffb, ffmism and flmism.
8897 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8898 (unspec_ffmism): Ditto.
8899 (unspec_flmism): Ditto.
8900 (nds32_expand_builtin_impl): Check if string extension is available.
8901 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8902 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8904 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8907 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8910 * lra-assigns.c (assign_by_spills): Return a flag of reload
8911 assignment failure. Do not process the reload assignment
8912 failures. Do not spill other reload pseudos if they has the same
8914 (lra_assign): Add a return arg. Set up from the result of
8915 assign_by_spills call.
8916 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8917 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8918 usage_insns if it is not NULL.
8919 (spill_hard_reg_in_range): New function.
8920 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8921 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8922 function prototypes.
8923 (lra_assign): Change prototype.
8924 * lra.c (lra): Add code to deal with fails by splitting hard reg
8927 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8930 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8932 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8935 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8936 (dw_val_node): Add val_symbolic_view.
8937 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8938 (symview_upper_bound): New.
8939 (new_line_info_table): Initialize symviews_since_reset.
8940 (dwarf2out_source_line): Count symviews_since_reset and set
8941 symview_upper_bound.
8942 (dw_val_equal_p): Handle symview.
8943 (add_AT_symview): New.
8944 (print_dw_val): Handle symview.
8945 (attr_checksum, attr_checksum_ordered): Likewise.
8946 (same_dw_val_p, size_of_die): Likewise.
8947 (value_format, output_die): Likewise.
8948 (add_high_low_attributes): Use add_AT_symview for entry_view.
8949 (dwarf2out_finish): Reset symview_upper_bound, clear
8952 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8955 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8956 Add strict argument and use it.
8957 (rs6000_split_multireg_move): Update for new strict argument.
8958 (mem_operand_gpr): Disallow all non-offsettable addresses.
8959 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8961 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8964 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8965 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8966 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8969 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8970 decl, use remap_type if we want to use the type.
8972 2018-03-09 Martin Sebor <msebor@redhat.com>
8974 PR tree-optimization/84526
8975 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8977 (builtin_access::generic_overlap): Be prepared to handle non-array
8980 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8982 PR rtl-optimization/84682
8983 * lra-constraints.c (process_address_1): Check is_address flag
8984 for address constraints.
8985 (process_alt_operands): Likewise.
8986 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8987 preprocess_constraints.
8988 * recog.h (preprocess_constraints): Add oploc parameter.
8990 * recog.c (preprocess_constraints): Test address_operand for
8991 CT_ADDRESS constraints.
8993 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8996 * lra-assigns.c (assign_by_spills): Return a flag of reload
8997 assignment failure. Do not process the reload assignment
8998 failures. Do not spill other reload pseudos if they has the same
9000 (lra_assign): Add a return arg. Set up from the result of
9001 assign_by_spills call.
9002 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9003 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9004 usage_insns if it is not NULL.
9005 (spill_hard_reg_in_range): New function.
9006 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9007 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9008 function prototypes.
9009 (lra_assign): Change prototype.
9010 * lra.c (lra): Add code to deal with fails by splitting hard reg
9013 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9016 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9017 Accept complain bool parameter. Only emit errors if it is true.
9018 (arm_parse_cpu_option_name): Likewise.
9019 (arm_target_thumb_only): Adjust callers of the above.
9020 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9021 prototype to take a default true bool parameter.
9022 (arm_parse_arch_option_name): Likewise.
9024 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9025 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9029 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9030 * configure: Regenerate.
9031 * configure.ac ("linker --version-script option"): New.
9032 ("linker soname option"): New.
9034 2018-03-09 Richard Biener <rguenther@suse.de>
9036 PR tree-optimization/84775
9037 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9038 immediate uses of predicate stmts and mark them modified.
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 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9051 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9052 when the function accesses prior frames.
9054 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9057 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9058 gen_llsym, otherwise call maybe_gen_llsym.
9061 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9062 has ',' character inside of it.
9064 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9067 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9068 as clobbering CC_REGNUM.
9070 2018-03-08 Richard Biener <rguenther@suse.de>
9073 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9074 (follow_copies_to_constant): Do not follow SSA names registered
9077 2018-03-08 Richard Biener <rguenther@suse.de>
9079 PR tree-optimization/84178
9080 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9082 (version_loop_for_if_conversion): Delay update_ssa call.
9083 (tree_if_conversion): Delay update_ssa until after predicate
9086 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9088 PR tree-optimization/84178
9089 * tree-if-conv.c (release_bb_predicate): Remove the
9090 the assertion that the stmts have NULL use_ops.
9091 Discard the statements, asserting that they haven't
9092 yet been added to a BB.
9094 2018-03-08 Richard Biener <rguenther@suse.de>
9096 PR tree-optimization/84746
9097 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9098 (phi_translate): Pass in destination ANTIC_OUT set.
9099 (phi_translate_1): Likewise. For a simplified result lookup
9100 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9101 (phi_translate_set): Adjust.
9102 (do_pre_regular_insertion): Likewise.
9103 (do_pre_partial_partial_insertion): Likewise.
9105 2018-03-08 Martin Liska <mliska@suse.cz>
9107 PR gcov-profile/84735
9108 * doc/gcov.texi: Document usage of profile files.
9109 * gcov-io.h: Document changes in the format.
9111 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9115 * dwarf2out.c (struct dw_line_info_table): Update comments for
9117 (FORCE_RESET_NEXT_VIEW): New.
9118 (FORCE_RESETTING_VIEW_P): New.
9119 (RESETTING_VIEW_P): Check for -1 too.
9120 (ZERO_VIEW_P): Likewise.
9121 (new_line_info_table): Force-reset next view.
9122 (dwarf2out_begin_function): Likewise.
9123 (dwarf2out_source_line): Simplify zero_view_p initialization.
9124 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9125 view directly. Omit view when omitting .loc at line 0.
9127 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9129 PR tree-optimization/84740
9130 * tree-switch-conversion.c (process_switch): Call build_constructors
9131 only if info.phi_count is non-zero.
9133 PR tree-optimization/84739
9134 * tree-tailcall.c (find_tail_calls): Check call arguments against
9135 DECL_ARGUMENTS (current_function_decl) rather than
9136 DECL_ARGUMENTS (func) when checking for tail recursion.
9138 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9140 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9141 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9142 Volker Reichelt's entry and add entries for people that perform
9143 GCC fuzzy testing and report numerous bugs.
9145 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9148 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9149 readonly data in sdata, if that is disabled.
9150 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9151 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9152 -mreadonly-in-sdata option.
9154 2018-03-07 Martin Sebor <msebor@redhat.com>
9156 PR tree-optimization/84468
9157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9158 basic block when looking for nul assignment.
9160 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9163 * except.h (output_function_exception_table): Adjust prototype.
9164 * except.c (output_function_exception_table): Remove FNNAME parameter
9165 and add SECTION parameter. Ouput one part of the table at a time.
9166 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9167 the first part of the exception table and emit unwind directives.
9168 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9169 (i386_pe_seh_cold_init): Likewise.
9170 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9171 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9172 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9173 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9174 * config/i386/winnt.c: Include except.h.
9175 (struct seh_frame_state): Add reg_offset, after_prologue and
9176 in_cold_section fields.
9177 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9178 (i386_pe_seh_cold_init): New function.
9179 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9180 to seh->in_cold_section.
9181 (seh_emit_push): Record the offset of the push.
9182 (seh_emit_save): Record the offet of the save.
9183 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9184 Test seh->after_prologue to disregard the epilogue.
9185 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9186 (i386_pe_end_cold_function): New function.
9188 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9191 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9192 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9195 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9197 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9198 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9201 * multiple_target.c: Include tree-inline.h and intl.h.
9202 (expand_target_clones): Diagnose and fail if node->definition and
9203 !tree_versionable_function_p (node->decl).
9205 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9207 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9209 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9210 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9211 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9213 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9216 * combine.c (try_combine): Use reg_or_subregno instead of handling
9217 just paradoxical SUBREGs and REGs.
9219 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9221 * config/arc/arc.c (arc_finalize_pic): Remove function.
9222 (arc_must_save_register): We use single base PIC register, remove
9223 checks to save/restore the PIC register.
9224 (arc_expand_prologue): Likewise.
9225 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9227 (arc_verify_short): Likewise.
9228 (arc_attr_type): Likewise.
9229 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9230 (walk_stores): Likewise.
9231 (arc_address_cost): Make it static.
9232 (arc_verify_short): Likewise.
9233 (branch_dest): Likewise.
9234 (arc_attr_type): Likewise.
9235 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9236 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9237 (arc_final_prescan_insn): Remove inserting the nops due to
9238 hardware hazards. It is done in reorg step.
9239 (insn_length_variant_t): Remove.
9240 (insn_length_parameters_t): Likewise.
9241 (arc_insn_length_parameters): Likewise.
9242 (arc_get_insn_variants): Likewise.
9243 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9245 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9248 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9251 PR tree-optimization/84687
9252 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9254 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9256 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9258 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9259 Rename to ppc_speculation_barrier.
9260 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9261 __builtin_ppc_speculation_barrier.
9263 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9266 * combine.c (combine_simplify_rtx): Don't try to simplify if
9267 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9270 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9272 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9273 to 32 bytes when compiling for POWER9.
9275 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9278 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9279 regparm >= 3 with no arg reg available also for calls with
9280 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9283 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9285 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9287 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9290 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9292 2018-03-05 Richard Biener <rguenther@suse.de>
9294 PR tree-optimization/84486
9295 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9296 When inserting a __builtin_assume_aligned call set the LHS
9297 SSA name alignment info accordingly.
9299 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9301 PR tree-optimization/84114
9302 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9303 Avoid reassociation of FLOAT_MODE addition.
9305 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9307 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9308 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9309 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9310 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9311 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9312 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9313 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9315 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9316 __builtin_ia32_wbinvd): New builtins.
9317 (SPECIAL_ARGS2): New.
9318 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9319 (SPECIAL_ARGS2): New.
9320 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9321 (ix86_valid_target_attribute_inner_p): Ditto.
9322 (ix86_init_mmx_sse_builtins): Add special_args2.
9323 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9324 TARGET_WBNOINVD_P): New.
9325 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9326 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9327 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9328 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9329 * config/i386/pconfigintrin.h: New file.
9330 * config/i386/wbnoinvdintrin.h: Ditto.
9331 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9333 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9335 2018-03-05 Richard Biener <rguenther@suse.de>
9337 PR tree-optimization/84670
9338 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9340 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9341 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9342 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9343 doesn't grow if all successors (recursively) were visited at least
9346 2018-03-05 Richard Biener <rguenther@suse.de>
9348 PR tree-optimization/84650
9349 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9350 if executed in the loop pipeline.
9352 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9354 * doc/configfiles.texi (Configuration Files): Move info about
9355 conditionalizing $target-protos.h to...
9356 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9357 differs from $target-protos.h.
9359 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9360 Chung-Ju Wu <jasonwucj@gmail.com>
9362 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9363 * config/nds32/nds32-multiple.md (setmemsi): Define.
9364 * config/nds32/nds32-memory-manipulation.c
9365 (nds32_gen_dup_4_byte_to_word_value): New.
9366 (emit_setmem_word_loop): New.
9367 (emit_setmem_byte_loop): New.
9368 (nds32_expand_setmem_loop): New.
9369 (nds32_expand_setmem_loop_v3m): New.
9370 (nds32_expand_setmem_unroll): New.
9371 (nds32_expand_setmem): New.
9373 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9374 Chung-Ju Wu <jasonwucj@gmail.com>
9376 * config/nds32/nds32-memory-manipulation.c
9377 (nds32_emit_load_store): New.
9378 (nds32_emit_post_inc_load_store): New.
9379 (nds32_emit_mem_move): New.
9380 (nds32_emit_mem_move_block): New.
9381 (nds32_expand_movmemsi_loop_unknown_size): New.
9382 (nds32_expand_movmemsi_loop_known_size): New.
9383 (nds32_expand_movmemsi_loop): New.
9384 (nds32_expand_movmemsi_unroll): New.
9385 (nds32_expand_movmemqi): Rename ...
9386 (nds32_expand_movmemsi): ... to this.
9387 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9388 (movmemsi): ... to this.
9389 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9390 (nds32_expand_movmemsi): ... to this.
9392 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9393 Monk Chiang <sh.chiang04@gmail.com>
9394 Chung-Ju Wu <jasonwucj@gmail.com>
9396 * config/nds32/nds32-protos.h
9397 (nds32_expand_load_multiple): New arguments.
9398 (nds32_expand_store_multiple): Ditto.
9399 (nds32_valid_multiple_load_store): Rename ...
9400 (nds32_valid_multiple_load_store_p): ... to this.
9401 * config/nds32/nds32-memory-manipulation.c
9402 (nds32_expand_load_multiple): Refine implementation.
9403 (nds32_expand_store_multiple): Ditto.
9404 * config/nds32/nds32-multiple.md
9405 (load_multiple): Update nds32_expand_load_multiple interface.
9406 (store_multiple): Update nds32_expand_store_multiple interface.
9407 * config/nds32/nds32-predicates.c
9408 (nds32_valid_multiple_load_store): Rename ...
9409 (nds32_valid_multiple_load_store_p): ... to this and refine
9411 * config/nds32/predicates.md
9412 (nds32_load_multiple_and_update_address_operation): New predicate.
9413 (nds32_store_multiple_and_update_address_operation): New predicate.
9415 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9416 Chung-Ju Wu <jasonwucj@gmail.com>
9418 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9419 (combo): New attribute.
9420 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9422 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9424 * config/nds32/nds32.opt: Change -mcmodel= default value.
9426 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9427 Monk Chiang <sh.chiang04@gmail.com>
9428 Chung-Ju Wu <jasonwucj@gmail.com>
9430 * config/nds32/constants.md (unspec_element): New enum.
9431 * config/nds32/constraints.md (Umw): New constraint.
9432 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9433 * config/nds32/nds32-intrinsic.md: Likewise.
9434 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9435 (nds32_valid_smw_lwm_base_p): New.
9436 (nds32_output_smw_single_word): New.
9437 (nds32_output_lmw_single_word): New.
9438 (nds32_expand_unaligned_load): New.
9439 (nds32_expand_unaligned_store): New.
9440 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9441 (nds32_output_smw_single_word): Declare.
9442 (nds32_output_lmw_single_word): Declare.
9443 (nds32_expand_unaligned_load): Declare.
9444 (nds32_expand_unaligned_store): Declare.
9445 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9446 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9447 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9448 NDS32_BUILTIN_UASTORE_DW.
9449 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9452 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9453 Kito Cheng <kito.cheng@gmail.com>
9454 Chung-Ju Wu <jasonwucj@gmail.com>
9456 * config/nds32/nds32-intrinsic.c
9457 (nds32_expand_builtin_null_ftype_reg): Delete.
9458 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9459 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9460 (nds32_read_argument): New.
9461 (nds32_legitimize_target): Ditto.
9462 (nds32_legitimize_argument): Ditto.
9463 (nds32_check_constant_argument): Ditto.
9464 (nds32_expand_unop_builtin): Ditto.
9465 (nds32_expand_unopimm_builtin): Ditto.
9466 (nds32_expand_binop_builtin): Ditto.
9467 (nds32_builtin_decl_impl): Ditto.
9468 (builtin_description): Ditto.
9469 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9470 (nds32_init_builtins_impl): Ditto.
9471 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9472 (nds32_builtin_decl): New.
9473 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9474 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9476 2018-03-02 Jeff Law <law@redhat.com>
9478 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9479 (redundant_insn, fill_simple_delay_slots): Likewise.
9480 (fill_slots_from_thread): Likewise.
9481 * resource.c (mark_referenced_resources): Likewise.
9482 (mark_set_resources, find_dead_or_set_registers): Likewise.
9484 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9486 * substring-locations.h (format_warning_va): Formatting fix for
9488 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9490 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9491 * substring-locations.c: Include intl.h.
9492 (format_warning_va): Turned into small wrapper around
9493 format_warning_n_va, renamed to ...
9494 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9495 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9497 (format_warning_at_substring_n): New function.
9498 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9499 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9500 format_warning_at_substring with just a shorter name instead of
9501 const function pointer.
9502 (fmtwarn_n): New function.
9503 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9504 appropriate, get rid of all the fmtstr temporaries, move conditionals
9505 with G_() wrapped string literals directly into fmtwarn arguments,
9506 cast dir.len to (int), formatting fixes.
9508 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9510 * doc/invoke.texi: Remove "Cilk Plus" references.
9512 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9513 Richard Biener <rguenther@suse.de>
9516 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9517 for error or warning attributes if CALL_FROM_THUNK_P is set.
9520 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9523 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9524 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9527 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9528 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9530 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9531 instead of -1U in last predictors element's probability member.
9533 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9536 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9537 arguments if they are comparable.
9539 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9541 PR tree-optimization/84634
9542 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9543 masks and masked_loop_p with a single loop_masks, making sure it's
9544 null for bb vectorization.
9546 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9548 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9549 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9550 loop->force_vectorize to check whether there is no alias.
9552 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9555 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9557 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9559 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9560 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9561 instead of a loop around prev_real_insn.
9562 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9566 * config/i386/i386.c (ix86_print_operand): Use conditional
9567 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9570 2018-03-02 Richard Biener <rguenther@suse.de>
9572 PR tree-optimization/84427
9573 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9574 (bitmap_set_subtract_values): Rewrite to handle multiple
9577 (prune_clobbered_mems): Likewise.
9578 (phi_translate): Take edge instead of pred/phiblock.
9579 (phi_translate_1): Likewise.
9580 (phi_translate_set): Likewise. Insert all translated
9581 exprs for a value into the set, keeping possibly multiple
9582 expressions per value.
9583 (compute_antic_aux): Adjust for phi_translate changes.
9584 When intersecting union the expressions and prune those
9585 not in the final value set, keeping possibly multiple
9586 expressions per value. Do not use value-insertion
9587 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9588 all expressions. Add verification that the value-sets
9589 only shrink during iteration.
9590 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9591 (do_pre_regular_insertion): Likewise.
9592 (do_pre_partial_partial_insertion): Likewise.
9594 2018-03-02 Richard Biener <rguenther@suse.de>
9597 * config/darwin.c (saved_debug_info_level): New static global.
9598 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9599 (darwin_asm_lto_end): Restore debug info generation settings.
9601 2018-03-01 Martin Liska <mliska@suse.cz>
9604 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9607 2018-03-01 Richard Biener <rguenther@suse.de>
9610 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9611 type annotation with LTO when debug was disabled at compile-time.
9613 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9615 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9617 (mips_final_postscan_insn): Likewise.
9619 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9621 PR rtl-optimization/84528
9622 * alias.c (init_alias_target): Add commentary.
9623 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9624 a unique base value if the frame pointer is not eliminated
9625 to the stack pointer.
9627 2018-03-01 Tom de Vries <tom@codesourcery.com>
9629 PR rtl-optimization/83327
9630 * lra-int.h (hard_regs_spilled_into): Declare.
9631 * lra.c (hard_regs_spilled_into): Define.
9632 (init_reg_info): Init hard_regs_spilled_into.
9633 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9634 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9635 (process_bb_lives): Handle hard_regs_spilled_into.
9636 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9637 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9639 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9641 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9642 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9643 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9644 * config/rs6000/aix72.h: New file.
9646 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9648 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9649 instead of warning_at with conditional singular and plural messages
9653 * stor-layout.c (update_alignment_for_field): For
9654 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9655 && !DECL_PACKED (field), do the alignment update, just use
9656 only desired_align instead of MAX (type_align, desired_align)
9658 (place_field): Don't do known_align < desired_align handling
9659 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9660 is non-NULL, instead do it after rli->prev_field handling and
9661 only if not within a bitfield word. For DECL_PACKED (field)
9662 use type_align of BITS_PER_UNIT.
9664 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9666 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9667 superfluous parentheses and trailing spaces.
9669 2018-02-28 Richard Biener <rguenther@suse.de>
9671 PR tree-optimization/84584
9672 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9673 SCoPs with fake exit edge.
9675 2018-02-28 Martin Liska <mliska@suse.cz>
9678 * timevar.c (timer::print): Fix format to properly print 100%
9681 2018-02-28 Richard Biener <rguenther@suse.de>
9684 * genmatch.c (capture_info::walk_match): Do not mark
9685 captured expressions without operands as expr_p given
9686 they act more like predicates and should be subject to
9687 "lost tail" side-effect preserving.
9689 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9691 PR rtl-optimization/81611
9692 * auto-inc-dec.c (attempt_change): Move dead note from
9693 mem_insn if it's the next use of regno
9694 (find_address): Take address use of reg holding
9695 non-incremented value. Add parm to limit search to the named
9697 (merge_in_block): Attempt to use a mem insn that is the next
9698 use of the original regno.
9700 2018-02-27 Martin Sebor <msebor@redhat.com>
9703 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9704 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9706 2018-02-27 Martin Sebor <msebor@redhat.com>
9708 PR translation/84207
9709 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9710 n argument to unsigned HOST_WIDE_INT.
9711 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9712 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9713 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9716 2018-02-27 Richard Biener <rguenther@suse.de>
9718 PR tree-optimization/84512
9719 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9720 Do not use the estimate returned from record_stmt_cost for
9721 the scalar iteration cost but sum properly using add_stmt_cost.
9723 2018-02-27 Richard Biener <rguenther@suse.de>
9725 PR tree-optimization/84466
9726 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9727 Adjust last change to less strictly validate use operands.
9729 2018-02-27 Martin Liska <mliska@suse.cz>
9731 PR gcov-profile/84548
9732 * gcov.c (process_file): Allow partial overlap and consider it
9733 also as group functions.
9734 (output_lines): Properly calculate range of lines for a group.
9736 2018-02-27 Martin Liska <mliska@suse.cz>
9738 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9739 'ggc' suffixes. Change first column width.
9740 (timer::print): Fix formatting of the column.
9742 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9744 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9745 preserve inline entry blocks for the sake of debug inline
9746 entry point markers alone.
9747 (remove_unused_locals): Suggest in comments a better place to
9748 force the preservation of inline entry blocks that are
9749 otherwise unused, but do not preserve them.
9751 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9753 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9755 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9758 * config/i386/constraints.md (Bs): Replace
9759 ix86_indirect_branch_register with
9760 TARGET_INDIRECT_BRANCH_REGISTER.
9762 * config/i386/i386.md (indirect_jump): Likewise.
9763 (tablejump): Likewise.
9764 (*sibcall_memory): Likewise.
9765 (*sibcall_value_memory): Likewise.
9766 Peepholes of indirect call and jump via memory: Likewise.
9767 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9768 (*sibcall_value_GOT_32): Likewise.
9769 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9770 (GOT_memory_operand): Likewise.
9771 (call_insn_operand): Likewise.
9772 (sibcall_insn_operand): Likewise.
9773 (GOT32_symbol_operand): Likewise.
9774 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9776 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9778 PR rtl-optimization/83496
9779 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9780 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9781 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9782 redundant insn, if any.
9783 (relax_delay_slots): Likewise.
9784 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9786 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9788 PR tree-optimization/83965
9789 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9790 that grouped statements are part of a reduction chain. Return
9791 true if the statement is not marked as a reduction itself but
9793 (vect_recog_dot_prod_pattern): Don't check whether the statement
9794 is part of a group here.
9795 (vect_recog_sad_pattern): Likewise.
9796 (vect_recog_widen_sum_pattern): Likewise.
9798 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9801 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9803 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9806 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9808 (ix86_output_indirect_function_return): New prototype.
9809 (ix86_split_simple_return_pop_internal): Likewise.
9810 * config/i386/i386.c (indirect_return_via_cx): New.
9811 (indirect_return_via_cx_bnd): Likewise.
9812 (indirect_thunk_name): Handle return va CX_REG.
9813 (output_indirect_thunk_function): Create alias for
9814 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9815 (ix86_output_indirect_jmp): Remove the bool argument.
9816 (ix86_output_indirect_function_return): New function.
9817 (ix86_split_simple_return_pop_internal): Likewise.
9818 * config/i386/i386.md (*indirect_jump): Don't pass false
9819 to ix86_output_indirect_jmp.
9820 (*tablejump_1): Likewise.
9821 (simple_return_pop_internal): Change it to define_insn_and_split.
9822 Call ix86_split_simple_return_pop_internal to split it for
9824 (simple_return_indirect_internal): Call
9825 ix86_output_indirect_function_return instead of
9826 ix86_output_indirect_jmp.
9828 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9831 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9832 memset and value initialization afterwards.
9834 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9836 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9838 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9841 * common/config/aarch64/aarch64-common.c
9842 (aarch_option_optimization_table[]): Switch
9843 off fomit-frame-pointer
9845 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9846 Chung-Ju Wu <jasonwucj@gmail.com>
9848 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9850 (store_multiple): Ditto.
9852 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9854 * config.gcc: Add --with-cpu support for nds32 target.
9855 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9856 * config/nds32/nds32.opt: Add -mcpu= option.
9858 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9860 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9861 isel=yes): Warn for these deprecated options.
9863 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9865 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9866 ISA_2_5_MASKS_EMBEDDED.
9868 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9870 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9871 p->max as pointers rather than using iterative_hash_expr.
9873 2018-02-23 Carl Love <cel@us.ibm.com>
9875 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9876 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9878 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9879 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9880 P8V_BUILTIN_VEC_VUNSIGNED2.
9882 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9885 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9886 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9888 * lra-constraints.c (curr_insn_transform): Set up
9889 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9891 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9892 * lra-eliminations.c (spill_pseudos): Ditto.
9893 (process_insn_for_elimination): Ditto.
9894 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9895 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9896 LRA_NON_CLOBBERED_ALT.
9898 2018-02-22 Martin Sebor <msebor@redhat.com>
9900 PR tree-optimization/84480
9901 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9902 to maybe_diag_stxncpy_trunc. Call it.
9903 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9904 from gimple_fold_builtin_strcpy. Print inlining stack.
9905 (handle_builtin_stxncpy): Print inlining stack.
9906 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9908 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9911 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9912 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9913 and -fcheck-pointer-bounds are used together.
9914 (indirect_thunk_prefix): New enum.
9915 (indirect_thunk_need_prefix): New function.
9916 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9917 "_nt" instead of "_bnd" for NOTRACK prefix.
9918 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9919 (output_indirect_thunk_function): Likewise.
9921 (ix86_code_end): Update output_indirect_thunk_function calls.
9922 (ix86_output_indirect_branch_via_reg): Replace
9923 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9924 (ix86_output_indirect_branch_via_push): Likewise.
9925 (ix86_output_function_return): Likewise.
9926 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9927 incompatible with -fcf-protection=branch and
9928 -fcheck-pointer-bounds.
9930 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9933 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9934 Change gcc_assert call to output_operand_lossage.
9936 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9938 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9940 2018-02-22 DJ Delorie <dj@redhat.com>
9941 Sebastian Perta <sebastian.perta@renesas.com>
9942 Oleg Endo <olegendo@gcc.gnu.org>
9944 * config/rx/rx.c (rx_rtx_costs): New function.
9945 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9947 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9949 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9951 2018-02-22 Martin Liska <mliska@suse.cz>
9954 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9955 Add "native" as a possible value.
9957 2018-02-22 Martin Liska <mliska@suse.cz>
9960 * config/i386/i386.c (ix86_option_override_internal):
9961 Add "native" as a possible value for -march and -mtune.
9963 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9966 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9967 to all type variants.
9969 PR tree-optimization/84503
9970 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9971 width as info->bitpos + info->bitsize - start.
9972 (merged_store_group::merge_overlapping): Simplify width computation.
9973 (check_no_overlap): New function.
9974 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9975 start + width and last_order of the group, fail if check_no_overlap
9977 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9978 to group if check_no_overlap fails.
9980 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9982 * config/rs6000/altivec.md: Delete contraint arguments to
9983 define_expand, define_split, and define_peephole2, and in
9984 define_insn_and_split if always unused.
9985 * config/rs6000/darwin.md: Ditto.
9986 * config/rs6000/dfp.md: Ditto.
9987 * config/rs6000/rs6000.md: Ditto.
9988 * config/rs6000/sync.md: Ditto.
9989 * config/rs6000/vector.md: Ditto.
9990 * config/rs6000/vsx.md: Ditto.
9992 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9994 * config/rs6000/altivec.md: Write output control strings as braced
9995 blocks instead of double-quoted strings.
9996 * config/rs6000/darwin.md: Ditto.
9997 * config/rs6000/rs6000.md: Ditto.
9998 * config/rs6000/vector.md: Ditto.
9999 * config/rs6000/vsx.md: Ditto.
10001 2018-02-21 Jason Merrill <jason@redhat.com>
10003 PR c++/84314 - ICE with templates and fastcall attribute.
10004 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10006 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10008 * ipa-cp.c (determine_versionability): Fix comment typos.
10010 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10013 * ipa-cp.c (determine_versionability): Do not version functions caling
10016 2018-02-21 Martin Liska <mliska@suse.cz>
10019 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10020 Add "native" as a possible value.
10021 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10022 the macro when native cpu detection is available.
10024 2018-02-21 Martin Liska <mliska@suse.cz>
10027 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10028 Add "native" as a possible value.
10029 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10030 when native cpu detection is available.
10032 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10033 Martin Sebor <msebor@redhat.com>
10035 PR tree-optimization/84478
10036 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10038 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10039 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10040 support which is conservatively correct, for 2 only stay conservative
10041 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10042 argument to the 2 argument get_range_strlen, adjust 6 arg
10043 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10045 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10046 (gimple_fold_builtin_strlen): Pass true as last argument to
10049 2018-02-20 Martin Sebor <msebor@redhat.com>
10051 PR middle-end/84095
10052 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10053 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10054 (builtin_memref::builtin_memref): Factor out parts into
10055 set_base_and_offset and call it.
10057 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10059 PR middle-end/84406
10060 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10061 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10062 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10063 search at the associated MODE_INT.
10065 2018-02-20 Jeff Law <law@redhat.com>
10067 PR middle-end/82123
10068 PR tree-optimization/81592
10069 PR middle-end/79257
10070 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10071 for range data rather than using global data.
10072 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10073 range data rather than using global data.
10074 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10075 pass it to children as needed.
10076 (struct directive::fmtresult): Similarly.
10077 (struct directive::set_width): Similarly.
10078 (struct directive::set_precision): Similarly.
10079 (format_integer, format_directive, parse_directive): Similarly.
10080 (format_none): Accept unnamed vr_values parameter.
10081 (format_percent, format_floating, format_character): Similarly.
10082 (format_string, format_plain): Similarly.
10083 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10084 the EVRP range analyzer for range data rather than using global data.
10085 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10086 gimple-ssa-evrp-analyze.h
10087 (class sprintf_dom_walker): Add after_dom_children member function.
10088 Add evrp_range_analyzer member.
10089 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10090 range analyzer as needed.
10091 (sprintf_dom_walker::after_dom_children): New member function.
10092 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10094 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10095 (evrp_range_analyzer::pop_to_marker): Likewise.
10097 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10099 PR tree-optimization/84419
10100 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10101 with the required type if its current type is compatible but
10104 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10106 PR middle-end/82004
10107 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10108 after vectorization.
10110 2018-02-20 Martin Liska <mliska@suse.cz>
10113 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10114 possible values if we don't have a hint.
10116 2018-02-20 Martin Liska <mliska@suse.cz>
10120 * final.c (shorten_branches): Build align_tab array with one
10122 * opts.c (finish_options): Add alignment option limit check.
10123 (MAX_CODE_ALIGN): Likewise.
10124 (MAX_CODE_ALIGN_VALUE): Likewise.
10125 * doc/invoke.texi: Document maximum allowed option value for
10126 all -falign-* options.
10128 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10131 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10132 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10133 * var-tracking.c (emit_note_insn_var_location): Remove all references
10134 to NOTE_INSN_CALL_ARG_LOCATION.
10135 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10136 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10137 Use copy_rtx_if_shared.
10138 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10139 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10140 (dwarf2out_var_location): Remove handling of
10141 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10143 * final.c (final_scan_insn): Remove all references to
10144 NOTE_INSN_CALL_ARG_LOCATION.
10145 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10146 before dumping final insns.
10147 * except.c (emit_note_eh_region_end): Remove all references to
10148 NOTE_INSN_CALL_ARG_LOCATION.
10149 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10150 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10151 * config/arc/arc.c (hwloop_optimize): Likewise.
10152 * config/arm/arm.c (create_fix_barrier): Likewise.
10153 * config/s390/s390.c (s390_chunkify_start): Likewise.
10154 * config/sh/sh.c (find_barrier): Likewise.
10155 * config/i386/i386.c (rest_of_insert_endbranch,
10156 ix86_seh_fixup_eh_fallthru): Likewise.
10157 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10158 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10159 * config/frv/frv.c (frv_function_prologue): Likewise.
10160 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10162 (note_outside_basic_block_p): Remove all references to
10163 NOTE_INSN_CALL_ARG_LOCATION.
10164 * gengtype.c (adjust_field_rtx_def): Likewise.
10165 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10167 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10168 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10171 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10174 PR tree-optimization/84452
10175 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10176 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10179 2018-02-19 Martin Liska <mliska@suse.cz>
10182 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10184 2018-02-19 Martin Liska <mliska@suse.cz>
10185 Richard Sandiford <richard.sandiford@linaro.org>
10187 PR tree-optimization/82491
10188 * gimple-fold.c (get_base_constructor): Make earlier bail out
10191 2018-02-19 Carl Love <cel@us.ibm.com>
10193 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10194 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10196 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10197 P8V_BUILTIN_VEC_NEG.
10199 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10201 * config/rl78/rl78.md (movdf): New define expand.
10203 2018-02-19 Martin Liska <mliska@suse.cz>
10206 * doc/invoke.texi: Fix typo.
10207 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10209 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10211 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10212 handle rs6000_single_float and rs6000_double_float specially for
10215 2018-02-16 Jeff Law <law@redhat.com>
10217 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10218 (pop_regs): Use it.
10220 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10223 * ipa-inline.c (inline_small_functions): Fix a typo.
10225 2018-02-16 Nathan Sidwell <nathan@acm.org>
10227 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10229 2018-02-16 Carl Love <cel@us.ibm.com>
10231 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10232 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10233 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10234 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10235 expansion to P8V_BUILTIN_VEC_FLOAT2.
10237 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10239 PR rtl-optimization/70023
10240 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10241 src_regno into account.
10243 2018-02-16 Carl Love <cel@us.ibm.com>
10245 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10246 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10247 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10248 * config/rs6000/rs6000.c: Remove case statements for
10249 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10250 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10251 and P9V_BUILTIN_VEC_VINSERT4B.
10252 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10253 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10254 * config/rs6000/vsx.md:
10255 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10258 2018-02-16 Carl Love <cel@us.ibm.com>
10260 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10262 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10264 * config/rs6000/rs6000-c.c: Add the definitions for
10265 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10266 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10267 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10268 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10269 definition for insert4b and define insn *insert3b_internal.
10270 * doc/extend.texi: Add documentation for vec_extract4b.
10272 2018-02-16 Nathan Sidwell <nathan@acm.org>
10274 * doc/extend.texi (Backwards Compatibility): Mention friend
10275 injection. Note for-scope is deprecated.
10276 * doc/invoke.texi (-ffriend-injection): Deprecate.
10278 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10280 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10281 that moved to I2, also allow destinations that are a paradoxical
10282 subreg (instead of a normal reg).
10284 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10287 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10290 2018-02-16 Richard Biener <rguenther@suse.de>
10292 PR tree-optimization/84037
10293 PR tree-optimization/84016
10295 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10296 Adjust vec_construct for the fact we need additional higher latency
10297 128bit inserts for AVX256 and AVX512 vector builds.
10298 (ix86_add_stmt_cost): Scale vector construction cost for
10301 2018-02-16 Richard Biener <rguenther@suse.de>
10303 PR tree-optimization/84417
10304 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10305 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10306 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10308 2018-02-16 Martin Liska <mliska@suse.cz>
10311 * internal-fn.def (ASAN_CHECK): Set proper flags.
10312 (ASAN_MARK): Likewise.
10314 2018-02-16 Julia Koval <julia.koval@intel.com>
10316 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10317 from PTA_CANNONLAKE.
10319 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10322 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10323 Use ++iter rather than iter++ for std::list iterators.
10324 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10325 defer deleting them until all nodes in the forest are processed. Do
10326 free even leaf nodes. Change to_process into auto_vec.
10329 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10330 * vec.h (vec_default_construct): Use memset instead of placement new
10331 if BROKEN_VALUE_INITIALIZATION is defined.
10332 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10333 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10336 PR rtl-optimization/83723
10337 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10338 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10339 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10341 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10343 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10345 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10347 PR rtl-optimization/81443
10348 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10349 from inner REGs to paradoxical SUBREGs.
10351 2018-02-16 Richard Biener <rguenther@suse.de>
10353 PR tree-optimization/84399
10354 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10355 For operands we can analyze at their definition make sure we can
10356 analyze them at each use as well.
10358 2018-02-16 Richard Biener <rguenther@suse.de>
10360 PR tree-optimization/84190
10361 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10362 volatile accesses if the decl isn't volatile.
10364 2018-02-15 Jason Merrill <jason@redhat.com>
10366 PR c++/84314 - ICE with templates and fastcall attribute.
10367 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10368 TYPE_CANONICAL on an existing type.
10370 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10372 PR tree-optimization/84383
10373 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10374 dstoff nor call operand_equal_p if dstbase is NULL.
10376 PR tree-optimization/84334
10377 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10378 also a CONSTANT_CLASS_P, punt.
10380 2018-02-14 Jim Wilson <jimw@sifive.com>
10382 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10383 first SMALL_OPERAND check. New local min_second_step. Move assert
10384 to where locals are set. Add TARGET_RVC support.
10385 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10387 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10389 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10391 2018-02-14 Martin Sebor <msebor@redhat.com>
10393 PR tree-optimization/83698
10394 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10395 arrays constrain the offset range to their bounds.
10396 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10397 (builtin_access::overlap): Avoid setting the size of overlap if it's
10399 (maybe_diag_overlap): Also consider arrays when deciding what values
10400 of offsets to include in diagnostics.
10402 2018-02-14 Martin Sebor <msebor@redhat.com>
10405 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10406 that correspond to the kind of a declaration.
10408 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10411 * config/pa/pa.md: Load address of PIC label using the linkage table
10412 if the label is nonlocal.
10414 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10416 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10417 warning message if user requests -maltivec=be.
10418 * doc/invoke.texi: Document deprecation of -maltivec=be.
10420 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10423 * config/rs6000/rs6000-c.c: Update definitions for
10424 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10425 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10427 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10430 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10431 add _get_ssp intrinsics. Remove argument from
10432 __builtin_ia32_rdssp[d|q].
10433 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10434 * config/i386/i386-builtin.def: Remove argument from
10435 __builtin_ia32_rdssp[d|q].
10436 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10437 ix86_expand_special_args_builtin for _rdssp[d|q].
10438 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10439 Clear register before usage.
10440 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10441 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10443 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10445 PR tree-optimization/84357
10446 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10447 operand 1 of an ARRAY_REF too.
10449 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10452 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10453 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10455 (set_of_reg): New struct.
10456 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10457 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10458 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10460 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10461 Split into bitclr, bitset, bitinvert patterns if appropriate.
10462 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10463 use rx_fuse_in_memory_bitop.
10464 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10465 to named insn, correct maximum insn length.
10467 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10470 * machmode.def: Define a complex mode for PARTIAL_INT.
10471 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10473 * doc/rtl.texi: Document CSPImode.
10474 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10476 (msp430_hard_regno_nregs_with_padding): Likewise.
10478 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10481 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10483 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10485 PR rtl-optimization/84169
10486 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10487 we generated a parallel as new i3 and we split that to new i2 and i3
10488 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10489 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10490 those to i2, not i1. Partially rewrite this scan code.
10492 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10495 * stor-layout.c (place_field): For variable length fields, adjust
10496 offset_align afterwards not just based on the field's alignment,
10497 but also on the size.
10499 PR middle-end/84309
10500 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10501 of exps and logs in the use_exp2 case.
10503 2018-02-13 Jeff Law <law@redhat.com>
10505 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10506 entry for "vector".
10508 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10511 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10515 * common.opt (gas-loc-support, gas-locview-support): New.
10516 (ginline-points, ginternal-reset-location-views): New.
10517 * doc/invoke.texi: Document them. Use @itemx where intended.
10518 (gvariable-location-views): Adjust.
10519 * target.def (reset_location_view): New.
10520 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10521 (TARGET_RESET_LOCATION_VIEW): New.
10522 * doc/tm.texi: Rebuilt.
10523 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10524 (dwarf2out_default_as_locview_support): New.
10525 (output_asm_line_debug_info): Use option variables.
10526 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10527 (output_loc_list): Likewise.
10528 (add_high_low_attributes): Check option variables.
10529 Don't output entry view attribute in strict mode.
10530 (gen_inlined_subroutine_die): Check option variables.
10531 (dwarf2out_inline_entry): Likewise.
10532 (init_sections_and_labels): Likewise.
10533 (dwarf2out_early_finish): Likewise.
10534 (maybe_reset_location_view): New, from...
10535 (dwarf2out_var_location): ... here. Call it.
10536 * debug.h (dwarf2out_default_as_loc_support): Declare.
10537 (dwarf2out_default_as_locview_support): Declare.
10538 * hooks.c (hook_int_rtx_insn_0): New.
10539 * hooks.h (hook_int_rtx_insn_0): Declare.
10540 * toplev.c (process_options): Take -gas-loc-support and
10541 -gas-locview-support from dwarf2out. Enable
10542 -gvariable-location-views by default only with locview
10543 assembler support. Enable -ginternal-reset-location-views by
10544 default only if the target defines the corresponding hook.
10545 Enable -ginline-points by default if location views are
10546 enabled; force it disabled if statement frontiers are
10548 * tree-inline.c (expand_call_inline): Check option variables.
10549 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10551 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10553 PR tree-optimization/84321
10554 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10555 handling. Also check whether the anti-range contains any values
10556 that satisfy the mask; switch to a VR_RANGE if not.
10558 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10561 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10563 2018-02-13 Martin Jambor <mjambor@suse.cz>
10566 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10567 of call statements, also set location of a load to a temporary.
10569 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10571 * config/rl78/rl78.c (add_vector_labels): New function.
10572 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10573 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10574 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10575 which checks that no arguments are passed.
10576 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10577 * doc/extend.texi: Documentation for the new attribute.
10579 2018-02-13 Andreas Schwab <schwab@suse.de>
10581 * config/riscv/linux.h (CPP_SPEC): Define.
10583 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10586 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10587 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10588 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10589 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10590 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10591 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10592 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10593 temporarily for AES and PCLMUL builtins.
10595 PR tree-optimization/84339
10596 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10597 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10600 PR middle-end/84309
10601 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10602 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10603 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10605 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10607 * omp-simd-clone.h: New file.
10608 * omp-simd-clone.c: Include omp-simd-clone.h.
10609 (expand_simd_clones): No longer static.
10610 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10611 cgraph.h and omp-simd-clone.h.
10612 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10613 (vect_recog_widen_shift_pattern): Formatting fix.
10614 (vect_pattern_recog_1): Don't check optab for calls.
10617 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10618 operands[2] into a REG before using gen_lowpart on it.
10620 2018-02-12 Jeff Law <law@redhat.com>
10623 * config/sh/sh.c (find_barrier): Consider a sibling call
10626 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10627 successfully back substituting a reg.
10629 2018-02-12 Richard Biener <rguenther@suse.de>
10631 PR tree-optimization/84037
10632 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10633 parameter, move visited init to caller.
10634 (vect_slp_analyze_operations): Separate cost from validity
10635 check, initialize visited once for all instances.
10636 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10638 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10640 (vect_model_promotion_demotion_cost): Likewise.
10641 (vectorizable_bswap): Guard cost modeling with !slp_node
10642 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10644 (vectorizable_call): Likewise.
10645 (vectorizable_conversion): Likewise.
10646 (vectorizable_assignment): Likewise.
10647 (vectorizable_shift): Likewise.
10648 (vectorizable_operation): Likewise.
10649 (vectorizable_store): Likewise.
10650 (vectorizable_load): Likewise.
10651 (vectorizable_condition): Likewise.
10652 (vectorizable_comparison): Likewise.
10654 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10657 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10658 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10659 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10661 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10663 PR middle-end/83665
10664 * params.def (inline-min-speedup): Increase from 8 to 15.
10665 (max-inline-insns-auto): Decrease from 40 to 30.
10666 * ipa-split.c (consider_split): Add some buffer for function to
10667 be considered inlining candidate.
10668 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10671 2018-02-12 Richard Biener <rguenther@suse.de>
10673 PR tree-optimization/84037
10674 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10675 matched stmts if we cannot swap the non-matched ones.
10677 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10679 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10680 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10681 _mm_maskz_scalef_round_ss): New intrinsics.
10682 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10683 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10684 __builtin_ia32_scalefss_round): Remove.
10685 (__builtin_ia32_scalefsd_mask_round,
10686 __builtin_ia32_scalefss_mask_round): New intrinsics.
10687 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10688 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10689 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10690 "<round_constraint>")): Changed to ...
10691 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10692 "<round_scalar_constraint>")): ... this.
10693 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10694 %0, %1, %2<round_op3>}"): Changed to ...
10695 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10696 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10697 %2<round_scalar_mask_op3>}"): ... this.
10698 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10700 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10702 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10703 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10704 (_mm_maskz_sqrt_round_ss): New intrinsics.
10705 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10706 (__builtin_ia32_sqrtsd_mask_round)
10707 (__builtin_ia32_sqrtss_mask_round): New builtins.
10708 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10709 (__builtin_ia32_sqrtss_round): Remove.
10710 (__builtin_ia32_sqrtsd_mask_round)
10711 (__builtin_ia32_sqrtss_mask_round): New builtins.
10712 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10713 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10714 ((match_operand:VF_128 1 "vector_operand"
10715 "xBm,<round_constraint>")): Changed to ...
10716 ((match_operand:VF_128 1 "vector_operand"
10717 "xBm,<round_scalar_constraint>")): ... this.
10718 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10719 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10720 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10721 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10722 %<iptr>1<round_scalar_mask_op3>}): ... this.
10723 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10724 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10726 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10729 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10730 Cast vec_cmpeq result to correct type.
10731 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10732 Cast vec_cmpgt result to correct type.
10734 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10736 * final.c (final_scan_insn_1): Renamed from...
10737 (final_scan_insn): ... this. New wrapper, to recover
10738 seen from the outermost call in recursive ones.
10739 * config/sparc/sparc.c (output_return): Drop seen from call.
10740 (output_sibcall): Likewise.
10741 * config/visium/visium.c (output_branch): Likewise.
10743 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10745 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10748 2018-02-10 Alan Modra <amodra@gmail.com>
10751 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10752 Specify LR as an input.
10754 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10757 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10758 remove_member_access_dummy_vars): New functions.
10759 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10760 lower_omp_1, execute_lower_omp): Use them.
10762 PR rtl-optimization/84308
10763 * shrink-wrap.c (spread_components): Release todo vector.
10765 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10767 PR rtl-optimization/57193
10768 * ira-color.c (struct allocno_color_data): Add member
10769 conflict_allocno_hard_prefs.
10770 (update_conflict_allocno_hard_prefs): New.
10771 (bucket_allocno_compare_func): Add a preference based on
10772 conflict_allocno_hard_prefs.
10773 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10774 (color_allocnos): Remove a dead code. Initiate
10775 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10777 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10780 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10781 constraint from =wa to wa. Avoid a subreg on the output operand,
10782 instead use a pseudo and subreg it in a move.
10783 (p9_xxbrd_<mode>): Changed to ...
10784 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10785 (p9_xxbrd_v2df): New expander.
10786 (p9_xxbrw_<mode>): Changed to ...
10787 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10788 (p9_xxbrw_v4sf): New expander.
10790 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10792 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10794 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10797 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10798 multiply in 32-bit mode.
10799 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10800 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10803 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10805 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10806 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10807 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10808 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10810 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10813 * dwarf2out.c (is_trivial_indirect_ref): New function.
10814 (dwarf2out_late_global_decl): Do not generate a location
10815 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10816 and that are not defined in the current unit.
10818 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10820 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10821 instead of a libcall for UNORDERED.
10823 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10826 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10827 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10829 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10832 * config/s390/s390.c (s390_set_current_function): Invoke
10833 s390_indirect_branch_settings also if fndecl didn't change.
10835 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10837 * config/rs6000/rs6000.md (blockage): Set length to zero.
10839 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10841 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10843 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10846 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10847 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10851 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10852 PARALLEL incoming that failed vt_get_decl_and_offset check.
10854 PR middle-end/84237
10855 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10856 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10858 (get_variable_section): For decls in named .bss* sections pass true as
10859 second argument to bss_initializer_p.
10861 2018-02-09 Marek Polacek <polacek@redhat.com>
10862 Jakub Jelinek <jakub@redhat.com>
10865 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10866 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10867 Sync some changes from cxx_fold_indirect_ref.
10869 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10871 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10873 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10874 (BLOCK_INLINE_ENTRY_LABEL): New.
10875 (dwarf2out_var_location): Disregard inline entry markers.
10876 (inline_entry_data): New struct.
10877 (inline_entry_data_hasher): New hashtable type.
10878 (inline_entry_data_hasher::hash): New.
10879 (inline_entry_data_hasher::equal): New.
10880 (inline_entry_data_table): New variable.
10881 (add_high_low_attributes): Add DW_AT_entry_pc and
10882 DW_AT_GNU_entry_view attributes if a pending entry is found
10883 in inline_entry_data_table. Add old entry_pc attribute only
10884 if debug nonbinding markers are disabled.
10885 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10886 markers are enabled.
10887 (block_within_block_p, dwarf2out_inline_entry): New.
10888 (dwarf2out_finish): Check that no entries remained in
10889 inline_entry_data_table.
10890 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10891 (final_scan_insn, notice_source_line): Likewise.
10892 (rest_of_clean_state): Skip inline entry markers.
10893 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10895 * gimple.c (gimple_build_debug_inline_entry): New.
10896 * gimple.h (enum gimple_debug_subcode): Add
10897 GIMPLE_DEBUG_INLINE_ENTRY.
10898 (gimple_build_debug_inline_entry): Declare.
10899 (gimple_debug_inline_entry_p): New.
10900 (gimple_debug_nonbind_marker_p): Adjust.
10901 * insn-notes.def (INLINE_ENTRY): New.
10902 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10903 inline entry marker notes.
10904 (print_insn): Likewise.
10905 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10906 (INSN_DEBUG_MARKER_KIND): Likewise.
10907 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10908 * tree-inline.c (expand_call_inline): Build and insert
10909 debug_inline_entry stmt.
10910 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10911 inline entry blocks early, if nonbind markers are enabled.
10912 (dump_scope_block): Dump fragment info.
10913 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10914 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10915 (gimple_build_debug_inline_entry): New.
10916 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10917 Enable/disable inline entry points too.
10918 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10919 (DEBUG_INSN): Describe inline entry markers.
10921 * common.opt (gvariable-location-views): New.
10922 (gvariable-location-views=incompat5): New.
10923 * config.in: Rebuilt.
10924 * configure: Rebuilt.
10925 * configure.ac: Test assembler for view support.
10926 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10927 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10928 * dwarf2out.c (var_loc_view): New typedef.
10929 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10930 (dwarf2out_locviews_in_attribute): New.
10931 (dwarf2out_locviews_in_loclist): New.
10932 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10933 (enum dw_line_info_opcode): Add LI_adv_address.
10934 (struct dw_line_info_table): Add view.
10935 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10936 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10937 (zero_view_p): New variable.
10938 (ZERO_VIEW_P): New macro.
10939 (output_asm_line_debug_info): New.
10940 (struct var_loc_node): Add view.
10941 (add_AT_view_list, AT_loc_list): New.
10942 (add_var_loc_to_decl): Add view param. Test it against last.
10943 (new_loc_list): Add view params. Record them.
10944 (AT_loc_list_ptr): Handle loc and view lists.
10945 (view_list_to_loc_list_val_node): New.
10946 (print_dw_val): Handle dw_val_class_view_list.
10947 (size_of_die): Likewise.
10948 (value_format): Likewise.
10949 (loc_list_has_views): New.
10950 (gen_llsym): Set vl_symbol too.
10951 (maybe_gen_llsym, skip_loc_list_entry): New.
10952 (dwarf2out_maybe_output_loclist_view_pair): New.
10953 (output_loc_list): Output view list or entries too.
10954 (output_view_list_offset): New.
10955 (output_die): Handle dw_val_class_view_list.
10956 (output_dwarf_version): New.
10957 (output_compilation_unit_header): Use it.
10958 (output_skeleton_debug_sections): Likewise.
10959 (output_rnglists, output_line_info): Likewise.
10960 (output_pubnames, output_aranges): Update version comments.
10961 (output_one_line_info_table): Output view numbers in asm comments.
10962 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10963 Call maybe_gen_llsym.
10964 (loc_list_from_tree_1): Adjust.
10965 (add_AT_location_description): Create view list attribute if
10966 needed, check it's absent otherwise.
10967 (convert_cfa_to_fb_loc_list): Adjust.
10968 (maybe_emit_file): Call output_asm_line_debug_info for test.
10969 (dwarf2out_var_location): Reset views as needed. Precompute
10970 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10971 attribute. Set view.
10972 (new_line_info_table): Reset next view.
10973 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10974 (dwarf2out_source_line): Likewise. Output view resets and labels to
10975 the assembler, or select appropriate line info opcodes.
10976 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10977 (optimize_string_length): Catch it. Adjust.
10978 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10979 dw_val_class_view_list, and remove it if no longer needed.
10980 (hash_loc_list): Hash view numbers.
10981 (loc_list_hasher::equal): Compare them.
10982 (optimize_location_lists): Check whether a view list symbol is
10983 needed, and whether the locview attribute is present, and
10984 whether they match. Remove the locview attribute if no longer
10986 (index_location_lists): Call skip_loc_list_entry for test.
10987 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10988 Use output_dwarf_version.
10989 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10990 (struct dw_val_node): Add val_view_list.
10991 * final.c (SEEN_NEXT_VIEW): New.
10992 (set_next_view_needed): New.
10993 (clear_next_view_needed): New.
10994 (maybe_output_next_view): New.
10995 (final_start_function): Rename to...
10996 (final_start_function_1): ... this. Take pointer to FIRST,
10997 add SEEN parameter. Emit param bindings in the initial view.
10998 (final_start_function): Reintroduce SEEN-less interface.
10999 (final): Rename to...
11000 (final_1): ... this. Take SEEN parameter. Output final pending
11001 next view at the end.
11002 (final): Reintroduce seen-less interface.
11003 (final_scan_insn): Output pending next view before switching
11004 sections or ending a block. Mark the next view as needed when
11005 outputting variable locations. Notify debug backend of section
11006 changes, and of location view changes.
11007 (rest_of_handle_final): Adjust.
11008 * toplev.c (process_options): Autodetect value for debug variable
11009 location views option. Warn on incompat5 without -gdwarf-5.
11010 * doc/invoke.texi (gvariable-location-views): New.
11011 (gvariable-location-views=incompat5): New.
11012 (gno-variable-location-views): New.
11014 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11016 PR tree-optimization/84136
11017 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11018 that the result of find_edge is non-NULL.
11020 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11023 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11024 storing integer register in SImode. Fix cost of 256 and 512
11025 byte aligned SSE register store.
11027 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11029 * config/i386/i386.c (ix86_multiplication_cost): Fix
11030 multiplication cost for TARGET_AVX512DQ.
11032 2018-02-08 Marek Polacek <polacek@redhat.com>
11034 PR tree-optimization/84238
11035 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11038 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11040 PR tree-optimization/84265
11041 * tree-vect-stmts.c (vectorizable_store): Don't treat
11042 VMAT_CONTIGUOUS accesses as grouped.
11043 (vectorizable_load): Likewise.
11045 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11047 PR tree-optimization/81635
11048 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11049 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11050 (test_round_for_mask): New functions.
11051 (wide_int_cc_tests): Call test_round_for_mask.
11052 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11053 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11054 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11055 range returned by get_range_info.
11057 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11060 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11061 * symtab.c: Include builtins.h
11062 (symtab_node::output_to_lto_symbol_table_p): Move here
11063 from lto-streamer-out.c:output_symbol_p.
11064 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11065 (output_symbol_p): Move all logic to symtab.c
11066 (produce_symtab): Update.
11068 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11070 * config/s390/s390-opts.h (enum indirect_branch): Define.
11071 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11072 (s390_indirect_branch_via_thunk)
11073 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11074 (enum s390_indirect_branch_type): Define.
11075 * config/s390/s390.c (struct s390_frame_layout, struct
11076 machine_function): Remove.
11077 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11078 (indirect_branch_table_label_no, indirect_branch_table_name):
11080 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11081 (enum s390_indirect_branch_option): Define.
11082 (s390_return_addr_from_memory): New function.
11083 (s390_handle_string_attribute): New function.
11084 (s390_attribute_table): Add new attribute handler.
11085 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11086 (s390_indirect_branch_via_thunk): New function.
11087 (s390_indirect_branch_via_inline_thunk): New function.
11088 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11089 sibling call optimization for non z10 compiles.
11090 (s390_emit_call): Force indirect branch target to be a single
11091 register. Add r1 clobber for non-z10 compiles.
11092 (s390_emit_epilogue): Emit return jump via return_use expander.
11093 (s390_reorg): Handle JUMP_INSNs as execute targets.
11094 (s390_option_override_internal): Perform validity checks for the
11095 new command line options.
11096 (s390_indirect_branch_attrvalue): New function.
11097 (s390_indirect_branch_settings): New function.
11098 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11099 (s390_output_indirect_thunk_function): New function.
11100 (s390_code_end): Implement target hook.
11101 (s390_case_values_threshold): Implement target hook.
11102 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11104 * config/s390/s390.h (struct s390_frame_layout)
11105 (struct machine_function): Move here from s390.c.
11106 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11107 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11108 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11109 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11110 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11111 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11112 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11113 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11114 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11115 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11116 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11117 (mnemonic attribute): Add values which aren't recognized
11119 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11120 pattern for branch conversion. Fix mnemonic attribute.
11121 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11122 indirect branch via thunk if requested.
11123 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11124 ("*indirect_jump"): Disable for branch conversion using out of
11126 ("indirect_jump_via_thunk<mode>_z10")
11127 ("indirect_jump_via_thunk<mode>")
11128 ("indirect_jump_via_inlinethunk<mode>_z10")
11129 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11130 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11131 ("casesi_jump_via_inlinethunk<mode>_z10")
11132 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11133 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11134 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11135 ("*indirect2_jump"): Disable for branch conversion.
11136 ("casesi_jump"): Turn into expander and expand patterns for branch
11138 ("return_use"): New expander.
11139 ("*return"): Emit return via thunk and rename it to ...
11140 ("*return<mode>"): ... this one.
11141 * config/s390/s390.opt: Add new options and and enum for the
11144 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11146 * lra-constraints.c (match_reload): Unconditionally use
11147 gen_lowpart_SUBREG, rather than selecting between that
11148 and equivalent gen_rtx_SUBREG code.
11150 2018-02-08 Richard Biener <rguenther@suse.de>
11152 PR tree-optimization/84233
11153 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11154 changed flag instead of boguously re-using phi_inserted.
11156 2018-02-08 Martin Jambor <mjambor@suse.cz>
11158 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11159 static local variables.
11161 2018-02-08 Richard Biener <rguenther@suse.de>
11163 PR tree-optimization/84278
11164 * tree-vect-stmts.c (vectorizable_store): When looking for
11165 smaller vector types to perform grouped strided loads/stores
11166 make sure the mode is supported by the target.
11167 (vectorizable_load): Likewise.
11169 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11171 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11172 Increase LDP/STP opportunities by adding adjacent callee-saves.
11174 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11176 PR rtl-optimization/84068
11177 PR rtl-optimization/83459
11178 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11180 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11182 PR tree-optimization/84224
11183 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11184 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11185 non-zero arguments.
11187 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11190 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11191 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11192 count, remove one USE.
11194 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11196 * doc/install.texi (Configuration): Document the
11197 --with-long-double-format={ibm,ieee} PowerPC configuration
11201 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11202 Convert from define_expand to be define_insn_and_split. Rework
11203 float/double/_Float128 conversions to QI/HI/SImode to work with
11204 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11205 conversions to QI/HImode types did a store and then a load to
11206 truncate the value. For conversions to VSX registers, don't split
11207 the insn, instead emit the code directly. Use the code iterator
11208 any_fix to combine signed and unsigned conversions.
11209 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11210 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11211 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11212 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11213 (fix_<mode>di2_hw): Likewise.
11214 (fixuns_<mode>di2_hw): Likewise.
11215 (fix_<mode>si2_hw): Likewise.
11216 (fixuns_<mode>si2_hw): Likewise.
11217 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11218 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11219 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11220 fix<uns>_trunc<SFDF:mode>si2_p8.
11221 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11222 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11223 (fix<uns>_<mode>_mem): Likewise.
11224 (fctiw<u>z_<mode>_mem): Likewise.
11225 (fix<uns>_<mode>_mem): Likewise.
11226 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11227 the register allocator from doing a direct move to the GPRs to do
11228 a store, and instead use the ISA 3.0 store byte/half-word from
11229 vector register instruction. For IEEE 128-bit floating point,
11230 also optimize stores of 32-bit ints.
11231 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11233 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11235 * genextract.c (push_pathstr_operand): New function to support
11237 (walk_rtx): Call push_pathstr_operand.
11238 (print_path): Support [a-zA-Z].
11240 2018-02-07 Richard Biener <rguenther@suse.de>
11242 PR tree-optimization/84037
11243 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11244 (cse_and_gimplify_to_preheader): Declare.
11245 (vect_get_place_in_interleaving_chain): Likewise.
11246 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11248 (_loop_vec_info::~_loop_vec_info): Delete it.
11249 (cse_and_gimplify_to_preheader): New function.
11250 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11251 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11252 (vectorizable_load): Likewise. For grouped stores always base
11253 the IV on the first element.
11254 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11255 condition before gimplifying.
11257 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11259 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11260 *DIV_EXPR and *MOD_EXPR.
11262 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11265 * config/i386/i386.c (ix86_option_override_internal): Mask out
11266 the CF_SET bit when checking -fcf-protection.
11268 2018-02-07 Tom de Vries <tom@codesourcery.com>
11271 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11274 2018-02-07 Richard Biener <rguenther@suse.de>
11276 PR tree-optimization/84204
11277 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11280 PR tree-optimization/84205
11281 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11282 special-case isl_ast_op_zdiv_r.
11284 PR tree-optimization/84223
11285 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11286 Only add conditions from within the region.
11287 (gather_bbs::after_dom_children): Adjust.
11289 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11292 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11293 * config/avr/avr.md: Only post-reload split REG-REG moves if
11294 either register is GENERAL_REG_P.
11296 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11298 PR tree-optimization/84235
11299 * tree-ssa-scopedtables.c
11300 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11301 if the subtraction is performed in floating point type where NaNs are
11302 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11303 build 1. Formatting fix.
11305 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11308 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11309 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11310 and skip it regardless of bb boundaries. Use CALL_P macro,
11311 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11312 unnecessarily, formatting fix.
11314 2018-02-06 Michael Collison <michael.collison@arm.com>
11316 * config/arm/thumb2.md:
11317 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11318 (*thumb_mov_notscc): Ditto.
11320 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11323 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11324 unsigned_fix, not "s".
11326 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11328 * configure.ac (gcc_fn_eh_frame_ro): New function.
11329 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11330 correct .eh_frame permissions.
11331 * configure: Regenerate.
11333 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11335 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11336 irrelevant options.
11338 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11340 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11341 Display warning message for -mno-speculate-indirect-jumps.
11343 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11345 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11347 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11349 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11351 PR tree-optimization/84225
11352 * tree-eh.c (find_trapping_overflow): Only call
11353 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11355 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11358 * config/i386/i386.c: Reimplement the check of possible options
11359 -mibt/-mshstk conbination. Change error messages.
11360 * doc/invoke.texi: Fix a typo: remove extra '='.
11362 2018-02-06 Marek Polacek <polacek@redhat.com>
11364 PR tree-optimization/84228
11365 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11367 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11370 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11371 emitted arch directives.
11372 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11373 __ARM_FEATURE_COPROC before changing architectures.
11375 2018-02-06 Richard Biener <rguenther@suse.de>
11377 * config/i386/i386.c (print_reg): Fix typo.
11378 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11380 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11382 * configure: Regenerate.
11384 2018-02-05 Martin Sebor <msebor@redhat.com>
11386 PR tree-optimization/83369
11387 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11390 2018-02-05 Martin Liska <mliska@suse.cz>
11392 * doc/invoke.texi: Cherry-pick upstream r323995.
11394 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11396 * ira.c (ira_init_register_move_cost): Adjust comment.
11398 2018-02-05 Martin Liska <mliska@suse.cz>
11400 PR gcov-profile/84137
11401 * doc/gcov.texi: Fix typo in documentation.
11403 2018-02-05 Martin Liska <mliska@suse.cz>
11405 PR gcov-profile/83879
11406 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11407 using dlopen functionality.
11409 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11411 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11412 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11413 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11414 _mm_maskz_range_round_ss): New intrinsics.
11415 (__builtin_ia32_rangesd128_round)
11416 (__builtin_ia32_rangess128_round): Remove.
11417 (__builtin_ia32_rangesd128_mask_round,
11418 __builtin_ia32_rangess128_mask_round): New builtins.
11419 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11420 __builtin_ia32_rangess128_round): Remove.
11421 (__builtin_ia32_rangesd128_mask_round,
11422 __builtin_ia32_rangess128_mask_round): New builtins.
11423 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11424 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11425 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11426 "<round_saeonly_constraint>")): Changed to ...
11427 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11428 "<round_saeonly_scalar_constraint>")): ... this.
11429 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11430 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11431 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11432 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11433 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11435 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11437 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11439 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11440 Remove all values except native, 8540 and 8548.
11442 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11444 * config/i386/i386.c (ix86_output_function_return): Pass
11445 INVALID_REGNUM, instead of -1, as invalid register number to
11446 indirect_thunk_name and output_indirect_thunk.
11448 2018-02-02 Julia Koval <julia.koval@intel.com>
11450 * config.gcc: Add -march=icelake.
11451 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11452 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11453 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11454 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11455 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11456 (processor_target_table): Add icelake.
11457 (ix86_option_override_internal): Handle new PTAs.
11458 (get_builtin_code_for_version): Handle icelake.
11459 (M_INTEL_COREI7_ICELAKE): New.
11460 (fold_builtin_cpu): Handle icelake.
11461 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11462 * doc/invoke.texi: Add -march=icelake.
11464 2018-02-02 Julia Koval <julia.koval@intel.com>
11466 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11467 to wide_int_bitmask.
11468 * wide-int-bitmask.h: New.
11470 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11473 * config/i386/i386.md: Replace Pmode with word_mode in
11474 builtin_setjmp_setup and builtin_longjmp to support x32.
11476 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11480 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11482 (rs6000_supported_cpu_names): New static variable.
11483 (linux_cpu_translation_table): Likewise.
11484 (elf_platform) <cpu>: Define new static variable and use it.
11485 Translate kernel AT_PLATFORM name to canonical name if needed.
11486 Error if platform name is unknown.
11488 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11491 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11493 2018-02-01 Jeff Law <law@redhat.com>
11496 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11497 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11498 the scratch if RELEASE_VIA_POP is false.
11499 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11500 If we have to save a temporary register, decrement SIZE appropriately.
11501 Pass new arguments to release_scratch_register_on_entry.
11502 (ix86_adjust_stack_and_probe): Likewise.
11503 (ix86_emit_probe_stack_range): Pass new arguments to
11504 release_scratch_register_on_entry.
11506 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11508 PR rtl-optimization/84157
11509 * combine.c (change_zero_ext): Use REG_P predicate in
11510 front of HARD_REGISTER_P predicate.
11512 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11514 * config/avr/avr.c (avr_option_override): Move disabling of
11515 -fdelete-null-pointer-checks to...
11516 * common/config/avr/avr-common.c (avr_option_optimization_table):
11519 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11521 PR tree-optimization/81635
11522 * tree-data-ref.c (split_constant_offset_1): For types that
11523 wrap on overflow, try to use range info to prove that wrapping
11526 2018-02-01 Renlin Li <renlin.li@arm.com>
11529 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11530 TAILCALL_ADDR_REGS.
11531 (aarch64_register_move_cost): Likewise.
11532 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11533 TAILCALL_ADDR_REGS.
11534 (REG_CLASS_NAMES): Likewise.
11535 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11536 TAILCALL_ADDR_REGS. Remove IP registers.
11537 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11539 2018-02-01 Richard Biener <rguenther@suse.de>
11541 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11542 for specifying RPO order and allow NULL for that.
11543 * domwalk.c (dom_walker::dom_walker): Likewise.
11544 (dom_walker::walk): Handle NULL RPO order.
11545 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11547 (rewrite_update_dom_walker): Likewise.
11548 (mark_def_dom_walker): Likewise.
11550 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11552 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11553 (aarch64_maybe_expand_sve_subreg_move): Declare.
11554 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11555 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11557 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11558 that are semantically a reverse operation.
11559 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11560 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11561 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11563 (aarch64_can_change_mode_class): For big-endian, forbid changes
11564 between two SVE modes if they have different element sizes.
11566 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11568 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11569 the TImode handling for big-endian targets.
11571 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11573 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11574 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11576 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11577 Remove BSWAP handing for big-endian targets and use the form of
11578 LD1RQ appropariate for the mode.
11580 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11582 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11583 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11584 duplicated element.
11586 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11589 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11590 check for operands that need to go through aarch64_sve_reload_be.
11592 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11594 PR tree-optimization/81661
11595 PR tree-optimization/84117
11596 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11597 * tree-eh.c: Include gimplify.h.
11598 (find_trapping_overflow, replace_trapping_overflow,
11599 rewrite_to_non_trapping_overflow): New functions.
11600 * tree-vect-loop.c: Include tree-eh.h.
11601 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11602 * tree-data-ref.c: Include tree-eh.h.
11603 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11605 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11607 PR rtl-optimization/84123
11608 * combine.c (change_zero_ext): Check if hard register satisfies
11609 can_change_dest_mode before calling gen_lowpart_SUBREG.
11611 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11614 * ira.c (ira_init_register_move_cost): Remove assert.
11616 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11618 PR rtl-optimization/84071
11619 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11620 * doc/tm.texi: Regenerate.
11622 2018-01-31 Richard Biener <rguenther@suse.de>
11624 PR tree-optimization/84132
11625 * tree-data-ref.c (analyze_miv_subscript): Properly
11626 check whether evolution_function_is_affine_multivariate_p
11627 before calling gcd_of_steps_may_divide_p.
11629 2018-01-31 Julia Koval <julia.koval@intel.com>
11632 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11633 * config/i386/i386.md (rdpid_rex64) New.
11634 (rdpid): Make 32bit only.
11636 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11639 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11640 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11642 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11645 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11647 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11649 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11651 PR rtl-optimization/84071
11652 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11653 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11655 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11657 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11658 (arc_attribute_table): Add 'aux' attribute.
11659 (arc_in_small_data_p): Consider aux like variables.
11660 (arc_is_aux_reg_p): New function.
11661 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11662 (arc_get_aux_arg): New function.
11663 (prepare_move_operands): Handle aux-register access.
11664 (arc_handle_aux_attribute): New function.
11665 * doc/extend.texi (ARC Variable attributes): Add subsection.
11667 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11669 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11670 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11671 (arc_attribute_table): Add 'uncached' attribute.
11672 (arc_print_operand): Print '.di' flag for uncached memory
11674 (arc_in_small_data_p): Do not consider for small data the uncached
11676 (arc_is_uncached_mem_p): New function.
11677 * config/arc/predicates.md (compact_store_memory_operand): Check
11678 for uncached memory accesses.
11679 (nonvol_nonimm_operand): Likewise.
11680 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11682 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11685 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11686 falign-loops=): Add Optimization flag.
11688 2018-01-30 Jeff Law <law@redhat.com>
11691 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11692 INT_REGISTERS_SAVED. Check it prior to calling
11693 get_scratch_register_on_entry.
11694 (ix86_adjust_stack_and_probe): Similarly.
11695 (ix86_emit_probe_stack_range): Similarly.
11696 (ix86_expand_prologue): Corresponding changes.
11698 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11701 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11702 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11704 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11707 * lra-constraints.c (curr_insn_transform): Process AND in the
11710 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11712 PR rtl-optimization/83986
11713 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11714 dependence against last_pending_memory_flush in addition to
11715 pending_jump_insns.
11717 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11719 PR tree-optimization/81611
11720 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11723 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11726 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11729 2018-01-30 Richard Biener <rguenther@suse.de>
11730 Jakub Jelinek <jakub@redhat.com>
11732 PR tree-optimization/84111
11733 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11734 inner loops added during recursion, as they don't have up-to-date
11737 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11740 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11741 (can_inline_edge_by_limits_p): ... here.
11742 (can_early_inline_edge_p, check_callers,
11743 update_caller_keys, update_callee_keys, recursive_inlining,
11744 add_new_edges_to_heap, speculation_useful_p,
11745 inline_small_functions,
11746 inline_small_functions, flatten_function,
11747 inline_to_all_callers_1): Update.
11749 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11751 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11754 2018-01-30 Richard Biener <rguenther@suse.de>
11756 PR tree-optimization/83008
11757 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11758 invariant and constant vector uses in stmts when they need
11759 more than one stmt.
11761 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11764 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11765 * configure: Regenerate.
11767 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11769 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11771 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11772 Use gen_rtx_REG rather than gen_lowpart.
11774 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11776 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11777 rather than 0 when creating partial subregs.
11779 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11781 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11784 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11787 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11788 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11789 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11790 flags. This restores the settings used before the 2017-07-24.
11791 Turning off pre increment/decrement/modify allows IVOPTS to
11792 optimize DF/SF loops where the index is an int.
11794 2018-01-29 Richard Biener <rguenther@suse.de>
11795 Kelvin Nilsen <kelvin@gcc.gnu.org>
11798 * tree-vect-stmts.c (vectorizable_call): Don't call
11799 targetm.vectorize_builtin_md_vectorized_function if callee is
11802 2018-01-22 Carl Love <cel@us.ibm.com>
11804 * doc/extend.tex: Fix typo in second arg in
11805 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11807 2018-01-29 Richard Biener <rguenther@suse.de>
11809 PR tree-optimization/84086
11810 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11811 (flush_ssaname_freelist): When SSA names were released reset
11812 the SCEV hash table.
11814 2018-01-29 Richard Biener <rguenther@suse.de>
11816 PR tree-optimization/84057
11817 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11818 removed paths when removing edges.
11820 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11822 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11823 -mfunction-return=@var{choice}.
11825 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11827 PR diagnostic/84034
11828 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11829 Handle CR like TAB.
11830 (layout::print_source_line): Likewise.
11831 (test_get_line_width_without_trailing_whitespace): Add test cases.
11833 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11835 PR middle-end/84040
11836 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11839 2018-01-26 Jim Wilson <jimw@sifive.com>
11841 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11843 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11846 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11848 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11849 and CMP + SUB-immediate -> SUBS.
11851 2018-01-26 Martin Sebor <msebor@redhat.com>
11853 PR tree-optimization/83896
11854 * tree-ssa-strlen.c (get_string_len): Rename...
11855 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11856 Avoid assuming length is constant.
11857 (handle_char_store): Use HOST_WIDE_INT for string length.
11859 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11862 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11863 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11865 2018-01-26 Richard Biener <rguenther@suse.de>
11867 PR rtl-optimization/84003
11868 * dse.c (record_store): Only record redundant stores when
11869 the earlier store aliases at least all accesses the later one does.
11871 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11873 PR rtl-optimization/83985
11874 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11875 REG_CFA_RESTORE insns.
11876 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11877 REG_CFA_RESTORE insns here.
11880 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11881 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11883 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11885 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11886 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11887 (arc_init): Likewise.
11888 (arc_override_options): Likewise.
11889 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11891 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11893 * config/arc/arc.h (TARGET_DBNZ): Define.
11894 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11895 properly set the tune attribute.
11896 (dbnz): Use TARGET_DBNZ guard.
11897 * config/arc/arc.opt (mtune): Add core3 option.
11899 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11901 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11902 recognize new pic like addresses.
11903 (arc_delegitimize_address): Clean up.
11905 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11907 * config/arc/arc-arches.def: Option mrf16 valid for all
11909 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11910 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11911 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11912 * config/arc/arc-tables.opt: Regenerate.
11913 * config/arc/arc.c (arc_conditional_register_usage): Handle
11914 reduced register file case.
11915 (arc_file_start): Set must have build attributes.
11916 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11917 mrf16 option value.
11918 * config/arc/arc.opt (mrf16): Add new option.
11919 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11920 * config/arc/genmultilib.awk: Handle new mrf16 option.
11921 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11922 * config/arc/t-multilib: Regenerate.
11923 * doc/invoke.texi (ARC Options): Document mrf16 option.
11925 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11927 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11928 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11929 (arc_attribute_table): Add 'secure_call' attribute.
11930 (arc_print_operand): Print secure call operand.
11931 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11933 (arc_is_secure_call_p): New function. * config/arc/arc.md
11934 (call_i): Add support for sjli instruction.
11935 (call_value_i): Likewise.
11936 * config/arc/constraints.md (Csc): New constraint.
11938 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11939 John Eric Martin <John.Martin@emmicro-us.com>
11941 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11942 * config/arc/arc.c (_arc_jli_section): New struct.
11943 (arc_jli_section): New type.
11944 (rc_jli_sections): New static variable.
11945 (arc_handle_jli_attribute): New function.
11946 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11947 (arc_file_end): New function.
11948 (TARGET_ASM_FILE_END): Define.
11949 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11950 (arc_add_jli_section): New function.
11951 (jli_call_scan): Likewise.
11952 (arc_reorg): Call jli_call_scan.
11953 (arc_output_addsi): Remove 'S' from printing asm operand.
11954 (arc_is_jli_call_p): New function.
11955 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11957 (movhi_insn): Likewise.
11958 (movsi_insn): Likewise.
11959 (movsi_set_cc_insn): Likewise.
11960 (loadqi_update): Likewise.
11961 (load_zeroextendqisi_update): Likewise.
11962 (load_signextendqisi_update): Likewise.
11963 (loadhi_update): Likewise.
11964 (load_zeroextendhisi_update): Likewise.
11965 (load_signextendhisi_update): Likewise.
11966 (loadsi_update): Likewise.
11967 (loadsf_update): Likewise.
11968 (movsicc_insn): Likewise.
11969 (bset_insn): Likewise.
11970 (bxor_insn): Likewise.
11971 (bclr_insn): Likewise.
11972 (bmsk_insn): Likewise.
11973 (bicsi3_insn): Likewise.
11974 (cmpsi_cc_c_insn): Likewise.
11975 (movsi_ne): Likewise.
11976 (movsi_cond_exec): Likewise.
11977 (clrsbsi2): Likewise.
11978 (norm_f): Likewise.
11988 (call_i): Remove 'S' asm letter, add jli instruction.
11989 (call_value_i): Likewise.
11990 * config/arc/arc.op (mjli-always): New option.
11991 * config/arc/constraints.md (Cji): New constraint.
11992 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11994 (subsf3_fpx): Likewise.
11995 (mulsf3_fpx): Likewise.
11996 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11998 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11999 function attrbutes.
12000 * doc/invoke.texi (ARC): Document mjli-always option.
12002 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12004 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12005 avoid addition with 0 and use incw and decw where possible.
12007 2018-01-26 Richard Biener <rguenther@suse.de>
12009 PR tree-optimization/81082
12010 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12011 association if it requires casting to unsigned.
12012 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12013 from fold_plusminus_mult_expr to catch important cases late when
12014 range info is available.
12016 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12018 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12019 * configure.ac (hidden_linkonce): New test.
12020 * configure: Regenerate.
12021 * config.in: Regenerate.
12023 2018-01-26 Julia Koval <julia.koval@intel.com>
12025 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12026 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12027 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12028 _mm_mask_bitshuffle_epi64_mask): Fix type.
12029 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12030 USI_FTYPE_V4DI_V4DI_USI): Remove.
12031 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12032 __builtin_ia32_vpshufbitqmb256_mask,
12033 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12034 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12035 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12037 2018-01-26 Alan Modra <amodra@gmail.com>
12040 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12041 UNSPEC_VBPERMQ. Sort other unspecs.
12043 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12045 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12047 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12049 PR middle-end/83055
12050 * predict.c (drop_profile): Do not push/pop cfun; update also
12052 (handle_missing_profiles): Fix logic looking for zero profiles.
12054 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12056 PR middle-end/83977
12057 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12058 on functions with #pragma omp declare simd or functions with simd
12060 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12061 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12062 Remove trailing \n from warning_at calls.
12064 2018-01-25 Tom de Vries <tom@codesourcery.com>
12067 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12068 for neutered workers.
12070 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12073 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12074 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12076 2018-01-24 Jeff Law <law@redhat.com>
12079 * i386.c (get_probe_interval): Move to earlier point.
12080 (ix86_compute_frame_layout): If -fstack-clash-protection and
12081 the frame is larger than the probe interval, then use pushes
12082 to save registers rather than reg->mem moves.
12083 (ix86_expand_prologue): Remove conditional for int_registers_saved
12086 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12089 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12090 min/max for never referenced object.
12092 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12094 PR middle-end/83977
12095 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12097 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12098 attributes from DECL_ATTRIBUTES (decl) without affecting
12099 DECL_ATTRIBUTES (current_function_decl).
12100 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12101 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12103 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12105 PR tree-optimization/83979
12106 * fold-const.c (fold_comparison): Use constant_boolean_node
12107 instead of boolean_{true,false}_node.
12109 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12111 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12114 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12116 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12117 Simplify the clause that sets the length attribute.
12118 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12119 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12120 clause that sets the length attribute.
12121 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12123 2018-01-24 Tom de Vries <tom@codesourcery.com>
12126 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12127 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12128 Add strict parameter.
12129 (prevent_branch_around_nothing): Insert dummy insn between branch to
12130 label and label with no ptx insn inbetween.
12131 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12133 2018-01-24 Tom de Vries <tom@codesourcery.com>
12136 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12137 for neutered threads in warp.
12138 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12140 2018-01-24 Richard Biener <rguenther@suse.de>
12142 PR tree-optimization/83176
12143 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12146 2018-01-24 Richard Biener <rguenther@suse.de>
12148 PR tree-optimization/82819
12149 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12150 code generating pluses that are no-ops in the target precision.
12152 2018-01-24 Richard Biener <rguenther@suse.de>
12154 PR middle-end/84000
12155 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12157 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12159 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12160 to merge probabilities.
12161 * predict.c (probably_never_executed): Also mark as cold functions
12162 with global 0 profile and guessed local profile.
12163 * profile-count.c (profile_probability::combine_with_count): New
12165 * profile-count.h (profile_probability::operator*,
12166 profile_probability::operator*=, profile_probability::operator/,
12167 profile_probability::operator/=): Reduce precision to adjusted
12168 and set value to guessed on contradictory divisions.
12169 (profile_probability::combine_with_freq): Remove.
12170 (profile_probability::combine_wiht_count): Declare.
12171 (profile_count::force_nonzero):: Set to adjusted.
12172 (profile_count::probability_in):: Set quality to adjusted.
12173 * tree-ssa-tail-merge.c (replace_block_by): Use
12174 combine_with_count.
12176 2018-01-23 Andrew Waterman <andrew@sifive.com>
12177 Jim Wilson <jimw@sifive.com>
12179 * config/riscv/riscv.c (riscv_stack_boundary): New.
12180 (riscv_option_override): Set riscv_stack_boundary. Handle
12181 riscv_preferred_stack_boundary_arg.
12182 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12183 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12184 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12185 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12186 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12187 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12189 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12192 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12193 of struct ix86_frame.
12194 (ix86_expand_epilogue): Likewise. Add a local variable for
12195 the reg_save_offset field in struct ix86_frame.
12197 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12199 PR tree-optimization/82604
12200 * tree-loop-distribution.c (enum partition_kind): New enum item
12201 PKIND_PARTIAL_MEMSET.
12202 (partition_builtin_p): Support above new enum item.
12203 (generate_code_for_partition): Ditto.
12204 (compute_access_range): Differentiate cases that equality can be
12205 proven at all loops, the innermost loops or no loops.
12206 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12207 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12208 (finalize_partitions, distribute_loop): Don't fuse partition of
12209 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12210 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12211 parloop is enabled.
12213 2018-01-23 Martin Liska <mliska@suse.cz>
12215 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12216 order to ignore the predictor.
12217 (PRED_POLYMORPHIC_CALL): Likewise.
12218 (PRED_RECURSIVE_CALL): Likewise.
12220 2018-01-23 Martin Liska <mliska@suse.cz>
12222 * tree-profile.c (tree_profiling): Print function header to
12223 aware reader which function we are working on.
12224 * value-prof.c (gimple_find_values_to_profile): Do not print
12225 not interesting value histograms.
12227 2018-01-23 Martin Liska <mliska@suse.cz>
12229 * profile-count.h (enum profile_quality): Add
12230 profile_uninitialized as the first value. Do not number values
12231 as they are zero based.
12232 (profile_count::verify): Update sanity check.
12233 (profile_probability::verify): Likewise.
12235 2018-01-23 Nathan Sidwell <nathan@acm.org>
12237 * doc/invoke.texi (ffor-scope): Deprecate.
12239 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12241 PR tree-optimization/83510
12242 * domwalk.c (set_all_edges_as_executable): New function.
12243 (dom_walker::dom_walker): Convert bool param
12244 "skip_unreachable_blocks" to enum reachability. Move setup of
12245 edge flags to set_all_edges_as_executable and only do it when
12246 reachability is REACHABLE_BLOCKS.
12247 * domwalk.h (enum dom_walker::reachability): New enum.
12248 (dom_walker::dom_walker): Convert bool param
12249 "skip_unreachable_blocks" to enum reachability.
12250 (set_all_edges_as_executable): New decl.
12251 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12252 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12254 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12255 but converting true to REACHABLE_BLOCKS.
12256 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12258 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12259 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12260 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12262 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12263 if check_all_array_refs will be called.
12265 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12267 * tree.c (selftest::test_location_wrappers): Add more test
12270 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12272 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12273 (selftest::test_bit_in_range): Likewise.
12275 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12278 * doc/sourcebuild.texi (vect_float): Say that the selector
12279 only describes the situation when -funsafe-math-optimizations is on.
12280 (vect_float_strict): Document.
12282 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12284 PR tree-optimization/83965
12285 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12286 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12287 instead of checking only for a reduction.
12288 (vect_recog_widen_sum_pattern): Likewise.
12290 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12292 * predict.c (probably_never_executed): Only use precise profile info.
12293 (compute_function_frequency): Skip after inlining hack since we now
12294 have quality checking.
12296 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12298 * profile-count.h (profile_probability::very_unlikely,
12299 profile_probability::unlikely, profile_probability::even): Set
12300 precision to guessed.
12302 2018-01-23 Richard Biener <rguenther@suse.de>
12304 PR tree-optimization/83963
12305 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12306 Properly terminate dominator walk when crossing the exit edge not
12307 when visiting its source block.
12309 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12312 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12313 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12315 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12317 PR tree-optimization/83957
12318 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12319 semicolon after for body surrounded by braces.
12321 PR tree-optimization/83081
12322 * profile-count.h (profile_probability::split): New method.
12323 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12324 Use profile_probability::split.
12325 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12326 when splitting a single conditional jump into 2.
12328 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12330 PR tree-optimization/69452
12331 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12334 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12336 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12337 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12338 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12340 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12342 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12344 * config/rl78/rl78.md (movdi): New define_expand.
12345 * config/rl78/rl78.c (rl78_split_movdi): New function.
12347 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12350 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12352 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12353 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12354 128-bit to produce an UNSPEC move to get the double word with the
12355 signbit and then a shift directly to do signbit.
12356 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12357 implementation with a new version that just does either a direct
12358 move or a regular move. Move memory interface to separate insns.
12359 Move insns so they are next to the expander.
12360 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12361 with signbit move. Split big and little endian case.
12362 (signbit<mode>2_dm_mem_le): Likewise.
12363 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12364 (signbit<mode>2_dm2): Likewise.
12366 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12368 * config/rl78/rl78.md (anddi3): New define_expand.
12370 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12372 * config/rl78/rl78.md (umindi3): New define_expand.
12374 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12376 * config/rl78/rl78.md (smindi3): New define_expand.
12378 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12380 * config/rl78/rl78.md (smaxdi3): New define_expand.
12382 2018-01-22 Carl Love <cel@us.ibm.com>
12384 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12385 LVX_V1TI): Add macro expansion.
12386 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12387 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12388 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12389 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12390 Change check to determine if the instruction is a byte reversing
12391 entry. Fix typo in comment.
12392 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12393 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12394 Add def_builtin calls for new builtins.
12395 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12396 Add define_insn expansion.
12398 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12400 * config/rl78/rl78.md (umaxdi3): New define_expand.
12402 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12404 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12405 for non-QImode registers.
12407 2018-01-22 Richard Biener <rguenther@suse.de>
12409 PR tree-optimization/83963
12410 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12411 including the loop exit block.
12412 (scop_detection::merge_sese): Likewise.
12413 (scop_detection::add_scop): Do it here instead.
12415 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12417 * doc/sourcebuild.texi (arm_softfloat): Document.
12419 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12422 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12423 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12424 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12426 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12427 David Edelsohn <dje.gcc@gmail.com>
12430 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12431 Change "crset eq" to "crset 2".
12432 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12433 (*call_indirect_aix<mode>_nospec): Likewise.
12434 (*call_value_indirect_aix<mode>_nospec): Likewise.
12435 (*call_indirect_elfv2<mode>_nospec): Likewise.
12436 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12437 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12438 change assembly output from . to $.
12439 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12440 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12441 (*tablejump<mode>_internal1_nospec): Likewise.
12443 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12446 * config/sh/sh_optimize_sett_clrt.cc:
12447 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12449 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12451 PR tree-optimization/83940
12452 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12453 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12454 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12455 use it to pass back the definition type.
12456 (vect_check_store_rhs): Likewise rhs_dt_out.
12457 (vect_build_gather_load_calls): Add a mask_dt argument and use
12458 it instead of a call to vect_is_simple_use.
12459 (vectorizable_store): Update calls to vect_check_load_store_mask
12460 and vect_check_store_rhs. Use the dt returned by the latter instead
12461 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12462 instead of calls to vect_is_simple_use. Pass the scalar rather
12463 than the vector operand to vect_is_simple_use when handling
12464 second and subsequent copies of an rhs value.
12465 (vectorizable_load): Update calls to vect_check_load_store_mask
12466 and vect_build_gather_load_calls. Use the cached mask_dt and
12467 gs_info.offset_dt instead of calls to vect_is_simple_use.
12469 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12471 PR middle-end/83945
12472 * tree-emutls.c: Include gimplify.h.
12473 (lower_emutls_2): New function.
12474 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12475 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12476 it before further processing.
12479 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12480 UINTVAL (trueop1) instead of INTVAL (op1).
12482 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12486 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12487 INCOMING_FRAME_SP_OFFSET if not defined.
12488 (scan_trace): Add ENTRY argument. If true and
12489 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12490 emit a note to adjust the CFA offset.
12491 (create_cfi_notes): Adjust scan_trace callers.
12492 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12493 INCOMING_FRAME_SP_OFFSET in the CIE.
12494 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12495 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12497 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12498 * doc/tm.texi: Regenerated.
12500 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12502 PR rtl-optimization/83147
12503 * lra-constraints.c (remove_inheritance_pseudos): Use
12504 lra_substitute_pseudo_within_insn.
12506 2018-01-19 Tom de Vries <tom@codesourcery.com>
12507 Cesar Philippidis <cesar@codesourcery.com>
12510 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12512 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12515 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12516 spaces for function labels.
12518 2018-01-19 Martin Liska <mliska@suse.cz>
12520 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12521 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12522 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12523 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12524 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12525 (PRED_CONST_RETURN): Change from 69 to 65.
12526 (PRED_NULL_RETURN): Change from 91 to 71.
12527 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12528 (PRED_LOOP_GUARD): Change from 66 to 73.
12530 2018-01-19 Martin Liska <mliska@suse.cz>
12532 * predict.c (predict_insn_def): Add new assert.
12533 (struct branch_predictor): Change type to signed integer.
12534 (test_prediction_value_range): Amend test to cover
12535 PROB_UNINITIALIZED.
12536 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12537 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12538 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12539 (PRED_LOOP_IV_COMPARE): Likewise.
12540 * predict.h (PROB_UNINITIALIZED): Define new constant.
12542 2018-01-19 Martin Liska <mliska@suse.cz>
12544 * predict.c (dump_prediction): Add new format for
12545 analyze_brprob.py script which is enabled with -details
12547 * profile-count.h (precise_p): New function.
12549 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12551 PR tree-optimization/83922
12552 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12553 there are no statements that need masking.
12554 (vect_active_double_reduction_p): New function.
12555 (vect_analyze_loop_operations): Use it when handling phis that
12556 are not in the loop header.
12558 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12560 PR tree-optimization/83914
12561 * tree-vect-loop.c (vectorizable_induction): Don't convert
12562 init_expr or apply the peeling adjustment for inductions
12563 that are nested within the vectorized loop.
12565 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12567 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12570 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12574 * function.h (gimplify_parameters): Add gimple_seq * argument.
12575 * function.c: Include gimple.h and options.h.
12576 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12577 for the added local temporaries if needed.
12578 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12579 if there are any parameter cleanups, wrap whole body into a
12580 try/finally with the cleanups.
12582 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12585 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12586 Use GET_MODE_CLASS for scalar floating point.
12588 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12592 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12593 Fix call of call_cgraph_insertion_hooks.
12595 2018-01-18 Martin Sebor <msebor@redhat.com>
12597 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12599 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12602 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12605 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12608 * common.opt: (-ffile-prefix-map): New option.
12609 * opts.c (common_handle_option): Defer it.
12610 * opts-global.c (handle_common_deferred_options): Handle it.
12611 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12612 * file-prefix-map.h: New file.
12613 (remap_debug_filename, add_debug_prefix_map): ...here.
12614 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12615 * final.c (debug_prefix_map, add_debug_prefix_map
12616 remap_debug_filename): Move to...
12617 * file-prefix-map.c: New file.
12618 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12619 generalize, get rid of alloca(), use strrchr() instead of strchr().
12620 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12621 Implement in terms of add_prefix_map().
12622 (remap_macro_filename, remap_debug_filename): Implement in term of
12624 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12625 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12626 * dbxout.c: Include file-prefix-map.h.
12627 * varasm.c: Likewise.
12628 * vmsdbgout.c: Likewise.
12629 * xcoffout.c: Likewise.
12630 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12631 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12632 * doc/invoke.texi (-ffile-prefix-map): Document.
12633 (-fdebug-prefix-map): Update description.
12635 2018-01-18 Martin Liska <mliska@suse.cz>
12637 * config/i386/i386.c (indirect_thunk_name): Document that also
12639 (output_indirect_thunk): Document why both instructions
12640 (pause and lfence) are generated.
12642 2018-01-18 Richard Biener <rguenther@suse.de>
12644 PR tree-optimization/83887
12645 * graphite-scop-detection.c
12646 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12647 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12648 (scop_detection::merge_sese): Re-implement with a flood-fill
12649 algorithm that properly finds a SESE region if it exists.
12651 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12654 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12655 pointer_diff optimizations use view_convert instead of convert.
12657 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12659 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12660 Generate different code for -mno-speculate-indirect-jumps.
12661 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12662 (*call_indirect_aix<mode>): Disable for
12663 -mno-speculate-indirect-jumps.
12664 (*call_indirect_aix<mode>_nospec): New define_insn.
12665 (*call_value_indirect_aix<mode>): Disable for
12666 -mno-speculate-indirect-jumps.
12667 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12668 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12669 -mno-speculate-indirect-jumps.
12670 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12672 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12674 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12675 long double type, set the flags for noting the default long double
12676 type, even if we don't pass or return a long double type.
12678 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12681 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12684 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12686 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12687 support for merge[hl].
12688 (fold_mergehl_helper): New helper function.
12689 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12690 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12691 (altivec_vmrglw_direct): Add xxmrglw insn.
12693 2018-01-17 Andrew Waterman <andrew@sifive.com>
12695 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12696 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12698 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12701 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12702 call the lto_location_cache before reading the
12703 DECL_SOURCE_LOCATION of the types.
12705 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12706 Richard Sandiford <richard.sandiford@linaro.org>
12708 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12709 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12710 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12711 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12712 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12714 * config/aarch64/constraints.md (aarch64_movti_operand):
12716 * config/aarch64/predicates.md (Uti): Add new constraint.
12718 2018-01-17 Carl Love <cel@us.ibm.com>
12720 * config/rs6000/vsx.md (define_expand xl_len_r,
12721 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12722 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12724 (define_expand, define_insn): Move the shift left from the
12725 define_insn to the define_expand for lxvl and stxvl instructions.
12726 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12727 and XL_LEN_R definitions to PURE.
12729 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12731 * config/i386/i386.c (indirect_thunk_name): Declare regno
12732 as unsigned int. Compare regno with INVALID_REGNUM.
12733 (output_indirect_thunk): Ditto.
12734 (output_indirect_thunk_function): Ditto.
12735 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12736 in the call to output_indirect_thunk_function.
12738 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12740 PR middle-end/83884
12741 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12742 rather than the size of inner_type to determine the stack slot size
12743 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12745 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12748 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12751 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12753 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12754 endian Linux systems to optionally enable multilibs for selecting
12755 the long double type if the user configured an explicit type.
12756 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12757 have no long double multilibs if not defined.
12758 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12759 warn if the user used -mabi={ieee,ibm}longdouble and we built
12760 multilibs for long double.
12761 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12762 appropriate multilib option.
12763 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12765 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12766 for building long double multilibs.
12767 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12769 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12771 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12774 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12776 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12779 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12782 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12785 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12787 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12788 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12790 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12792 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12793 different rtl trees depending on TARGET_64BIT.
12794 (rs6000_gen_lvx): Likewise.
12796 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12798 * config/visium/visium.md (nop): Tweak comment.
12799 (hazard_nop): Likewise.
12801 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12803 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12804 -mspeculate-indirect-jumps.
12805 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12806 for -mno-speculate-indirect-jumps.
12807 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12808 (*call_value_indirect_elfv2<mode>): Disable for
12809 -mno-speculate-indirect-jumps.
12810 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12811 (indirect_jump): Emit different RTL for
12812 -mno-speculate-indirect-jumps.
12813 (*indirect_jump<mode>): Disable for
12814 -mno-speculate-indirect-jumps.
12815 (*indirect_jump<mode>_nospec): New define_insn.
12816 (tablejump): Emit different RTL for
12817 -mno-speculate-indirect-jumps.
12818 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12819 (tablejumpsi_nospec): New define_expand.
12820 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12821 (tablejumpdi_nospec): New define_expand.
12822 (*tablejump<mode>_internal1): Disable for
12823 -mno-speculate-indirect-jumps.
12824 (*tablejump<mode>_internal1_nospec): New define_insn.
12825 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12828 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12830 * caller-save.c (insert_save): Drop unnecessary parameter. All
12833 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12834 Richard Biener <rguenth@suse.de>
12837 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12838 return early, inline manually is_gimple_sizepos. Make sure if we
12839 call gimplify_expr we don't end up with a gimple constant.
12840 * tree.c (variably_modified_type_p): Don't return true for
12841 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12842 * gimplify.h (is_gimple_sizepos): Remove.
12844 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12846 PR tree-optimization/83857
12847 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12848 vectorizable_live_operation for pure SLP statements.
12849 (vectorizable_live_operation): Handle PHIs.
12851 2018-01-16 Richard Biener <rguenther@suse.de>
12853 PR tree-optimization/83867
12854 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12855 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12857 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12860 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12861 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12862 If off is not INTEGER_CST, issue a may not be aligned warning
12863 rather than isn't aligned. Use isn%'t rather than isn't.
12864 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12866 <case MULT_EXPR>: Improve the case when bottom and one of the
12867 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12868 operand, in that case check if the other operand is multiple of
12869 bottom divided by the INTEGER_CST operand.
12871 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12874 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12875 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12876 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12877 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12878 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12879 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12880 (pa_function_arg_size): New function.
12882 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12884 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12885 in a separate statement.
12887 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12889 PR tree-optimization/83847
12890 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12891 group gathers and scatters.
12893 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12895 PR rtl-optimization/86620
12896 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12898 PR rtl-optimization/83213
12899 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12900 to last if both are JUMP_INSNs.
12902 PR tree-optimization/83843
12903 * gimple-ssa-store-merging.c
12904 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12905 store_immediate_info for bswap/nop orig_stores.
12907 2018-01-15 Andrew Waterman <andrew@sifive.com>
12909 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12911 <UDIV>: Increase cost if !TARGET_DIV.
12913 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12915 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12916 (define_attr "cr_logical_3op"): New.
12917 (cceq_ior_compare): Adjust.
12918 (cceq_ior_compare_complement): Adjust.
12919 (*cceq_rev_compare): Adjust.
12920 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12921 (is_cracked_insn): Adjust.
12922 (insn_must_be_first_in_group): Adjust.
12923 * config/rs6000/40x.md: Adjust.
12924 * config/rs6000/440.md: Adjust.
12925 * config/rs6000/476.md: Adjust.
12926 * config/rs6000/601.md: Adjust.
12927 * config/rs6000/603.md: Adjust.
12928 * config/rs6000/6xx.md: Adjust.
12929 * config/rs6000/7450.md: Adjust.
12930 * config/rs6000/7xx.md: Adjust.
12931 * config/rs6000/8540.md: Adjust.
12932 * config/rs6000/cell.md: Adjust.
12933 * config/rs6000/e300c2c3.md: Adjust.
12934 * config/rs6000/e500mc.md: Adjust.
12935 * config/rs6000/e500mc64.md: Adjust.
12936 * config/rs6000/e5500.md: Adjust.
12937 * config/rs6000/e6500.md: Adjust.
12938 * config/rs6000/mpc.md: Adjust.
12939 * config/rs6000/power4.md: Adjust.
12940 * config/rs6000/power5.md: Adjust.
12941 * config/rs6000/power6.md: Adjust.
12942 * config/rs6000/power7.md: Adjust.
12943 * config/rs6000/power8.md: Adjust.
12944 * config/rs6000/power9.md: Adjust.
12945 * config/rs6000/rs64.md: Adjust.
12946 * config/rs6000/titan.md: Adjust.
12948 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12950 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12951 ix86_indirect_branch_register logic.
12953 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12955 * config/i386/constraints.md (Bs): Update
12956 ix86_indirect_branch_register check. Don't check
12957 ix86_indirect_branch_register with GOT_memory_operand.
12959 * config/i386/predicates.md (GOT_memory_operand): Don't check
12960 ix86_indirect_branch_register here.
12961 (GOT32_symbol_operand): Likewise.
12963 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12965 * config/i386/predicates.md (constant_call_address_operand):
12966 Rewrite ix86_indirect_branch_register logic.
12967 (sibcall_insn_operand): Likewise.
12969 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12971 * config/i386/constraints.md (Bs): Replace
12972 ix86_indirect_branch_thunk_register with
12973 ix86_indirect_branch_register.
12975 * config/i386/i386.md (indirect_jump): Likewise.
12976 (tablejump): Likewise.
12977 (*sibcall_memory): Likewise.
12978 (*sibcall_value_memory): Likewise.
12979 Peepholes of indirect call and jump via memory: Likewise.
12980 * config/i386/i386.opt: Likewise.
12981 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12982 (GOT_memory_operand): Likewise.
12983 (call_insn_operand): Likewise.
12984 (sibcall_insn_operand): Likewise.
12985 (GOT32_symbol_operand): Likewise.
12987 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12989 PR middle-end/83837
12990 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12991 type rather than type addr's type points to.
12992 (expand_omp_atomic_mutex): Likewise.
12993 (expand_omp_atomic): Likewise.
12995 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12998 * config/i386/i386.c (output_indirect_thunk_function): Use
12999 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13000 for __x86_return_thunk.
13002 2018-01-15 Richard Biener <rguenther@suse.de>
13004 PR middle-end/83850
13005 * expmed.c (extract_bit_field_1): Fix typo.
13007 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13010 * config/arm/iterators.md (VF): New mode iterator.
13011 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13012 Remove integer-related logic from pattern.
13013 (neon_vabd<mode>_3): Likewise.
13015 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13017 PR middle-end/82694
13018 * common.opt (fstrict-overflow): No longer an alias.
13019 (fwrapv-pointer): New option.
13020 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13021 also for pointer types based on flag_wrapv_pointer.
13022 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13023 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13024 opts->x_flag_wrapv got set.
13025 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13026 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13027 POINTER_TYPE_OVERFLOW_UNDEFINED.
13028 * match.pd: Likewise in address comparison pattern.
13029 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13031 2018-01-15 Richard Biener <rguenther@suse.de>
13034 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13035 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13036 Reset type names to their identifier if their TYPE_DECL doesn't
13037 have linkage (and thus is used for ODR and devirt).
13038 (save_debug_info_for_decl): Remove.
13039 (save_debug_info_for_type): Likewise.
13040 (add_tree_to_fld_list): Adjust.
13041 * tree-pretty-print.c (dump_generic_node): Make dumping of
13042 type names more robust.
13044 2018-01-15 Richard Biener <rguenther@suse.de>
13046 * BASE-VER: Bump to 8.0.1.
13048 2018-01-14 Martin Sebor <msebor@redhat.com>
13051 * builtins.c (check_access): Avoid warning when the no-warning bit
13054 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13056 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13057 * ira-color (allocno_hard_regs_compare): Likewise.
13059 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13062 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13063 Use .pushsection/.popsection.
13065 2018-01-14 Martin Sebor <msebor@redhat.com>
13068 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13070 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13072 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13073 entry from extra_headers.
13074 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13075 extra_headers, make the list bitwise identical to the i?86-*-* one.
13077 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13079 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13080 -mcmodel=large with -mindirect-branch=thunk,
13081 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13082 -mfunction-return=thunk-extern.
13083 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13084 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13085 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13087 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13089 * config/i386/i386.c (print_reg): Print the name of the full
13090 integer register without '%'.
13091 (ix86_print_operand): Handle 'V'.
13092 * doc/extend.texi: Document 'V' modifier.
13094 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13096 * config/i386/constraints.md (Bs): Disallow memory operand for
13097 -mindirect-branch-register.
13099 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13100 (GOT_memory_operand): Likewise.
13101 (call_insn_operand): Likewise.
13102 (sibcall_insn_operand): Likewise.
13103 (GOT32_symbol_operand): Likewise.
13104 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13105 for -mindirect-branch-register.
13106 (tablejump): Likewise.
13107 (*sibcall_memory): Likewise.
13108 (*sibcall_value_memory): Likewise.
13109 Disallow peepholes of indirect call and jump via memory for
13110 -mindirect-branch-register.
13111 (*call_pop): Replace m with Bw.
13112 (*call_value_pop): Likewise.
13113 (*sibcall_pop_memory): Replace m with Bs.
13114 * config/i386/i386.opt (mindirect-branch-register): New option.
13115 * doc/invoke.texi: Document -mindirect-branch-register option.
13117 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13119 * config/i386/i386-protos.h (ix86_output_function_return): New.
13120 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13121 set function_return_type.
13122 (indirect_thunk_name): Add ret_p to indicate thunk for function
13124 (output_indirect_thunk_function): Pass false to
13125 indirect_thunk_name.
13126 (ix86_output_indirect_branch_via_reg): Likewise.
13127 (ix86_output_indirect_branch_via_push): Likewise.
13128 (output_indirect_thunk_function): Create alias for function
13129 return thunk if regno < 0.
13130 (ix86_output_function_return): New function.
13131 (ix86_handle_fndecl_attribute): Handle function_return.
13132 (ix86_attribute_table): Add function_return.
13133 * config/i386/i386.h (machine_function): Add
13134 function_return_type.
13135 * config/i386/i386.md (simple_return_internal): Use
13136 ix86_output_function_return.
13137 (simple_return_internal_long): Likewise.
13138 * config/i386/i386.opt (mfunction-return=): New option.
13139 (indirect_branch): Mention -mfunction-return=.
13140 * doc/extend.texi: Document function_return function attribute.
13141 * doc/invoke.texi: Document -mfunction-return= option.
13143 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13145 * config/i386/i386-opts.h (indirect_branch): New.
13146 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13147 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13148 with local indirect jump when converting indirect call and jump.
13149 (ix86_set_indirect_branch_type): New.
13150 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13151 (indirectlabelno): New.
13152 (indirect_thunk_needed): Likewise.
13153 (indirect_thunk_bnd_needed): Likewise.
13154 (indirect_thunks_used): Likewise.
13155 (indirect_thunks_bnd_used): Likewise.
13156 (INDIRECT_LABEL): Likewise.
13157 (indirect_thunk_name): Likewise.
13158 (output_indirect_thunk): Likewise.
13159 (output_indirect_thunk_function): Likewise.
13160 (ix86_output_indirect_branch_via_reg): Likewise.
13161 (ix86_output_indirect_branch_via_push): Likewise.
13162 (ix86_output_indirect_branch): Likewise.
13163 (ix86_output_indirect_jmp): Likewise.
13164 (ix86_code_end): Call output_indirect_thunk_function if needed.
13165 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13167 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13168 (ix86_attribute_table): Add indirect_branch.
13169 * config/i386/i386.h (machine_function): Add indirect_branch_type
13170 and has_local_indirect_jump.
13171 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13173 (tablejump): Likewise.
13174 (*indirect_jump): Use ix86_output_indirect_jmp.
13175 (*tablejump_1): Likewise.
13176 (simple_return_indirect_internal): Likewise.
13177 * config/i386/i386.opt (mindirect-branch=): New option.
13178 (indirect_branch): New.
13181 (thunk-inline): Likewise.
13182 (thunk-extern): Likewise.
13183 * doc/extend.texi: Document indirect_branch function attribute.
13184 * doc/invoke.texi: Document -mindirect-branch= option.
13186 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13189 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13191 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13193 * ipa-inline.c (want_inline_small_function_p): Return false if
13194 inlining has already failed with CIF_FINAL_ERROR.
13195 (update_caller_keys): Call want_inline_small_function_p before
13197 (update_callee_keys): Likewise.
13199 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13201 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13203 (rs6000_quadword_masked_address_p): Likewise.
13204 (quad_aligned_load_p): Likewise.
13205 (quad_aligned_store_p): Likewise.
13206 (const_load_sequence_p): Add comment to describe the outer-most loop.
13207 (mimic_memory_attributes_and_flags): New function.
13208 (rs6000_gen_stvx): Likewise.
13209 (replace_swapped_aligned_store): Likewise.
13210 (rs6000_gen_lvx): Likewise.
13211 (replace_swapped_aligned_load): Likewise.
13212 (replace_swapped_load_constant): Capitalize argument name in
13213 comment describing this function.
13214 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13215 and stores that access quad-word aligned addresses and replace
13216 with stvx or lvx instructions when appropriate.
13217 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13218 New function prototype.
13219 (rs6000_quadword_masked_address_p): Likewise.
13220 (rs6000_gen_lvx): Likewise.
13221 (rs6000_gen_stvx): Likewise.
13222 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13223 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13224 when memory address is aligned.
13225 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13226 this split to select lvx instruction when memory address is aligned.
13227 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13228 instruction when memory address is aligned.
13229 (*vsx_le_perm_load_v16qi): Likewise.
13230 (four unnamed splitters): Modify to select the stvx instruction
13231 when memory is aligned.
13233 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13235 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13236 which appears in the queue multiple times.
13238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13239 Alan Hayward <alan.hayward@arm.com>
13240 David Sherwood <david.sherwood@arm.com>
13242 * tree-vectorizer.h (vec_lower_bound): New structure.
13243 (_loop_vec_info): Add check_nonzero and lower_bounds.
13244 (LOOP_VINFO_CHECK_NONZERO): New macro.
13245 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13246 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13247 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13248 fields. Make seg_len the distance travelled, not including the
13250 (dr_direction_indicator): Declare.
13251 (dr_zero_step_indicator): Likewise.
13252 (dr_known_forward_stride_p): Likewise.
13253 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13255 (runtime_alias_check_p): Allow runtime alias checks with
13257 (operator ==): Compare access_size and align.
13258 (prune_runtime_alias_test_list): Rework for new distinction between
13259 the access_size and seg_len.
13260 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13262 (get_segment_min_max): New function.
13263 (create_intersect_range_checks): Use it.
13264 (dr_step_indicator): New function.
13265 (dr_direction_indicator): Likewise.
13266 (dr_zero_step_indicator): Likewise.
13267 (dr_known_forward_stride_p): Likewise.
13268 * tree-loop-distribution.c (data_ref_segment_size): Return
13269 DR_STEP * (niters - 1).
13270 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13272 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13273 (vect_preserves_scalar_order_p): New function, split out from...
13274 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13275 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13276 (vect_vfa_access_size): New function.
13277 (vect_vfa_align): Likewise.
13278 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13279 (dump_lower_bound): New function.
13280 (vect_check_lower_bound): Likewise.
13281 (vect_small_gap_p): Likewise.
13282 (vectorizable_with_step_bound_p): Likewise.
13283 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13284 depencies if the vectorization factor is 1. Convert the checks
13285 for nonzero steps into checks on the bounds of DR_STEP. Try using
13286 a bunds check for variable steps if the minimum required step is
13287 relatively small. Update calls to the dr_with_seg_len
13288 constructor and to vect_compile_time_alias.
13289 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13291 (vect_loop_versioning): Call it.
13292 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13294 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13297 Alan Hayward <alan.hayward@arm.com>
13298 David Sherwood <david.sherwood@arm.com>
13300 * doc/sourcebuild.texi (vect_scatter_store): Document.
13301 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13303 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13305 * genopinit.c (main): Add supports_vec_scatter_store and
13306 supports_vec_scatter_store_cached to target_optabs.
13307 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13308 IFN_MASK_SCATTER_STORE.
13309 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13311 * internal-fn.h (internal_store_fn_p): Declare.
13312 (internal_fn_stored_value_index): Likewise.
13313 * internal-fn.c (scatter_store_direct): New macro.
13314 (expand_scatter_store_optab_fn): New function.
13315 (direct_scatter_store_optab_supported_p): New macro.
13316 (internal_store_fn_p): New function.
13317 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13318 IFN_MASK_SCATTER_STORE.
13319 (internal_fn_mask_index): Likewise.
13320 (internal_fn_stored_value_index): New function.
13321 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13322 for scatter stores.
13323 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13324 * optabs-query.c (supports_vec_scatter_store_p): New function.
13325 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13326 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13327 true for scatter stores.
13328 (vect_gather_scatter_fn_p): Handle scatter stores too.
13329 (vect_check_gather_scatter): Consider using scatter stores if
13330 supports_vec_scatter_store_p.
13331 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13332 scatter stores too.
13333 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13334 internal_fn_stored_value_index.
13335 (check_load_store_masking): Handle scatter stores too.
13336 (vect_get_store_rhs): Make public.
13337 (vectorizable_call): Use internal_store_fn_p.
13338 (vectorizable_store): Handle scatter store internal functions.
13339 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13340 when deciding whether the end of the group has been reached.
13341 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13342 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13343 (mask_scatter_store<mode>): New insns.
13345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13346 Alan Hayward <alan.hayward@arm.com>
13347 David Sherwood <david.sherwood@arm.com>
13349 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13350 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13351 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13353 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13354 Use vect_truncate_gather_scatter_offset if we can't treat the
13355 operation as a normal gather load or scatter store.
13356 (get_group_load_store_type): Take the gather_scatter_info
13357 as argument. Try using a gather load or scatter store for
13358 single-element groups.
13359 (get_load_store_type): Update calls to get_group_load_store_type
13360 and vect_use_strided_gather_scatters_p.
13362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13363 Alan Hayward <alan.hayward@arm.com>
13364 David Sherwood <david.sherwood@arm.com>
13366 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13367 optional tree argument.
13368 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13370 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13371 but continue to use the current value as a fallback.
13372 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13373 to compare the updates.
13374 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13375 (get_load_store_type): Use it when handling a strided access.
13376 (vect_get_strided_load_store_ops): New function.
13377 (vect_get_data_ptr_increment): Likewise.
13378 (vectorizable_load): Handle strided gather loads. Always pass
13379 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13382 Alan Hayward <alan.hayward@arm.com>
13383 David Sherwood <david.sherwood@arm.com>
13385 * doc/md.texi (gather_load@var{m}): Document.
13386 (mask_gather_load@var{m}): Likewise.
13387 * genopinit.c (main): Add supports_vec_gather_load and
13388 supports_vec_gather_load_cached to target_optabs.
13389 * optabs-tree.c (init_tree_optimization_optabs): Use
13390 ggc_cleared_alloc to allocate target_optabs.
13391 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13392 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13394 * internal-fn.h (internal_load_fn_p): Declare.
13395 (internal_gather_scatter_fn_p): Likewise.
13396 (internal_fn_mask_index): Likewise.
13397 (internal_gather_scatter_fn_supported_p): Likewise.
13398 * internal-fn.c (gather_load_direct): New macro.
13399 (expand_gather_load_optab_fn): New function.
13400 (direct_gather_load_optab_supported_p): New macro.
13401 (direct_internal_fn_optab): New function.
13402 (internal_load_fn_p): Likewise.
13403 (internal_gather_scatter_fn_p): Likewise.
13404 (internal_fn_mask_index): Likewise.
13405 (internal_gather_scatter_fn_supported_p): Likewise.
13406 * optabs-query.c (supports_at_least_one_mode_p): New function.
13407 (supports_vec_gather_load_p): Likewise.
13408 * optabs-query.h (supports_vec_gather_load_p): Declare.
13409 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13410 and memory_type field.
13411 (NUM_PATTERNS): Bump to 15.
13412 * tree-vect-data-refs.c: Include internal-fn.h.
13413 (vect_gather_scatter_fn_p): New function.
13414 (vect_describe_gather_scatter_call): Likewise.
13415 (vect_check_gather_scatter): Try using internal functions for
13416 gather loads. Recognize existing calls to a gather load function.
13417 (vect_analyze_data_refs): Consider using gather loads if
13418 supports_vec_gather_load_p.
13419 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13420 (vect_get_gather_scatter_offset_type): Likewise.
13421 (vect_convert_mask_for_vectype): Likewise.
13422 (vect_add_conversion_to_patterm): Likewise.
13423 (vect_try_gather_scatter_pattern): Likewise.
13424 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13425 (vect_vect_recog_func_ptrs): Add it.
13426 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13427 internal_fn_mask_index and internal_gather_scatter_fn_p.
13428 (check_load_store_masking): Take the gather_scatter_info as an
13429 argument and handle gather loads.
13430 (vect_get_gather_scatter_ops): New function.
13431 (vectorizable_call): Check internal_load_fn_p.
13432 (vectorizable_load): Likewise. Handle gather load internal
13434 (vectorizable_store): Update call to check_load_store_masking.
13435 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13436 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13437 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13438 (aarch64_gather_scale_operand_d): New predicates.
13439 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13440 (mask_gather_load<mode>): New insns.
13442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13443 Alan Hayward <alan.hayward@arm.com>
13444 David Sherwood <david.sherwood@arm.com>
13446 * optabs.def (fold_left_plus_optab): New optab.
13447 * doc/md.texi (fold_left_plus_@var{m}): Document.
13448 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13449 * internal-fn.c (fold_left_direct): Define.
13450 (expand_fold_left_optab_fn): Likewise.
13451 (direct_fold_left_optab_supported_p): Likewise.
13452 * fold-const-call.c (fold_const_fold_left): New function.
13453 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13454 * tree-parloops.c (valid_reduction_p): New function.
13455 (gather_scalar_reductions): Use it.
13456 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13457 (vect_finish_replace_stmt): Declare.
13458 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13459 (needs_fold_left_reduction_p): New function, split out from...
13460 (vect_is_simple_reduction): ...here. Accept reductions that
13461 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13462 (vect_force_simple_reduction): Also store the reduction type in
13463 the assignment's STMT_VINFO_REDUC_TYPE.
13464 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13465 (merge_with_identity): New function.
13466 (vect_expand_fold_left): Likewise.
13467 (vectorize_fold_left_reduction): Likewise.
13468 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13469 scalar phi in place for it. Check for target support and reject
13470 cases that would reassociate the operation. Defer the transform
13471 phase to vectorize_fold_left_reduction.
13472 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13473 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13474 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13478 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13479 call to ifc_temp_var.
13481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13482 Alan Hayward <alan.hayward@arm.com>
13483 David Sherwood <david.sherwood@arm.com>
13485 * target.def (legitimize_address_displacement): Take the original
13486 offset as a poly_int.
13487 * targhooks.h (default_legitimize_address_displacement): Update
13489 * targhooks.c (default_legitimize_address_displacement): Likewise.
13490 * doc/tm.texi: Regenerate.
13491 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13492 as an argument, moving assert of ad->disp == ad->disp_term to...
13493 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13494 Try calling targetm.legitimize_address_displacement before expanding
13495 the address rather than afterwards, and adjust for the new interface.
13496 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13497 Match the new hook interface. Handle SVE addresses.
13498 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13499 new hook interface.
13501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13503 * Makefile.in (OBJS): Add early-remat.o.
13504 * target.def (select_early_remat_modes): New hook.
13505 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13506 * doc/tm.texi: Regenerate.
13507 * targhooks.h (default_select_early_remat_modes): Declare.
13508 * targhooks.c (default_select_early_remat_modes): New function.
13509 * timevar.def (TV_EARLY_REMAT): New timevar.
13510 * passes.def (pass_early_remat): New pass.
13511 * tree-pass.h (make_pass_early_remat): Declare.
13512 * early-remat.c: New file.
13513 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13515 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13518 Alan Hayward <alan.hayward@arm.com>
13519 David Sherwood <david.sherwood@arm.com>
13521 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13522 vfm1 with a bound_epilog parameter.
13523 (vect_do_peeling): Update calls accordingly, and move the prologue
13524 call earlier in the function. Treat the base bound_epilog as 0 for
13525 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13526 this base when peeling for gaps.
13527 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13528 with fully-masked loops.
13529 (vect_estimate_min_profitable_iters): Handle the single peeled
13530 iteration in that case.
13532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13533 Alan Hayward <alan.hayward@arm.com>
13534 David Sherwood <david.sherwood@arm.com>
13536 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13537 single-element interleaving even if the size is not a power of 2.
13538 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13539 accesses for single-element interleaving if the group size is
13542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13543 Alan Hayward <alan.hayward@arm.com>
13544 David Sherwood <david.sherwood@arm.com>
13546 * doc/md.texi (fold_extract_last_@var{m}): Document.
13547 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13548 * optabs.def (fold_extract_last_optab): New optab.
13549 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13550 * internal-fn.c (fold_extract_direct): New macro.
13551 (expand_fold_extract_optab_fn): Likewise.
13552 (direct_fold_extract_optab_supported_p): Likewise.
13553 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13554 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13555 EXTRACT_LAST_REDUCTION.
13556 (get_initial_def_for_reduction): Do not create an initial vector
13557 for EXTRACT_LAST_REDUCTION reductions.
13558 (vectorizable_reduction): Leave the scalar phi in place for
13559 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13560 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13561 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13562 transform phase to vectorizable_condition.
13563 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13565 (vect_finish_stmt_generation): ...here.
13566 (vect_finish_replace_stmt): New function.
13567 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13568 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13570 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13573 Alan Hayward <alan.hayward@arm.com>
13574 David Sherwood <david.sherwood@arm.com>
13576 * doc/md.texi (extract_last_@var{m}): Document.
13577 * optabs.def (extract_last_optab): New optab.
13578 * internal-fn.def (EXTRACT_LAST): New internal function.
13579 * internal-fn.c (cond_unary_direct): New macro.
13580 (expand_cond_unary_optab_fn): Likewise.
13581 (direct_cond_unary_optab_supported_p): Likewise.
13582 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13583 loops using EXTRACT_LAST.
13584 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13585 (extract_last_<mode>): ...this optab.
13586 (vec_extract<mode><Vel>): Update accordingly.
13588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13589 Alan Hayward <alan.hayward@arm.com>
13590 David Sherwood <david.sherwood@arm.com>
13592 * target.def (empty_mask_is_expensive): New hook.
13593 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13594 * doc/tm.texi: Regenerate.
13595 * targhooks.h (default_empty_mask_is_expensive): Declare.
13596 * targhooks.c (default_empty_mask_is_expensive): New function.
13597 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13598 if the target says that empty masks are expensive.
13599 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13601 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13604 Alan Hayward <alan.hayward@arm.com>
13605 David Sherwood <david.sherwood@arm.com>
13607 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13608 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13609 (vect_use_loop_mask_for_alignment_p): New function.
13610 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13611 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13612 niters_skip argument. Make sure that the first niters_skip elements
13613 of the first iteration are inactive.
13614 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13615 Update call to vect_set_loop_masks_directly.
13616 (get_misalign_in_elems): New function, split out from...
13617 (vect_gen_prolog_loop_niters): ...here.
13618 (vect_update_init_of_dr): Take a code argument that specifies whether
13619 the adjustment should be added or subtracted.
13620 (vect_update_init_of_drs): Likewise.
13621 (vect_prepare_for_masked_peels): New function.
13622 (vect_do_peeling): Skip prologue peeling if we're using a mask
13623 instead. Update call to vect_update_inits_of_drs.
13624 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13626 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13627 alignment. Do not include the number of peeled iterations in
13628 the minimum threshold in that case.
13629 (vectorizable_induction): Adjust the start value down by
13630 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13631 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13632 Take the number of skipped iterations into account when calculating
13634 * tree-vect-stmts.c (vect_gen_while_not): New function.
13636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13637 Alan Hayward <alan.hayward@arm.com>
13638 David Sherwood <david.sherwood@arm.com>
13640 * doc/sourcebuild.texi (vect_fully_masked): Document.
13641 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13642 default value to 0.
13643 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13645 (vect_analyze_loop_2): ...here. Don't check the vectorization
13646 factor against the number of loop iterations if the loop is
13649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13650 Alan Hayward <alan.hayward@arm.com>
13651 David Sherwood <david.sherwood@arm.com>
13653 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13654 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13655 (dump_groups): Update accordingly.
13656 (iv_use::mem_type): New member variable.
13657 (address_p): New function.
13658 (record_use): Add a mem_type argument and initialize the new
13660 (record_group_use): Add a mem_type argument. Use address_p.
13661 Remove obsolete null checks of base_object. Update call to record_use.
13662 (find_interesting_uses_op): Update call to record_group_use.
13663 (find_interesting_uses_cond): Likewise.
13664 (find_interesting_uses_address): Likewise.
13665 (get_mem_type_for_internal_fn): New function.
13666 (find_address_like_use): Likewise.
13667 (find_interesting_uses_stmt): Try find_address_like_use before
13668 calling find_interesting_uses_op.
13669 (addr_offset_valid_p): Use the iv mem_type field as the type
13670 of the addressed memory.
13671 (add_autoinc_candidates): Likewise.
13672 (get_address_cost): Likewise.
13673 (split_small_address_groups_p): Use address_p.
13674 (split_address_groups): Likewise.
13675 (add_iv_candidate_for_use): Likewise.
13676 (autoinc_possible_for_pair): Likewise.
13677 (rewrite_groups): Likewise.
13678 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13679 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13680 (get_alias_ptr_type_for_ptr_address): New function.
13681 (rewrite_use_address): Rewrite address uses in calls that were
13682 identified by find_address_like_use.
13684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13685 Alan Hayward <alan.hayward@arm.com>
13686 David Sherwood <david.sherwood@arm.com>
13688 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13690 * gimple-expr.h (is_gimple_addressable: Likewise.
13691 * gimple-expr.c (is_gimple_address): Likewise.
13692 * internal-fn.c (expand_call_mem_ref): New function.
13693 (expand_mask_load_optab_fn): Use it.
13694 (expand_mask_store_optab_fn): Likewise.
13696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13697 Alan Hayward <alan.hayward@arm.com>
13698 David Sherwood <david.sherwood@arm.com>
13700 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13701 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13702 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13703 (cond_umax@var{mode}): Document.
13704 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13705 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13706 (cond_umin_optab, cond_umax_optab): New optabs.
13707 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13708 (COND_IOR, COND_XOR): New internal functions.
13709 * internal-fn.h (get_conditional_internal_fn): Declare.
13710 * internal-fn.c (cond_binary_direct): New macro.
13711 (expand_cond_binary_optab_fn): Likewise.
13712 (direct_cond_binary_optab_supported_p): Likewise.
13713 (get_conditional_internal_fn): New function.
13714 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13715 Cope with reduction statements that are vectorized as calls rather
13717 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13718 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13719 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13720 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13721 (UNSPEC_COND_EOR): New unspecs.
13722 (optab): Add mappings for them.
13723 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13724 (sve_int_op, sve_fp_op): New int attributes.
13726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13727 Alan Hayward <alan.hayward@arm.com>
13728 David Sherwood <david.sherwood@arm.com>
13730 * optabs.def (while_ult_optab): New optab.
13731 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13732 * internal-fn.def (WHILE_ULT): New internal function.
13733 * internal-fn.h (direct_internal_fn_supported_p): New override
13734 that takes two types as argument.
13735 * internal-fn.c (while_direct): New macro.
13736 (expand_while_optab_fn): New function.
13737 (convert_optab_supported_p): Likewise.
13738 (direct_while_optab_supported_p): New macro.
13739 * wide-int.h (wi::udiv_ceil): New function.
13740 * tree-vectorizer.h (rgroup_masks): New structure.
13741 (vec_loop_masks): New typedef.
13742 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13743 and fully_masked_p.
13744 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13745 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13746 (vect_max_vf): New function.
13747 (slpeel_make_loop_iterate_ntimes): Delete.
13748 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13749 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13750 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13751 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13752 internal-fn.h, stor-layout.h and optabs-query.h.
13753 (vect_set_loop_mask): New function.
13754 (add_preheader_seq): Likewise.
13755 (add_header_seq): Likewise.
13756 (interleave_supported_p): Likewise.
13757 (vect_maybe_permute_loop_masks): Likewise.
13758 (vect_set_loop_masks_directly): Likewise.
13759 (vect_set_loop_condition_masked): Likewise.
13760 (vect_set_loop_condition_unmasked): New function, split out from
13761 slpeel_make_loop_iterate_ntimes.
13762 (slpeel_make_loop_iterate_ntimes): Rename to..
13763 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13764 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13765 (vect_do_peeling): Update call accordingly.
13766 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13768 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13769 mask_compare_type, can_fully_mask_p and fully_masked_p.
13770 (release_vec_loop_masks): New function.
13771 (_loop_vec_info): Use it to free the loop masks.
13772 (can_produce_all_loop_masks_p): New function.
13773 (vect_get_max_nscalars_per_iter): Likewise.
13774 (vect_verify_full_masking): Likewise.
13775 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13776 retries, and free the mask rgroups before retrying. Check loop-wide
13777 reasons for disallowing fully-masked loops. Make the final decision
13778 about whether use a fully-masked loop or not.
13779 (vect_estimate_min_profitable_iters): Do not assume that peeling
13780 for the number of iterations will be needed for fully-masked loops.
13781 (vectorizable_reduction): Disable fully-masked loops.
13782 (vectorizable_live_operation): Likewise.
13783 (vect_halve_mask_nunits): New function.
13784 (vect_double_mask_nunits): Likewise.
13785 (vect_record_loop_mask): Likewise.
13786 (vect_get_loop_mask): Likewise.
13787 (vect_transform_loop): Handle the case in which the final loop
13788 iteration might handle a partial vector. Call vect_set_loop_condition
13789 instead of slpeel_make_loop_iterate_ntimes.
13790 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13791 (check_load_store_masking): New function.
13792 (prepare_load_store_mask): Likewise.
13793 (vectorizable_store): Handle fully-masked loops.
13794 (vectorizable_load): Likewise.
13795 (supportable_widening_operation): Use vect_halve_mask_nunits for
13797 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13798 (vect_gen_while): New function.
13799 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13800 (aarch64_uqdec<mode>): New insn.
13802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13803 Alan Hayward <alan.hayward@arm.com>
13804 David Sherwood <david.sherwood@arm.com>
13806 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13807 (reduc_xor_scal_optab): New optabs.
13808 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13809 (reduc_xor_scal_@var{m}): Document.
13810 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13811 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13812 internal functions.
13813 * fold-const-call.c (fold_const_call): Handle them.
13814 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13815 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13816 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13817 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13818 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13819 (UNSPEC_XORV): New unspecs.
13820 (optab): Add entries for them.
13821 (BITWISEV): New int iterator.
13822 (bit_reduc_op): New int attributes.
13824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13825 Alan Hayward <alan.hayward@arm.com>
13826 David Sherwood <david.sherwood@arm.com>
13828 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13829 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13830 * optabs.def (vec_shl_insert_optab): New optab.
13831 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13832 (duplicate_and_interleave): Likewise.
13833 * tree-vect-loop.c: Include internal-fn.h.
13834 (neutral_op_for_slp_reduction): New function, split out from
13835 get_initial_defs_for_reduction.
13836 (get_initial_def_for_reduction): Handle option 2 for variable-length
13837 vectors by loading the neutral value into a vector and then shifting
13838 the initial value into element 0.
13839 (get_initial_defs_for_reduction): Replace the code argument with
13840 the neutral value calculated by neutral_op_for_slp_reduction.
13841 Use gimple_build_vector for constant-length vectors.
13842 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13843 but the first group_size elements have a neutral value.
13844 Use duplicate_and_interleave otherwise.
13845 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13846 Update call to get_initial_defs_for_reduction. Handle SLP
13847 reductions for variable-length vectors by creating one vector
13848 result for each scalar result, with the elements associated
13849 with other scalar results stubbed out with the neutral value.
13850 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13851 Require IFN_VEC_SHL_INSERT for double reductions on
13852 variable-length vectors, or SLP reductions that have
13853 a neutral value. Require can_duplicate_and_interleave_p
13854 support for variable-length unchained SLP reductions if there
13855 is no neutral value, such as for MIN/MAX reductions. Also require
13856 the number of vector elements to be a multiple of the number of
13857 SLP statements when doing variable-length unchained SLP reductions.
13858 Update call to vect_create_epilog_for_reduction.
13859 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13860 and remove initial values.
13861 (duplicate_and_interleave): Make public.
13862 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13863 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13866 Alan Hayward <alan.hayward@arm.com>
13867 David Sherwood <david.sherwood@arm.com>
13869 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13870 (can_duplicate_and_interleave_p): New function.
13871 (vect_get_and_check_slp_defs): Take the vector of statements
13872 rather than just the current one. Remove excess parentheses.
13873 Restriction rejectinon of vect_constant_def and vect_external_def
13874 for variable-length vectors to boolean types, or types for which
13875 can_duplicate_and_interleave_p is false.
13876 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13877 (duplicate_and_interleave): New function.
13878 (vect_get_constant_vectors): Use gimple_build_vector for
13879 constant-length vectors and suitable variable-length constant
13880 vectors. Use duplicate_and_interleave for other variable-length
13881 vectors. Don't defer the update when inserting new statements.
13883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13884 Alan Hayward <alan.hayward@arm.com>
13885 David Sherwood <david.sherwood@arm.com>
13887 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13888 min_profitable_iters doesn't go negative.
13890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13895 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13896 * optabs.def (vec_mask_load_lanes_optab): New optab.
13897 (vec_mask_store_lanes_optab): Likewise.
13898 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13899 (MASK_STORE_LANES): Likewise.
13900 * internal-fn.c (mask_load_lanes_direct): New macro.
13901 (mask_store_lanes_direct): Likewise.
13902 (expand_mask_load_optab_fn): Handle masked operations.
13903 (expand_mask_load_lanes_optab_fn): New macro.
13904 (expand_mask_store_optab_fn): Handle masked operations.
13905 (expand_mask_store_lanes_optab_fn): New macro.
13906 (direct_mask_load_lanes_optab_supported_p): Likewise.
13907 (direct_mask_store_lanes_optab_supported_p): Likewise.
13908 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13910 (vect_load_lanes_supported): Likewise.
13911 * tree-vect-data-refs.c (strip_conversion): New function.
13912 (can_group_stmts_p): Likewise.
13913 (vect_analyze_data_ref_accesses): Use it instead of checking
13914 for a pair of assignments.
13915 (vect_store_lanes_supported): Take a masked_p parameter.
13916 (vect_load_lanes_supported): Likewise.
13917 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13918 vect_store_lanes_supported and vect_load_lanes_supported.
13919 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13920 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13921 parameter. Don't allow gaps for masked accesses.
13922 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13923 and vect_load_lanes_supported.
13924 (get_load_store_type): Take a masked_p parameter and update
13925 call to get_group_load_store_type.
13926 (vectorizable_store): Update call to get_load_store_type.
13927 Handle IFN_MASK_STORE_LANES.
13928 (vectorizable_load): Update call to get_load_store_type.
13929 Handle IFN_MASK_LOAD_LANES.
13931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13937 * config/aarch64/aarch64-protos.h
13938 (aarch64_sve_struct_memory_operand_p): Declare.
13939 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13940 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13941 (VPRED, vpred): Handle SVE structure modes.
13942 * config/aarch64/constraints.md (Utx): New constraint.
13943 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13944 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13945 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13946 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13947 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13948 structure modes. Split into pieces after RA.
13949 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13950 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13952 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13953 SVE structure modes.
13954 (aarch64_classify_address): Likewise.
13955 (sizetochar): Move earlier in file.
13956 (aarch64_print_operand): Handle SVE register lists.
13957 (aarch64_array_mode): New function.
13958 (aarch64_sve_struct_memory_operand_p): Likewise.
13959 (TARGET_ARRAY_MODE): Redefine.
13961 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13962 Alan Hayward <alan.hayward@arm.com>
13963 David Sherwood <david.sherwood@arm.com>
13965 * target.def (array_mode): New target hook.
13966 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13967 * doc/tm.texi: Regenerate.
13968 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13969 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13970 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13971 targetm.array_mode.
13972 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13976 Alan Hayward <alan.hayward@arm.com>
13977 David Sherwood <david.sherwood@arm.com>
13979 * fold-const.c (fold_binary_loc): Check the argument types
13980 rather than the result type when testing for a vector operation.
13982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13984 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13985 * doc/tm.texi: Regenerate.
13987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13988 Alan Hayward <alan.hayward@arm.com>
13989 David Sherwood <david.sherwood@arm.com>
13991 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13992 (sve): Document new AArch64 extension.
13993 * doc/md.texi (w): Extend the description of the AArch64
13994 constraint to include SVE vectors.
13995 (Upl, Upa): Document new AArch64 predicate constraints.
13996 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13998 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13999 (msve-vector-bits=): New option.
14000 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14001 SVE when these are disabled.
14002 (sve): New extension.
14003 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14004 modes. Adjust their number of units based on aarch64_sve_vg.
14005 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14006 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14007 aarch64_addr_query_type.
14008 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14009 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14010 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14011 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14012 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14013 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14014 (aarch64_simd_imm_zero_p): Delete.
14015 (aarch64_check_zero_based_sve_index_immediate): Declare.
14016 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14017 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14018 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14019 (aarch64_sve_float_mul_immediate_p): Likewise.
14020 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14021 rather than an rtx.
14022 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14023 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14024 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14025 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14026 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14027 (aarch64_regmode_natural_size): Likewise.
14028 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14029 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14031 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14032 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14033 for VG and the SVE predicate registers.
14034 (V_ALIASES): Add a "z"-prefixed alias.
14035 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14036 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14037 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14038 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14039 (REG_CLASS_NAMES): Add entries for them.
14040 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14041 and the predicate registers.
14042 (aarch64_sve_vg): Declare.
14043 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14044 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14045 (REGMODE_NATURAL_SIZE): Define.
14046 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14048 * config/aarch64/aarch64.c: Include cfgrtl.h.
14049 (simd_immediate_info): Add a constructor for series vectors,
14050 and an associated step field.
14051 (aarch64_sve_vg): New variable.
14052 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14053 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14054 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14055 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14056 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14057 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14058 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14059 (aarch64_get_mask_mode): New functions.
14060 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14061 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14062 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14063 predicate modes and predicate registers. Explicitly restrict
14064 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14065 to store a vector mode if it is recognized by
14066 aarch64_classify_vector_mode.
14067 (aarch64_regmode_natural_size): New function.
14068 (aarch64_hard_regno_caller_save_mode): Return the original mode
14070 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14071 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14072 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14073 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14075 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14076 does not overlap dest if the function is frame-related. Handle
14078 (aarch64_split_add_offset): New function.
14079 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14080 them aarch64_add_offset.
14081 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14082 and update call to aarch64_sub_sp.
14083 (aarch64_add_cfa_expression): New function.
14084 (aarch64_expand_prologue): Pass extra temporary registers to the
14085 functions above. Handle the case in which we need to emit new
14086 DW_CFA_expressions for registers that were originally saved
14087 relative to the stack pointer, but now have to be expressed
14088 relative to the frame pointer.
14089 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14091 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14092 IP0 and IP1 values for SVE frames.
14093 (aarch64_expand_vec_series): New function.
14094 (aarch64_expand_sve_widened_duplicate): Likewise.
14095 (aarch64_expand_sve_const_vector): Likewise.
14096 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14097 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14098 into the register, rather than emitting a SET directly.
14099 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14100 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14101 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14102 (offset_9bit_signed_scaled_p): New functions.
14103 (aarch64_replicate_bitmask_imm): New function.
14104 (aarch64_bitmask_imm): Use it.
14105 (aarch64_cannot_force_const_mem): Reject expressions involving
14106 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14107 (aarch64_classify_index): Handle SVE indices, by requiring
14108 a plain register index with a scale that matches the element size.
14109 (aarch64_classify_address): Handle SVE addresses. Assert that
14110 the mode of the address is VOIDmode or an integer mode.
14111 Update call to aarch64_classify_symbol.
14112 (aarch64_classify_symbolic_expression): Update call to
14113 aarch64_classify_symbol.
14114 (aarch64_const_vec_all_in_range_p): New function.
14115 (aarch64_print_vector_float_operand): Likewise.
14116 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14117 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14118 and the FP immediates 1.0 and 0.5.
14119 (aarch64_print_address_internal): Handle SVE addresses.
14120 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14121 (aarch64_regno_regclass): Handle predicate registers.
14122 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14124 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14125 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14126 (aarch64_convert_sve_vector_bits): New function.
14127 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14128 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14129 rather than an rtx.
14130 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14131 Handle SVE vector and predicate modes. Accept VL-based constants
14132 that need only one temporary register, and VL offsets that require
14133 no temporary registers.
14134 (aarch64_conditional_register_usage): Mark the predicate registers
14135 as fixed if SVE isn't available.
14136 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14137 Return true for SVE vector and predicate modes.
14138 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14139 rather than an unsigned int. Handle SVE modes.
14140 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14142 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14144 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14145 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14146 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14147 (aarch64_sve_float_mul_immediate_p): New functions.
14148 (aarch64_sve_valid_immediate): New function.
14149 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14150 Explicitly reject structure modes. Check for INDEX constants.
14151 Handle PTRUE and PFALSE constants.
14152 (aarch64_check_zero_based_sve_index_immediate): New function.
14153 (aarch64_simd_imm_zero_p): Delete.
14154 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14155 vector modes. Accept constants in the range of CNT[BHWD].
14156 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14157 ask for an Advanced SIMD mode.
14158 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14159 (aarch64_simd_vector_alignment): Handle SVE predicates.
14160 (aarch64_vectorize_preferred_vector_alignment): New function.
14161 (aarch64_simd_vector_alignment_reachable): Use it instead of
14163 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14164 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14166 (MAX_VECT_LEN): Delete.
14167 (expand_vec_perm_d): Add a vec_flags field.
14168 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14169 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14170 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14172 (aarch64_evpc_rev): Rename to...
14173 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14174 (aarch64_evpc_rev_global): New function.
14175 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14176 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14178 (aarch64_evpc_sve_tbl): New function.
14179 (aarch64_expand_vec_perm_const_1): Update after rename of
14180 aarch64_evpc_rev. Handle SVE permutes too, trying
14181 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14182 than aarch64_evpc_tbl.
14183 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14184 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14185 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14186 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14187 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14188 (aarch64_expand_sve_vcond): New functions.
14189 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14190 of aarch64_vector_mode_p.
14191 (aarch64_dwarf_poly_indeterminate_value): New function.
14192 (aarch64_compute_pressure_classes): Likewise.
14193 (aarch64_can_change_mode_class): Likewise.
14194 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14195 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14196 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14197 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14198 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14199 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14200 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14201 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14203 (Dn, Dl, Dr): Accept const as well as const_vector.
14204 (Dz): Likewise. Compare against CONST0_RTX.
14205 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14206 of "vector" where appropriate.
14207 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14208 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14209 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14210 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14211 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14212 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14213 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14214 (v_int_equiv): Extend to SVE modes.
14215 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14217 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14218 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14219 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14220 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14221 (SVE_COND_FP_CMP): New int iterators.
14222 (perm_hilo): Handle the new unpack unspecs.
14223 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14225 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14226 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14227 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14228 (aarch64_equality_operator, aarch64_constant_vector_operand)
14229 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14230 (aarch64_sve_nonimmediate_operand): Likewise.
14231 (aarch64_sve_general_operand): Likewise.
14232 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14233 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14234 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14235 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14236 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14237 (aarch64_sve_float_arith_immediate): Likewise.
14238 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14239 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14240 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14241 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14242 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14243 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14244 (aarch64_sve_float_arith_operand): Likewise.
14245 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14246 (aarch64_sve_float_mul_operand): Likewise.
14247 (aarch64_sve_vec_perm_operand): Likewise.
14248 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14249 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14250 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14251 as well as const_vector.
14252 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14253 in file. Use CONST0_RTX and CONSTM1_RTX.
14254 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14255 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14256 Use aarch64_simd_imm_zero.
14257 * config/aarch64/aarch64-sve.md: New file.
14258 * config/aarch64/aarch64.md: Include it.
14259 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14260 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14261 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14262 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14263 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14264 (sve): New attribute.
14265 (enabled): Disable instructions with the sve attribute unless
14267 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14268 aarch64_expand_mov_immediate.
14269 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14270 CNT[BHSD] immediates.
14271 (movti): Split CONST_POLY_INT moves into two halves.
14272 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14273 Split additions that need a temporary here if the destination
14274 is the stack pointer.
14275 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14276 (*add<mode>3_poly_1): New instruction.
14277 (set_clobber_cc): New expander.
14279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14281 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14282 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14283 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14284 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14285 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14286 Change innermode from fixed_mode_size to machine_mode.
14287 (simplify_subreg): Update call accordingly. Handle a constant-sized
14288 subreg of a variable-length CONST_VECTOR.
14290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14291 Alan Hayward <alan.hayward@arm.com>
14292 David Sherwood <david.sherwood@arm.com>
14294 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14295 (add_offset_to_base): New function, split out from...
14296 (create_mem_ref): ...here. When handling a scale other than 1,
14297 check first whether the address is valid without the offset.
14298 Add it into the base if so, leaving the index and scale as-is.
14300 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14303 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14304 fold_for_warn before checking if arg2 is INTEGER_CST.
14306 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14308 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14309 (store_multiple_operation): Delete.
14310 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14311 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14312 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14313 guarded by TARGET_STRING.
14314 (rs6000_output_load_multiple): Delete.
14315 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14316 OPTION_MASK_STRING / TARGET_STRING handling.
14317 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14318 (const rs6000_opt_masks) <"string">: Change mask to 0.
14319 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14320 (MASK_STRING): Delete.
14321 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14323 (load_multiple): Delete.
14330 (store_multiple): Delete.
14337 (movmemsi_8reg): Delete.
14338 (corresponding unnamed define_insn): Delete.
14339 (movmemsi_6reg): Delete.
14340 (corresponding unnamed define_insn): Delete.
14341 (movmemsi_4reg): Delete.
14342 (corresponding unnamed define_insn): Delete.
14343 (movmemsi_2reg): Delete.
14344 (corresponding unnamed define_insn): Delete.
14345 (movmemsi_1reg): Delete.
14346 (corresponding unnamed define_insn): Delete.
14347 * config/rs6000/rs6000.opt (mno-string): New.
14348 (mstring): Replace by deprecation warning stub.
14349 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14351 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14353 * regrename.c (regrename_do_replace): If replacing the same
14354 reg multiple times, try to reuse last created gen_raw_REG.
14357 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14358 main to workaround a bug in GDB.
14360 2018-01-12 Tom de Vries <tom@codesourcery.com>
14363 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14365 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14367 PR rtl-optimization/80481
14368 * ira-color.c (get_cap_member): New function.
14369 (allocnos_conflict_by_live_ranges_p): Use it.
14370 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14371 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14373 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14376 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14377 (*saddl_se_1): Ditto.
14378 (*ssubsi_1): Ditto.
14379 (*ssubl_se_1): Ditto.
14381 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14383 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14384 rather than wi::to_widest for DR_INITs.
14385 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14386 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14387 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14389 (vect_analyze_group_access_1): Note that here.
14391 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14393 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14394 polynomial type sizes.
14396 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14398 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14399 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14400 (gimple_add_tmp_var): Likewise.
14402 2018-01-12 Martin Liska <mliska@suse.cz>
14404 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14405 (gimple_alloc_sizes): Likewise.
14406 (dump_gimple_statistics): Use PRIu64 in printf format.
14407 * gimple.h: Change uint64_t to int.
14409 2018-01-12 Martin Liska <mliska@suse.cz>
14411 * tree-core.h: Use uint64_t instead of int.
14412 * tree.c (tree_node_counts): Likewise.
14413 (tree_node_sizes): Likewise.
14414 (dump_tree_statistics): Use PRIu64 in printf format.
14416 2018-01-12 Martin Liska <mliska@suse.cz>
14418 * Makefile.in: As qsort_chk is implemented in vec.c, add
14419 vec.o to linkage of gencfn-macros.
14420 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14421 passing the info to record_node_allocation_statistics.
14422 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14424 * ggc-common.c (struct ggc_usage): Add operator== and use
14425 it in operator< and compare function.
14426 * mem-stats.h (struct mem_usage): Likewise.
14427 * vec.c (struct vec_usage): Remove operator< and compare
14428 function. Can be simply inherited.
14430 2018-01-12 Martin Jambor <mjambor@suse.cz>
14433 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14434 * tree-ssa-math-opts.c: Include domwalk.h.
14435 (convert_mult_to_fma_1): New function.
14436 (fma_transformation_info): New type.
14437 (fma_deferring_state): Likewise.
14438 (cancel_fma_deferring): New function.
14439 (result_of_phi): Likewise.
14440 (last_fma_candidate_feeds_initial_phi): Likewise.
14441 (convert_mult_to_fma): Added deferring logic, split actual
14442 transformation to convert_mult_to_fma_1.
14443 (math_opts_dom_walker): New type.
14444 (math_opts_dom_walker::after_dom_children): New method, body moved
14445 here from pass_optimize_widening_mul::execute, added deferring logic
14447 (pass_optimize_widening_mul::execute): Moved most of code to
14448 math_opts_dom_walker::after_dom_children.
14449 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14450 * config/i386/i386.c (ix86_option_override_internal): Added
14451 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14453 2018-01-12 Richard Biener <rguenther@suse.de>
14456 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14457 inline instance vars.
14459 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14462 * config/rx/rx.c (rx_is_restricted_memory_address):
14463 Handle SUBREG case.
14465 2018-01-12 Richard Biener <rguenther@suse.de>
14467 PR tree-optimization/80846
14468 * target.def (split_reduction): New target hook.
14469 * targhooks.c (default_split_reduction): New function.
14470 * targhooks.h (default_split_reduction): Declare.
14471 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14472 target requests first reduce vectors by combining low and high
14474 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14475 (get_vectype_for_scalar_type_and_size): Export.
14476 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14477 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14478 * doc/tm.texi: Regenerate.
14479 * config/i386/i386.c (ix86_split_reduction): Implement
14480 TARGET_VECTORIZE_SPLIT_REDUCTION.
14482 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14485 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14486 in PIC mode except for TARGET_VXWORKS_RTP.
14487 * config/sparc/sparc.c: Include cfgrtl.h.
14488 (TARGET_INIT_PIC_REG): Define.
14489 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14490 (sparc_pic_register_p): New predicate.
14491 (sparc_legitimate_address_p): Use it.
14492 (sparc_legitimize_pic_address): Likewise.
14493 (sparc_delegitimize_address): Likewise.
14494 (sparc_mode_dependent_address_p): Likewise.
14495 (gen_load_pcrel_sym): Remove 4th parameter.
14496 (load_got_register): Adjust call to above. Remove obsolete stuff.
14497 (sparc_expand_prologue): Do not call load_got_register here.
14498 (sparc_flat_expand_prologue): Likewise.
14499 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14500 (sparc_use_pseudo_pic_reg): New function.
14501 (sparc_init_pic_reg): Likewise.
14502 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14503 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14505 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14507 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14508 Add item for branch_cost.
14510 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14512 PR rtl-optimization/83565
14513 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14514 not extend the result to a larger mode for rotate operations.
14515 (num_sign_bit_copies1): Likewise.
14517 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14520 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14522 Use values-Xc.o for -pedantic.
14523 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14525 2018-01-12 Martin Liska <mliska@suse.cz>
14528 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14530 (possible_polymorphic_call_targets): Use it.
14531 (ipa_devirt): Likewise.
14533 2018-01-12 Martin Liska <mliska@suse.cz>
14535 * profile-count.h (enum profile_quality): Use 0 as invalid
14536 enum value of profile_quality.
14538 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14540 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14541 -mext-string options.
14543 2018-01-12 Richard Biener <rguenther@suse.de>
14545 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14546 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14547 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14549 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14551 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14553 * configure.ac (--with-long-double-format): Add support for the
14554 configuration option to change the default long double format on
14556 * config.gcc (powerpc*-linux*-*): Likewise.
14557 * configure: Regenerate.
14558 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14559 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14560 used without modification.
14562 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14564 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14565 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14566 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14567 MISC_BUILTIN_SPEC_BARRIER.
14568 (rs6000_init_builtins): Likewise.
14569 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14571 (speculation_barrier): New define_insn.
14572 * doc/extend.texi: Document __builtin_speculation_barrier.
14574 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14577 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14578 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14579 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14581 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14582 integral modes instead of "ss" and "sd".
14583 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14584 vectors with 32-bit and 64-bit elements.
14585 (vecdupssescalarmodesuffix): New mode attribute.
14586 (vec_dup<mode>): Use it.
14588 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14591 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14592 frame if argument is passed on stack.
14594 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14597 * ree.c (combine_reaching_defs): Optimize also
14598 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14599 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14601 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14603 PR middle-end/83189
14604 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14606 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14608 PR middle-end/83718
14609 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14610 after they are computed.
14612 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14614 PR tree-optimization/83695
14615 * gimple-loop-linterchange.cc
14616 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14617 reset cached scev information after interchange.
14618 (pass_linterchange::execute): Remove call to scev_reset_htab.
14620 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14622 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14623 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14624 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14625 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14626 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14627 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14628 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14629 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14630 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14631 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14632 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14633 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14634 (V_lane_reg): Likewise.
14635 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14637 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14638 (vfmal_lane_low<mode>_intrinsic,
14639 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14640 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14641 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14642 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14643 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14644 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14646 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14648 * config/arm/arm-cpus.in (fp16fml): New feature.
14649 (ALL_SIMD): Add fp16fml.
14650 (armv8.2-a): Add fp16fml as an option.
14651 (armv8.3-a): Likewise.
14652 (armv8.4-a): Add fp16fml as part of fp16.
14653 * config/arm/arm.h (TARGET_FP16FML): Define.
14654 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14656 * config/arm/arm-modes.def (V2HF): Define.
14657 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14658 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14659 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14660 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14661 vfmsl_low, vfmsl_high): New set of builtins.
14662 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14663 (vfml_op): New code attribute.
14664 (VFMLHALVES): New int iterator.
14665 (VFML, VFMLSEL): New mode attributes.
14666 (V_reg): Define mapping for V2HF.
14667 (V_hi, V_lo): New mode attributes.
14668 (VF_constraint): Likewise.
14669 (vfml_half, vfml_half_selector): New int attributes.
14670 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14672 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14673 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14675 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14676 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14677 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14678 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14680 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14681 Document new effective target and option set.
14683 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14685 * config/arm/arm-cpus.in (armv8_4): New feature.
14686 (ARMv8_4a): New fgroup.
14687 (armv8.4-a): New arch.
14688 * config/arm/arm-tables.opt: Regenerate.
14689 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14690 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14691 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14692 Add matching rules for -march=armv8.4-a and extensions.
14693 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14695 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14698 * config/rx/rx.md (BW): New mode attribute.
14699 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14701 2018-01-11 Richard Biener <rguenther@suse.de>
14703 PR tree-optimization/83435
14704 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14705 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14706 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14708 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14709 Alan Hayward <alan.hayward@arm.com>
14710 David Sherwood <david.sherwood@arm.com>
14712 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14714 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14715 (aarch64_print_address_internal): Use it to check for a zero offset.
14717 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14718 Alan Hayward <alan.hayward@arm.com>
14719 David Sherwood <david.sherwood@arm.com>
14721 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14722 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14723 Return a poly_int64 rather than a HOST_WIDE_INT.
14724 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14725 rather than a HOST_WIDE_INT.
14726 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14727 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14728 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14729 final_offset from HOST_WIDE_INT to poly_int64.
14730 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14731 to_constant when getting the number of units in an Advanced SIMD
14733 (aarch64_builtin_vectorized_function): Check for a constant number
14735 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14737 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14738 attribute instead of GET_MODE_NUNITS.
14739 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14740 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14741 GET_MODE_SIZE for fixed-size registers.
14742 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14743 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14744 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14745 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14746 (aarch64_print_operand, aarch64_print_address_internal)
14747 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14748 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14749 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14750 Handle polynomial GET_MODE_SIZE.
14751 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14752 wider than SImode without modification.
14753 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14754 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14755 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14756 passing and returning SVE modes.
14757 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14758 rather than GEN_INT.
14759 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14760 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14761 (aarch64_allocate_and_probe_stack_space): Likewise.
14762 (aarch64_layout_frame): Cope with polynomial offsets.
14763 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14764 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14765 polynomial offsets.
14766 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14767 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14768 poly_int64 rather than a HOST_WIDE_INT.
14769 (aarch64_get_separate_components, aarch64_process_components)
14770 (aarch64_expand_prologue, aarch64_expand_epilogue)
14771 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14772 (aarch64_anchor_offset): New function, split out from...
14773 (aarch64_legitimize_address): ...here.
14774 (aarch64_builtin_vectorization_cost): Handle polynomial
14775 TYPE_VECTOR_SUBPARTS.
14776 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14778 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14779 number of elements from the PARALLEL rather than the mode.
14780 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14781 rather than GET_MODE_BITSIZE.
14782 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14783 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14784 (aarch64_expand_vec_perm_const_1): Handle polynomial
14785 d->perm.length () and d->perm elements.
14786 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14787 Apply to_constant to d->perm elements.
14788 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14789 polynomial CONST_VECTOR_NUNITS.
14790 (aarch64_move_pointer): Take amount as a poly_int64 rather
14792 (aarch64_progress_pointer): Avoid temporary variable.
14793 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14794 the mode attribute instead of GET_MODE.
14796 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14797 Alan Hayward <alan.hayward@arm.com>
14798 David Sherwood <david.sherwood@arm.com>
14800 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14801 x exists before using it.
14802 (aarch64_add_constant_internal): Rename to...
14803 (aarch64_add_offset_1): ...this. Replace regnum with separate
14804 src and dest rtxes. Handle the case in which they're different,
14805 including when the offset is zero. Replace scratchreg with an rtx.
14806 Use 2 additions if there is no spare register into which we can
14807 move a 16-bit constant.
14808 (aarch64_add_constant): Delete.
14809 (aarch64_add_offset): Replace reg with separate src and dest
14810 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14811 Use aarch64_add_offset_1.
14812 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14813 an rtx rather than an int. Take the delta as a poly_int64
14814 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14815 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14816 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14817 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14818 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14819 and aarch64_add_sp.
14820 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14821 aarch64_add_constant.
14823 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14825 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14826 Use scalar_float_mode.
14828 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14830 * config/aarch64/aarch64-simd.md
14831 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14832 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14833 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14834 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14835 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14836 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14837 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14838 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14839 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14840 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14842 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14845 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14846 targ_options->x_arm_arch_string is non NULL.
14848 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14850 * config/aarch64/aarch64.h
14851 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14853 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14856 * expmed.c (emit_store_flag_force): Swap if const op0
14857 and change VOIDmode to mode of op0.
14859 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14861 PR rtl-optimization/83761
14862 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14863 than bytes to mode_for_size.
14865 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14867 PR middle-end/83189
14868 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14869 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14872 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14874 PR middle-end/83575
14875 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14876 when in layout mode.
14877 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14878 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14881 2018-01-10 Michael Collison <michael.collison@arm.com>
14883 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14884 * config/aarch64/aarch64-option-extension.def: Add
14885 AARCH64_OPT_EXTENSION of 'fp16fml'.
14886 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14887 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14888 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14889 * config/aarch64/constraints.md (Ui7): New constraint.
14890 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14891 (VFMLA_SEL_W): Ditto.
14894 (VFMLA16_LOW): New int iterator.
14895 (VFMLA16_HIGH): Ditto.
14896 (UNSPEC_FMLAL): New unspec.
14897 (UNSPEC_FMLSL): Ditto.
14898 (UNSPEC_FMLAL2): Ditto.
14899 (UNSPEC_FMLSL2): Ditto.
14900 (f16mac): New code attribute.
14901 * config/aarch64/aarch64-simd-builtins.def
14902 (aarch64_fmlal_lowv2sf): Ditto.
14903 (aarch64_fmlsl_lowv2sf): Ditto.
14904 (aarch64_fmlalq_lowv4sf): Ditto.
14905 (aarch64_fmlslq_lowv4sf): Ditto.
14906 (aarch64_fmlal_highv2sf): Ditto.
14907 (aarch64_fmlsl_highv2sf): Ditto.
14908 (aarch64_fmlalq_highv4sf): Ditto.
14909 (aarch64_fmlslq_highv4sf): Ditto.
14910 (aarch64_fmlal_lane_lowv2sf): Ditto.
14911 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14912 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14913 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14914 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14915 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14916 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14917 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14918 (aarch64_fmlal_lane_highv2sf): Ditto.
14919 (aarch64_fmlsl_lane_highv2sf): Ditto.
14920 (aarch64_fmlal_laneq_highv2sf): Ditto.
14921 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14922 (aarch64_fmlalq_lane_highv4sf): Ditto.
14923 (aarch64_fmlsl_lane_highv4sf): Ditto.
14924 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14925 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14926 * config/aarch64/aarch64-simd.md:
14927 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14928 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14929 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14930 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14931 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14932 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14933 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14934 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14935 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14936 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14937 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14938 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14939 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14940 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14941 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14942 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14943 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14944 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14945 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14946 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14947 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14948 (vfmlsl_low_u32): Ditto.
14949 (vfmlalq_low_u32): Ditto.
14950 (vfmlslq_low_u32): Ditto.
14951 (vfmlal_high_u32): Ditto.
14952 (vfmlsl_high_u32): Ditto.
14953 (vfmlalq_high_u32): Ditto.
14954 (vfmlslq_high_u32): Ditto.
14955 (vfmlal_lane_low_u32): Ditto.
14956 (vfmlsl_lane_low_u32): Ditto.
14957 (vfmlal_laneq_low_u32): Ditto.
14958 (vfmlsl_laneq_low_u32): Ditto.
14959 (vfmlalq_lane_low_u32): Ditto.
14960 (vfmlslq_lane_low_u32): Ditto.
14961 (vfmlalq_laneq_low_u32): Ditto.
14962 (vfmlslq_laneq_low_u32): Ditto.
14963 (vfmlal_lane_high_u32): Ditto.
14964 (vfmlsl_lane_high_u32): Ditto.
14965 (vfmlal_laneq_high_u32): Ditto.
14966 (vfmlsl_laneq_high_u32): Ditto.
14967 (vfmlalq_lane_high_u32): Ditto.
14968 (vfmlslq_lane_high_u32): Ditto.
14969 (vfmlalq_laneq_high_u32): Ditto.
14970 (vfmlslq_laneq_high_u32): Ditto.
14971 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14972 (AARCH64_FL_FOR_ARCH8_4): New.
14973 (AARCH64_ISA_F16FML): New ISA flag.
14974 (TARGET_F16FML): New feature flag for fp16fml.
14975 (doc/invoke.texi): Document new fp16fml option.
14977 2018-01-10 Michael Collison <michael.collison@arm.com>
14979 * config/aarch64/aarch64-builtins.c:
14980 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14981 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14982 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14983 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14984 (AARCH64_ISA_SHA3): New ISA flag.
14985 (TARGET_SHA3): New feature flag for sha3.
14986 * config/aarch64/iterators.md (sha512_op): New int attribute.
14987 (CRYPTO_SHA512): New int iterator.
14988 (UNSPEC_SHA512H): New unspec.
14989 (UNSPEC_SHA512H2): Ditto.
14990 (UNSPEC_SHA512SU0): Ditto.
14991 (UNSPEC_SHA512SU1): Ditto.
14992 * config/aarch64/aarch64-simd-builtins.def
14993 (aarch64_crypto_sha512hqv2di): New builtin.
14994 (aarch64_crypto_sha512h2qv2di): Ditto.
14995 (aarch64_crypto_sha512su0qv2di): Ditto.
14996 (aarch64_crypto_sha512su1qv2di): Ditto.
14997 (aarch64_eor3qv8hi): Ditto.
14998 (aarch64_rax1qv2di): Ditto.
14999 (aarch64_xarqv2di): Ditto.
15000 (aarch64_bcaxqv8hi): Ditto.
15001 * config/aarch64/aarch64-simd.md:
15002 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15003 (aarch64_crypto_sha512su0qv2di): Ditto.
15004 (aarch64_crypto_sha512su1qv2di): Ditto.
15005 (aarch64_eor3qv8hi): Ditto.
15006 (aarch64_rax1qv2di): Ditto.
15007 (aarch64_xarqv2di): Ditto.
15008 (aarch64_bcaxqv8hi): Ditto.
15009 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15010 (vsha512h2q_u64): Ditto.
15011 (vsha512su0q_u64): Ditto.
15012 (vsha512su1q_u64): Ditto.
15013 (veor3q_u16): Ditto.
15014 (vrax1q_u64): Ditto.
15015 (vxarq_u64): Ditto.
15016 (vbcaxq_u16): Ditto.
15017 * config/arm/types.md (crypto_sha512): New type attribute.
15018 (crypto_sha3): Ditto.
15019 (doc/invoke.texi): Document new sha3 option.
15021 2018-01-10 Michael Collison <michael.collison@arm.com>
15023 * config/aarch64/aarch64-builtins.c:
15024 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15025 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15026 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15027 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15028 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15029 (AARCH64_ISA_SM4): New ISA flag.
15030 (TARGET_SM4): New feature flag for sm4.
15031 * config/aarch64/aarch64-simd-builtins.def
15032 (aarch64_sm3ss1qv4si): Ditto.
15033 (aarch64_sm3tt1aq4si): Ditto.
15034 (aarch64_sm3tt1bq4si): Ditto.
15035 (aarch64_sm3tt2aq4si): Ditto.
15036 (aarch64_sm3tt2bq4si): Ditto.
15037 (aarch64_sm3partw1qv4si): Ditto.
15038 (aarch64_sm3partw2qv4si): Ditto.
15039 (aarch64_sm4eqv4si): Ditto.
15040 (aarch64_sm4ekeyqv4si): Ditto.
15041 * config/aarch64/aarch64-simd.md:
15042 (aarch64_sm3ss1qv4si): Ditto.
15043 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15044 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15045 (aarch64_sm4eqv4si): Ditto.
15046 (aarch64_sm4ekeyqv4si): Ditto.
15047 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15048 (sm3part_op): Ditto.
15049 (CRYPTO_SM3TT): Ditto.
15050 (CRYPTO_SM3PART): Ditto.
15051 (UNSPEC_SM3SS1): New unspec.
15052 (UNSPEC_SM3TT1A): Ditto.
15053 (UNSPEC_SM3TT1B): Ditto.
15054 (UNSPEC_SM3TT2A): Ditto.
15055 (UNSPEC_SM3TT2B): Ditto.
15056 (UNSPEC_SM3PARTW1): Ditto.
15057 (UNSPEC_SM3PARTW2): Ditto.
15058 (UNSPEC_SM4E): Ditto.
15059 (UNSPEC_SM4EKEY): Ditto.
15060 * config/aarch64/constraints.md (Ui2): New constraint.
15061 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15062 * config/arm/types.md (crypto_sm3): New type attribute.
15063 (crypto_sm4): Ditto.
15064 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15065 (vsm3tt1aq_u32): Ditto.
15066 (vsm3tt1bq_u32): Ditto.
15067 (vsm3tt2aq_u32): Ditto.
15068 (vsm3tt2bq_u32): Ditto.
15069 (vsm3partw1q_u32): Ditto.
15070 (vsm3partw2q_u32): Ditto.
15071 (vsm4eq_u32): Ditto.
15072 (vsm4ekeyq_u32): Ditto.
15073 (doc/invoke.texi): Document new sm4 option.
15075 2018-01-10 Michael Collison <michael.collison@arm.com>
15077 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15078 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15079 (AARCH64_FL_FOR_ARCH8_4): New.
15080 (AARCH64_FL_V8_4): New flag.
15081 (doc/invoke.texi): Document new armv8.4-a option.
15083 2018-01-10 Michael Collison <michael.collison@arm.com>
15085 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15086 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15087 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15088 * config/aarch64/aarch64-option-extension.def: Add
15089 AARCH64_OPT_EXTENSION of 'sha2'.
15090 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15091 (crypto): Disable sha2 and aes if crypto disabled.
15092 (crypto): Enable aes and sha2 if enabled.
15093 (simd): Disable sha2 and aes if simd disabled.
15094 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15096 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15097 (TARGET_SHA2): New feature flag for sha2.
15098 (TARGET_AES): New feature flag for aes.
15099 * config/aarch64/aarch64-simd.md:
15100 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15101 conditional on TARGET_AES.
15102 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15103 (aarch64_crypto_sha1hsi): Make pattern conditional
15105 (aarch64_crypto_sha1hv4si): Ditto.
15106 (aarch64_be_crypto_sha1hv4si): Ditto.
15107 (aarch64_crypto_sha1su1v4si): Ditto.
15108 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15109 (aarch64_crypto_sha1su0v4si): Ditto.
15110 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15111 (aarch64_crypto_sha256su0v4si): Ditto.
15112 (aarch64_crypto_sha256su1v4si): Ditto.
15113 (doc/invoke.texi): Document new aes and sha2 options.
15115 2018-01-10 Martin Sebor <msebor@redhat.com>
15117 PR tree-optimization/83781
15118 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15121 2018-01-11 Martin Sebor <msebor@gmail.com>
15122 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15124 PR tree-optimization/83501
15125 PR tree-optimization/81703
15127 * tree-ssa-strlen.c (get_string_cst): Rename...
15128 (get_string_len): ...to this. Handle global constants.
15129 (handle_char_store): Adjust.
15131 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15132 Jim Wilson <jimw@sifive.com>
15134 * config/riscv/riscv-protos.h (riscv_output_return): New.
15135 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15136 (riscv_attribute_table, riscv_output_return),
15137 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15138 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15139 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15141 (riscv_expand_prologue): Add early return for naked function.
15142 (riscv_expand_epilogue): Likewise.
15143 (riscv_function_ok_for_sibcall): Return false for naked function.
15144 (riscv_set_current_function): New.
15145 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15146 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15147 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15148 * doc/extend.texi (RISC-V Function Attributes): New.
15150 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15152 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15153 check for 128-bit long double before checking TCmode.
15154 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15155 128-bit long doubles before checking TFmode or TCmode.
15156 (FLOAT128_IBM_P): Likewise.
15158 2018-01-10 Martin Sebor <msebor@redhat.com>
15160 PR tree-optimization/83671
15161 * builtins.c (c_strlen): Unconditionally return zero for the empty
15163 Use -Warray-bounds for warnings.
15164 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15165 for non-constant array indices with COMPONENT_REF, arrays of
15166 arrays, and pointers to arrays.
15167 (gimple_fold_builtin_strlen): Determine and set length range for
15168 non-constant character arrays.
15170 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15172 PR middle-end/81897
15173 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15176 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15178 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15180 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15183 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15184 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15185 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15186 indexed_or_indirect_operand predicate.
15187 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15188 (*vsx_le_perm_load_v8hi): Likewise.
15189 (*vsx_le_perm_load_v16qi): Likewise.
15190 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15191 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15192 (*vsx_le_perm_store_v8hi): Likewise.
15193 (*vsx_le_perm_store_v16qi): Likewise.
15194 (eight unnamed splitters): Likewise.
15196 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15198 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15199 * config/rs6000/emmintrin.h: Likewise.
15200 * config/rs6000/mmintrin.h: Likewise.
15201 * config/rs6000/xmmintrin.h: Likewise.
15203 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15206 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15208 * tree.c (tree_nop_conversion): Return true for location wrapper
15210 (maybe_wrap_with_location): New function.
15211 (selftest::check_strip_nops): New function.
15212 (selftest::test_location_wrappers): New function.
15213 (selftest::tree_c_tests): Call it.
15214 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15215 (maybe_wrap_with_location): New decl.
15216 (EXPR_LOCATION_WRAPPER_P): New macro.
15217 (location_wrapper_p): New inline function.
15218 (tree_strip_any_location_wrapper): New inline function.
15220 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15223 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15224 stack_realign_offset for the largest alignment of stack slot
15226 (ix86_find_max_used_stack_alignment): New function.
15227 (ix86_finalize_stack_frame_flags): Use it. Set
15228 max_used_stack_alignment if we don't realign stack.
15229 * config/i386/i386.h (machine_function): Add
15230 max_used_stack_alignment.
15232 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15234 * config/arm/arm.opt (-mbranch-cost): New option.
15235 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15238 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15241 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15242 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15244 2018-01-10 Richard Biener <rguenther@suse.de>
15247 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15248 early out so it also covers the case where we have a non-NULL
15251 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15253 PR tree-optimization/83753
15254 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15255 for non-strided grouped accesses if the number of elements is 1.
15257 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15260 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15261 * i386.h (TARGET_USE_GATHER): Define.
15262 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15264 2018-01-10 Martin Liska <mliska@suse.cz>
15267 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15268 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15270 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15271 CLEANUP_NO_PARTITIONING is not set.
15273 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15275 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15276 for vectors, as a partial revert of r254296.
15277 * rtl.h (const_vec_p): Delete.
15278 (const_vec_duplicate_p): Don't test for vector CONSTs.
15279 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15280 * expmed.c (make_tree): Likewise.
15283 * common.md (E, F): Use CONSTANT_P instead of checking for
15285 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15286 checking for CONST_VECTOR.
15288 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15290 PR middle-end/83575
15291 * predict.c (force_edge_cold): Handle in more sane way edges
15292 with no prediction.
15294 2018-01-09 Carl Love <cel@us.ibm.com>
15296 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15298 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15299 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15300 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15301 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15302 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15303 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15304 * config/rs6000/rs6000-protos.h: Add extern defition for
15305 rs6000_generate_float2_double_code.
15306 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15308 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15309 (float2_v2df): Add define_expand.
15311 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15314 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15315 op_mode in the force_to_mode call.
15317 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15319 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15320 instead of checking each element individually.
15321 (aarch64_evpc_uzp): Likewise.
15322 (aarch64_evpc_zip): Likewise.
15323 (aarch64_evpc_ext): Likewise.
15324 (aarch64_evpc_rev): Likewise.
15325 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15326 instead of checking each element individually. Return true without
15328 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15329 whether all selected elements come from the same input, instead of
15330 checking each element individually. Remove calls to gen_rtx_REG,
15331 start_sequence and end_sequence and instead assert that no rtl is
15334 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15336 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15337 order of HIGH and CONST checks.
15339 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15341 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15342 if the destination isn't an SSA_NAME.
15344 2018-01-09 Richard Biener <rguenther@suse.de>
15346 PR tree-optimization/83668
15347 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15349 (canonicalize_loop_form): ... here, renamed from ...
15350 (canonicalize_loop_closed_ssa_form): ... this and amended to
15351 swap successor edges for loop exit blocks to make us use
15352 the RPO order we need for initial schedule generation.
15354 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15356 PR tree-optimization/64811
15357 * match.pd: When optimizing comparisons with Inf, avoid
15358 introducing or losing exceptions from comparisons with NaN.
15360 2018-01-09 Martin Liska <mliska@suse.cz>
15363 * asan.c (shadow_mem_size): Add gcc_assert.
15365 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15367 Don't save registers in main().
15370 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15371 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15372 * config/avr/avr.c (avr_set_current_function): Don't error if
15373 naked, OS_task or OS_main are specified at the same time.
15374 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15376 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15378 * common/config/avr/avr-common.c (avr_option_optimization_table):
15379 Switch on -mmain-is-OS_task for optimizing compilations.
15381 2018-01-09 Richard Biener <rguenther@suse.de>
15383 PR tree-optimization/83572
15384 * graphite.c: Include cfganal.h.
15385 (graphite_transform_loops): Connect infinite loops to exit
15386 and remove fake edges at the end.
15388 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15390 * ipa-inline.c (edge_badness): Revert accidental checkin.
15392 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15395 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15396 symbols; not inline clones.
15398 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15401 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15402 hard registers. Formatting fixes.
15404 PR preprocessor/83722
15405 * gcc.c (try_generate_repro): Pass
15406 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15407 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15410 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15411 Kito Cheng <kito.cheng@gmail.com>
15413 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15414 (riscv_leaf_function_p): Delete.
15415 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15417 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15419 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15421 (do_ifelse): New function.
15422 (do_isel): New function.
15423 (do_sub3): New function.
15424 (do_add3): New function.
15425 (do_load_mask_compare): New function.
15426 (do_overlap_load_compare): New function.
15427 (expand_compare_loop): New function.
15428 (expand_block_compare): Call expand_compare_loop() when appropriate.
15429 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15430 option description.
15431 (-mblock-compare-inline-loop-limit): New option.
15433 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15436 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15437 Reverse order of second and third operands in first alternative.
15438 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15439 of first and second elements in UNSPEC_VPERMR vector.
15440 (altivec_expand_vec_perm_le): Likewise.
15442 2018-01-08 Jeff Law <law@redhat.com>
15444 PR rtl-optimizatin/81308
15445 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15446 (process_switch): If group_case_labels makes a change, then set
15448 (pass_convert_switch::execute): If a switch is converted, then
15449 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15451 PR rtl-optimization/81308
15452 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15455 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15457 PR target/83663 - Revert r255946
15458 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15459 generation for cases where splatting a value is not useful.
15460 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15461 across a vec_duplicate and a paradoxical subreg forming a vector
15462 mode to a vec_concat.
15464 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15466 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15467 -march=armv8.3-a variants.
15468 * config/arm/t-multilib: Likewise.
15469 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15471 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15473 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15475 (cceq_ior_compare_complement): Give it a name so I can use it, and
15476 change boolean_or_operator predicate to boolean_operator so it can
15477 be used to generate a crand.
15478 (eqne): New code iterator.
15479 (bd/bd_neg): New code_attrs.
15480 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15481 a single define_insn.
15482 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15483 decrement (bdnzt/bdnzf/bdzt/bdzf).
15484 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15485 with the new names of the branch decrement patterns, and added the
15486 names of the branch decrement conditional patterns.
15488 2018-01-08 Richard Biener <rguenther@suse.de>
15490 PR tree-optimization/83563
15491 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15494 2018-01-08 Richard Biener <rguenther@suse.de>
15496 PR middle-end/83713
15497 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15499 2018-01-08 Richard Biener <rguenther@suse.de>
15501 PR tree-optimization/83685
15502 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15503 references to abnormals.
15505 2018-01-08 Richard Biener <rguenther@suse.de>
15508 * dwarf2out.c (output_indirect_strings): Handle empty
15509 skeleton_debug_str_hash.
15510 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15512 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15514 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15515 (emit_store_direct): Likewise.
15516 (arc_trampoline_adjust_address): Likewise.
15517 (arc_asm_trampoline_template): New function.
15518 (arc_initialize_trampoline): Use asm_trampoline_template.
15519 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15520 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15521 * config/arc/arc.md (flush_icache): Delete pattern.
15523 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15525 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15526 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15529 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15532 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15533 by not USED_FOR_TARGET.
15534 (make_pass_resolve_sw_modes): Likewise.
15536 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15538 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15541 2018-01-08 Richard Biener <rguenther@suse.de>
15543 PR middle-end/83580
15544 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15546 2018-01-08 Richard Biener <rguenther@suse.de>
15548 PR middle-end/83517
15549 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15551 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15553 PR middle-end/81897
15554 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15555 basic blocks with a small number of successors.
15556 (convert_control_dep_chain_into_preds): Improve handling of
15558 (dump_predicates): Split apart into...
15559 (dump_pred_chain): ...here...
15560 (dump_pred_info): ...and here.
15561 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15562 (can_chain_union_be_invalidated_p): Improve check for invalidation
15564 (uninit_uses_cannot_happen): Avoid unnecessary if
15565 convert_control_dep_chain_into_preds yielded nothing.
15567 2018-01-06 Martin Sebor <msebor@redhat.com>
15569 PR tree-optimization/83640
15570 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15571 subtracting negative offset from size.
15572 (builtin_access::overlap): Adjust offset bounds of the access to fall
15573 within the size of the object if possible.
15575 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15577 PR rtl-optimization/83699
15578 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15579 extract_bit_field_as_subreg to cases in which the extracted
15580 value is also a vector.
15582 * lra-constraints.c (process_alt_operands): Test for the equivalence
15583 substitutions when detecting a possible reload cycle.
15585 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15588 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15589 by default if flag_selective_schedling{,2}. Formatting fixes.
15591 PR rtl-optimization/83682
15592 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15593 if it has non-VECTOR_MODE element mode.
15594 (vec_duplicate_p): Likewise.
15596 PR middle-end/83694
15597 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15598 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15600 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15603 * config/i386/i386-builtin.def
15604 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15605 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15606 Require also OPTION_MASK_ISA_AVX512F in addition to
15607 OPTION_MASK_ISA_GFNI.
15608 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15609 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15610 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15611 to OPTION_MASK_ISA_GFNI.
15612 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15613 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15614 OPTION_MASK_ISA_AVX512BW.
15615 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15616 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15617 addition to OPTION_MASK_ISA_GFNI.
15618 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15619 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15620 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15621 to OPTION_MASK_ISA_GFNI.
15622 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15623 a requirement for all ISAs rather than any of them with a few
15625 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15627 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15628 bitmasks to be enabled with 3 exceptions, instead of requiring any
15629 enabled ISA with lots of exceptions.
15630 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15631 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15632 Change avx512bw in isa attribute to avx512f.
15633 * config/i386/sgxintrin.h: Add license boilerplate.
15634 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15635 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15636 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15637 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15639 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15640 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15641 temporarily sse2 rather than sse if not enabled already.
15644 * config/i386/sse.md (VI248_VLBW): Rename to ...
15645 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15646 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15647 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15648 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15649 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15650 mode iterator instead of VI248_VLBW.
15652 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15654 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15655 (record_modified): Skip clobbers; add debug output.
15656 (param_change_prob): Use sreal frequencies.
15658 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15660 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15661 punt for user-aligned variables.
15663 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15665 * tree-chrec.c (chrec_contains_symbols): Return true for
15668 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15671 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15672 of (x|y) == x for BICS pattern.
15674 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15676 PR tree-optimization/83605
15677 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15678 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15681 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15683 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15684 * config/epiphany/rtems.h: New file.
15686 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15687 Uros Bizjak <ubizjak@gmail.com>
15690 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15691 QIreg_operand instead of register_operand predicate.
15692 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15693 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15694 comments instead of -fmitigate[-_]rop.
15696 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15699 * cgraphunit.c (symbol_table::compile): Switch to text_section
15700 before calling assembly_start debug hook.
15701 * run-rtl-passes.c (run_rtl_passes): Likewise.
15704 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15706 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15707 range_int_cst_p rather than !symbolic_range_p before calling
15708 extract_range_from_multiplicative_op_1.
15710 2018-01-04 Jeff Law <law@redhat.com>
15712 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15713 redundant test in assertion.
15715 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15717 * doc/rtl.texi: Document machine_mode wrapper classes.
15719 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15721 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15722 using tree_to_uhwi.
15724 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15726 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15727 the VEC_PERM_EXPR fold to fail.
15729 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15732 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15733 to switched_sections.
15735 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15738 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15739 test for d.testing.
15741 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15744 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15745 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15747 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15750 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15751 is BLKmode and bitpos not zero or mode change is needed.
15753 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15756 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15759 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15762 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15763 instead of MULT rtx. Update all corresponding splitters.
15764 (*saddl_se): Ditto.
15765 (*ssub<modesuffix>): Ditto.
15766 (*ssubl_se): Ditto.
15767 (*cmp_sadd_di): Update split patterns.
15768 (*cmp_sadd_si): Ditto.
15769 (*cmp_sadd_sidi): Ditto.
15770 (*cmp_ssub_di): Ditto.
15771 (*cmp_ssub_si): Ditto.
15772 (*cmp_ssub_sidi): Ditto.
15773 * config/alpha/predicates.md (const23_operand): New predicate.
15774 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15775 Look for ASHIFT, not MULT inner operand.
15776 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15778 2018-01-04 Martin Liska <mliska@suse.cz>
15780 PR gcov-profile/83669
15781 * gcov.c (output_intermediate_file): Add version to intermediate
15783 * doc/gcov.texi: Document new field 'version' in intermediate
15784 file format. Fix location of '-k' option of gcov command.
15786 2018-01-04 Martin Liska <mliska@suse.cz>
15789 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15791 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15793 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15795 2018-01-03 Martin Sebor <msebor@redhat.com>
15797 PR tree-optimization/83655
15798 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15799 checking calls with invalid arguments.
15801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15803 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15804 (vectorizable_mask_load_store): Delete.
15805 (vectorizable_call): Return false for masked loads and stores.
15806 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15807 instead of gimple_assign_rhs1.
15808 (vectorizable_load): Handle IFN_MASK_LOAD.
15809 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15813 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15815 (vectorizable_mask_load_store): ...here.
15816 (vectorizable_load): ...and here.
15818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15820 * tree-vect-stmts.c (vect_build_all_ones_mask)
15821 (vect_build_zero_merge_argument): New functions, split out from...
15822 (vectorizable_load): ...here.
15824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15826 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15828 (vectorizable_mask_load_store): ...here.
15829 (vectorizable_store): ...and here.
15831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15833 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15835 (vectorizable_mask_load_store): ...here.
15837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15839 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15840 (vect_model_store_cost): Take a vec_load_store_type instead of a
15842 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15843 (vect_model_store_cost): Take a vec_load_store_type instead of a
15845 (vectorizable_mask_load_store): Update accordingly.
15846 (vectorizable_store): Likewise.
15847 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15851 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15852 IFN_MASK_LOAD calls here rather than...
15853 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15856 Alan Hayward <alan.hayward@arm.com>
15857 David Sherwood <david.sherwood@arm.com>
15859 * expmed.c (extract_bit_field_1): For vector extracts,
15860 fall back to extract_bit_field_as_subreg if vec_extract
15863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15864 Alan Hayward <alan.hayward@arm.com>
15865 David Sherwood <david.sherwood@arm.com>
15867 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15868 they are variable or constant sized.
15869 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15870 slots for constant-sized data.
15872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15873 Alan Hayward <alan.hayward@arm.com>
15874 David Sherwood <david.sherwood@arm.com>
15876 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15877 handling COND_EXPRs with boolean comparisons, try to find a better
15878 basis for the mask type than the boolean itself.
15880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15882 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15883 is calculated and how it can be overridden.
15884 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15885 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15887 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15891 Alan Hayward <alan.hayward@arm.com>
15892 David Sherwood <david.sherwood@arm.com>
15894 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15895 Remove the mode argument.
15896 (aarch64_simd_valid_immediate): Remove the mode and inverse
15898 * config/aarch64/iterators.md (bitsize): New iterator.
15899 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15900 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15901 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15902 aarch64_simd_valid_immediate.
15903 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15904 (aarch64_reg_or_bic_imm): Likewise.
15905 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15906 with an insn_type enum and msl with a modifier_type enum.
15907 Replace element_width with a scalar_mode. Change the shift
15908 to unsigned int. Add constructors for scalar_float_mode and
15909 scalar_int_mode elements.
15910 (aarch64_vect_float_const_representable_p): Delete.
15911 (aarch64_can_const_movi_rtx_p)
15912 (aarch64_simd_scalar_immediate_valid_for_move)
15913 (aarch64_simd_make_constant): Update call to
15914 aarch64_simd_valid_immediate.
15915 (aarch64_advsimd_valid_immediate_hs): New function.
15916 (aarch64_advsimd_valid_immediate): Likewise.
15917 (aarch64_simd_valid_immediate): Remove mode and inverse
15918 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15919 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15920 and aarch64_float_const_representable_p on the result.
15921 (aarch64_output_simd_mov_immediate): Remove mode argument.
15922 Update call to aarch64_simd_valid_immediate and use of
15923 simd_immediate_info.
15924 (aarch64_output_scalar_simd_mov_immediate): Update call
15927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15928 Alan Hayward <alan.hayward@arm.com>
15929 David Sherwood <david.sherwood@arm.com>
15931 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15932 (mode_nunits): Likewise CONST_MODE_NUNITS.
15933 * machmode.def (ADJUST_NUNITS): Document.
15934 * genmodes.c (mode_data::need_nunits_adj): New field.
15935 (blank_mode): Update accordingly.
15936 (adj_nunits): New variable.
15937 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15939 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15940 listed in adj_nunits.
15941 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15942 listed in adj_nunits. Don't emit case statements for such modes.
15943 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15944 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15945 nothing if adj_nunits is nonnull.
15946 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15947 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15948 (emit_mode_fbit): Update use of print_maybe_const_decl.
15949 (emit_move_size): Likewise. Treat the array as non-const
15951 (emit_mode_adjustments): Handle adj_nunits.
15953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15955 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15956 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15957 (VECTOR_MODES): Use it.
15958 (make_vector_modes): Take the prefix as an argument.
15960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15961 Alan Hayward <alan.hayward@arm.com>
15962 David Sherwood <david.sherwood@arm.com>
15964 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15965 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15966 for MODE_VECTOR_BOOL.
15967 * machmode.def (VECTOR_BOOL_MODE): Document.
15968 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15969 (make_vector_bool_mode): New function.
15970 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15972 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15973 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15975 * stor-layout.c (int_mode_for_mode): Likewise.
15976 * tree.c (build_vector_type_for_mode): Likewise.
15977 * varasm.c (output_constant_pool_2): Likewise.
15978 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15979 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15980 for MODE_VECTOR_BOOL.
15981 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15982 of mode class checks.
15983 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15984 instead of a list of mode class checks.
15985 (expand_vector_scalar_condition): Likewise.
15986 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15989 Alan Hayward <alan.hayward@arm.com>
15990 David Sherwood <david.sherwood@arm.com>
15992 * machmode.h (mode_size): Change from unsigned short to
15994 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15995 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15996 or if measurement_type is not polynomial.
15997 (fixed_size_mode::includes_p): Check for constant-sized modes.
15998 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15999 return a poly_uint16 rather than an unsigned short.
16000 (emit_mode_size): Change the type of mode_size from unsigned short
16001 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16002 (emit_mode_adjustments): Cope with polynomial vector sizes.
16003 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16005 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16007 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16008 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16009 * caller-save.c (setup_save_areas): Likewise.
16010 (replace_reg_with_saved_mem): Likewise.
16011 * calls.c (emit_library_call_value_1): Likewise.
16012 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16013 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16014 (gen_lowpart_for_combine): Likewise.
16015 * convert.c (convert_to_integer_1): Likewise.
16016 * cse.c (equiv_constant, cse_insn): Likewise.
16017 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16018 (cselib_subst_to_values): Likewise.
16019 * dce.c (word_dce_process_block): Likewise.
16020 * df-problems.c (df_word_lr_mark_ref): Likewise.
16021 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16022 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16023 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16024 (rtl_for_decl_location): Likewise.
16025 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16026 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16027 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16028 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16029 (expand_expr_real_1): Likewise.
16030 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16031 (pad_below): Likewise.
16032 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16033 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16034 * ira.c (get_subreg_tracking_sizes): Likewise.
16035 * ira-build.c (ira_create_allocno_objects): Likewise.
16036 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16037 (ira_sort_regnos_for_alter_reg): Likewise.
16038 * ira-costs.c (record_operand_costs): Likewise.
16039 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16040 (resolve_simple_move): Likewise.
16041 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16042 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16043 (lra_constraints): Likewise.
16044 (CONST_POOL_OK_P): Reject variable-sized modes.
16045 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16046 (add_pseudo_to_slot, lra_spill): Likewise.
16047 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16048 * optabs-query.c (get_best_extraction_insn): Likewise.
16049 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16050 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16051 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16052 * recog.c (offsettable_address_addr_space_p): Likewise.
16053 * regcprop.c (maybe_mode_change): Likewise.
16054 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16055 * regrename.c (build_def_use): Likewise.
16056 * regstat.c (dump_reg_info): Likewise.
16057 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16058 (find_reloads, find_reloads_subreg_address): Likewise.
16059 * reload1.c (eliminate_regs_1): Likewise.
16060 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16061 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16062 (simplify_binary_operation_1, simplify_subreg): Likewise.
16063 * targhooks.c (default_function_arg_padding): Likewise.
16064 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16065 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16066 (verify_gimple_assign_ternary): Likewise.
16067 * tree-inline.c (estimate_move_cost): Likewise.
16068 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16069 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16070 (get_address_cost_ainc): Likewise.
16071 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16072 (vect_supportable_dr_alignment): Likewise.
16073 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16074 (vectorizable_reduction): Likewise.
16075 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16076 (vectorizable_operation, vectorizable_load): Likewise.
16077 * tree.c (build_same_sized_truth_vector_type): Likewise.
16078 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16079 * var-tracking.c (emit_note_insn_var_location): Likewise.
16080 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16081 (ADDR_VEC_ALIGN): Likewise.
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_to_bits): Return a poly_uint16 rather than an
16089 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16090 or if measurement_type is polynomial.
16091 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16092 * combine.c (make_extraction): Likewise.
16093 * dse.c (find_shift_sequence): Likewise.
16094 * dwarf2out.c (mem_loc_descriptor): Likewise.
16095 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16096 (extract_bit_field, extract_low_bits): Likewise.
16097 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16098 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16099 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16100 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16101 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16102 * reload.c (find_reloads): Likewise.
16103 * reload1.c (alter_reg): Likewise.
16104 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16105 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16106 * tree-if-conv.c (predicate_mem_writes): Likewise.
16107 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16108 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16109 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16110 * valtrack.c (dead_debug_insert_temp): Likewise.
16111 * varasm.c (mergeable_constant_section): Likewise.
16112 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16115 Alan Hayward <alan.hayward@arm.com>
16116 David Sherwood <david.sherwood@arm.com>
16118 * expr.c (expand_assignment): Cope with polynomial mode sizes
16119 when assigning to a CONCAT.
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 * machmode.h (mode_precision): Change from unsigned short to
16127 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16129 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16130 or if measurement_type is not polynomial.
16131 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16132 in which the mode is already known to be a scalar_int_mode.
16133 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16134 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16136 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16137 for GET_MODE_PRECISION.
16138 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16139 for GET_MODE_PRECISION.
16140 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16142 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16143 (expand_field_assignment, make_extraction): Likewise.
16144 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16145 (get_last_value): Likewise.
16146 * convert.c (convert_to_integer_1): Likewise.
16147 * cse.c (cse_insn): Likewise.
16148 * expr.c (expand_expr_real_1): Likewise.
16149 * lra-constraints.c (simplify_operand_subreg): Likewise.
16150 * optabs-query.c (can_atomic_load_p): Likewise.
16151 * optabs.c (expand_atomic_load): Likewise.
16152 (expand_atomic_store): Likewise.
16153 * ree.c (combine_reaching_defs): Likewise.
16154 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16155 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16156 * tree.h (type_has_mode_precision_p): Likewise.
16157 * ubsan.c (instrument_si_overflow): Likewise.
16159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16160 Alan Hayward <alan.hayward@arm.com>
16161 David Sherwood <david.sherwood@arm.com>
16163 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16164 polynomial numbers of units.
16165 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16166 (valid_vector_subparts_p): New function.
16167 (build_vector_type): Remove temporary shim and take the number
16168 of units as a poly_uint64 rather than an int.
16169 (build_opaque_vector_type): Take the number of units as a
16170 poly_uint64 rather than an int.
16171 * tree.c (build_vector_from_ctor): Handle polynomial
16172 TYPE_VECTOR_SUBPARTS.
16173 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16174 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16175 (build_vector_from_val): If the number of units is variable,
16176 use build_vec_duplicate_cst for constant operands and
16177 VEC_DUPLICATE_EXPR otherwise.
16178 (make_vector_type): Remove temporary is_constant ().
16179 (build_vector_type, build_opaque_vector_type): Take the number of
16180 units as a poly_uint64 rather than an int.
16181 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16183 * cfgexpand.c (expand_debug_expr): Likewise.
16184 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16185 (store_constructor, expand_expr_real_1): Likewise.
16186 (const_scalar_mask_from_tree): Likewise.
16187 * fold-const-call.c (fold_const_reduction): Likewise.
16188 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16189 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16190 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16191 (fold_relational_const): Likewise.
16192 (native_interpret_vector): Likewise. Change the size from an
16193 int to an unsigned int.
16194 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16195 TYPE_VECTOR_SUBPARTS.
16196 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16197 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16198 duplicating a non-constant operand into a variable-length vector.
16199 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16200 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16201 * ipa-icf.c (sem_variable::equals): Likewise.
16202 * match.pd: Likewise.
16203 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16204 * print-tree.c (print_node): Likewise.
16205 * stor-layout.c (layout_type): Likewise.
16206 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16207 * tree-cfg.c (verify_gimple_comparison): Likewise.
16208 (verify_gimple_assign_binary): Likewise.
16209 (verify_gimple_assign_ternary): Likewise.
16210 (verify_gimple_assign_single): Likewise.
16211 * tree-pretty-print.c (dump_generic_node): Likewise.
16212 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16213 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16214 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16215 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16216 (vect_shift_permute_load_chain): Likewise.
16217 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16218 (expand_vector_condition, optimize_vector_constructor): Likewise.
16219 (lower_vec_perm, get_compute_type): Likewise.
16220 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16221 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16222 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16223 (vect_recog_mask_conversion_pattern): Likewise.
16224 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16225 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16226 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16227 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16228 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16229 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16230 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16231 (supportable_widening_operation): Likewise.
16232 (supportable_narrowing_operation): Likewise.
16233 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16235 * varasm.c (output_constant): Likewise.
16237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16238 Alan Hayward <alan.hayward@arm.com>
16239 David Sherwood <david.sherwood@arm.com>
16241 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16242 so that both the length == 3 and length != 3 cases set up their
16243 own permute vectors. Add comments explaining why we know the
16244 number of elements is constant.
16245 (vect_permute_load_chain): Likewise.
16247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16248 Alan Hayward <alan.hayward@arm.com>
16249 David Sherwood <david.sherwood@arm.com>
16251 * machmode.h (mode_nunits): Change from unsigned char to
16253 (ONLY_FIXED_SIZE_MODES): New macro.
16254 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16255 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16256 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16258 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16259 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16260 or if measurement_type is not polynomial.
16261 * genmodes.c (ZERO_COEFFS): New macro.
16262 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16264 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16265 Use ZERO_COEFFS when emitting initializers.
16266 * data-streamer.h (bp_pack_poly_value): New function.
16267 (bp_unpack_poly_value): Likewise.
16268 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16269 for GET_MODE_NUNITS.
16270 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16271 for GET_MODE_NUNITS.
16272 * tree.c (make_vector_type): Remove temporary shim and make
16273 the real function take the number of units as a poly_uint64
16274 rather than an int.
16275 (build_vector_type_for_mode): Handle polynomial nunits.
16276 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16277 * emit-rtl.c (const_vec_series_p_1): Likewise.
16278 (gen_rtx_CONST_VECTOR): Likewise.
16279 * fold-const.c (test_vec_duplicate_folding): Likewise.
16280 * genrecog.c (validate_pattern): Likewise.
16281 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16282 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16283 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16284 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16285 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16286 * rtlanal.c (subreg_get_info): Likewise.
16287 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16288 (vect_grouped_load_supported): Likewise.
16289 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16290 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16291 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16292 (simplify_const_unary_operation, simplify_binary_operation_1)
16293 (simplify_const_binary_operation, simplify_ternary_operation)
16294 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16295 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16296 instead of CONST_VECTOR_NUNITS.
16297 * varasm.c (output_constant_pool_2): Likewise.
16298 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16299 explicit-encoded elements in the XVEC for variable-length vectors.
16301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16303 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16306 Alan Hayward <alan.hayward@arm.com>
16307 David Sherwood <david.sherwood@arm.com>
16309 * coretypes.h (fixed_size_mode): Declare.
16310 (fixed_size_mode_pod): New typedef.
16311 * builtins.h (target_builtins::x_apply_args_mode)
16312 (target_builtins::x_apply_result_mode): Change type to
16313 fixed_size_mode_pod.
16314 * builtins.c (apply_args_size, apply_result_size, result_vector)
16315 (expand_builtin_apply_args_1, expand_builtin_apply)
16316 (expand_builtin_return): Update accordingly.
16318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16320 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16321 * cselib.c (cselib_hash_rtx): Likewise.
16322 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16323 CONST_VECTOR encoding.
16325 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16326 Jeff Law <law@redhat.com>
16329 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16330 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16331 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16332 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16335 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16336 explicitly probe *sp in a noreturn function if there were any callee
16337 register saves or frame pointer is needed.
16339 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16342 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16343 BLKmode for ternary, binary or unary expressions.
16346 * var-tracking.c (delete_vta_debug_insn): New inline function.
16347 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16348 insns from get_insns () to NULL instead of each bb separately.
16349 Use delete_vta_debug_insn. No longer static.
16350 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16351 delete_vta_debug_insns callers.
16352 * rtl.h (delete_vta_debug_insns): Declare.
16353 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16354 instead of variable_tracking_main.
16356 2018-01-03 Martin Sebor <msebor@redhat.com>
16358 PR tree-optimization/83603
16359 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16360 arguments past the endof the argument list in functions declared
16361 without a prototype.
16362 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16363 Avoid checking when arguments are null.
16365 2018-01-03 Martin Sebor <msebor@redhat.com>
16368 * doc/extend.texi (attribute const): Fix a typo.
16369 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16370 issuing -Wsuggest-attribute for void functions.
16372 2018-01-03 Martin Sebor <msebor@redhat.com>
16374 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16375 offset_int::from instead of wide_int::to_shwi.
16376 (maybe_diag_overlap): Remove assertion.
16377 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16378 * gimple-ssa-sprintf.c (format_directive): Same.
16379 (parse_directive): Same.
16380 (sprintf_dom_walker::compute_format_length): Same.
16381 (try_substitute_return_value): Same.
16383 2018-01-03 Jeff Law <law@redhat.com>
16385 PR middle-end/83654
16386 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16387 non-constant residual for zero at runtime and avoid probing in
16388 that case. Reorganize code for trailing problem to mirror handling
16391 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16393 PR tree-optimization/83501
16394 * tree-ssa-strlen.c (get_string_cst): New.
16395 (handle_char_store): Call get_string_cst.
16397 2018-01-03 Martin Liska <mliska@suse.cz>
16399 PR tree-optimization/83593
16400 * tree-ssa-strlen.c: Include tree-cfg.h.
16401 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16402 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16403 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16405 (strlen_dom_walker::before_dom_children): Call
16406 gimple_purge_dead_eh_edges. Dump tranformation with details
16408 (strlen_dom_walker::before_dom_children): Update call by adding
16409 new argument cleanup_eh.
16410 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16412 2018-01-03 Martin Liska <mliska@suse.cz>
16415 * cif-code.def (VARIADIC_THUNK): New enum value.
16416 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16419 2018-01-03 Jan Beulich <jbeulich@suse.com>
16421 * sse.md (mov<mode>_internal): Tighten condition for when to use
16422 vmovdqu<ssescalarsize> for TI and OI modes.
16424 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16426 Update copyright years.
16428 2018-01-03 Martin Liska <mliska@suse.cz>
16431 * ipa-visibility.c (function_and_variable_visibility): Skip
16432 functions with noipa attribure.
16434 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16436 * gcc.c (process_command): Update copyright notice dates.
16437 * gcov-dump.c (print_version): Ditto.
16438 * gcov.c (print_version): Ditto.
16439 * gcov-tool.c (print_version): Ditto.
16440 * gengtype.c (create_file): Ditto.
16441 * doc/cpp.texi: Bump @copying's copyright year.
16442 * doc/cppinternals.texi: Ditto.
16443 * doc/gcc.texi: Ditto.
16444 * doc/gccint.texi: Ditto.
16445 * doc/gcov.texi: Ditto.
16446 * doc/install.texi: Ditto.
16447 * doc/invoke.texi: Ditto.
16449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16451 * vector-builder.h (vector_builder::m_full_nelts): Change from
16452 unsigned int to poly_uint64.
16453 (vector_builder::full_nelts): Update prototype accordingly.
16454 (vector_builder::new_vector): Likewise.
16455 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16456 (vector_builder::operator ==): Likewise.
16457 (vector_builder::finalize): Likewise.
16458 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16459 Take the number of elements as a poly_uint64 rather than an
16461 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16462 from unsigned int to poly_uint64.
16463 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16464 (vec_perm_indices::new_vector): Likewise.
16465 (vec_perm_indices::length): Likewise.
16466 (vec_perm_indices::nelts_per_input): Likewise.
16467 (vec_perm_indices::input_nelts): Likewise.
16468 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16469 number of elements per input as a poly_uint64 rather than an
16470 unsigned int. Use the original encoding for variable-length
16471 vectors, rather than clamping each individual element.
16472 For the second and subsequent elements in each pattern,
16473 clamp the step and base before clamping their sum.
16474 (vec_perm_indices::series_p): Handle polynomial element counts.
16475 (vec_perm_indices::all_in_range_p): Likewise.
16476 (vec_perm_indices_to_tree): Likewise.
16477 (vec_perm_indices_to_rtx): Likewise.
16478 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16479 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16480 (tree_vector_builder::new_binary_operation): Handle polynomial
16481 element counts. Return false if we need to know the number
16482 of elements at compile time.
16483 * fold-const.c (fold_vec_perm): Punt if the number of elements
16484 isn't known at compile time.
16486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16488 * vec-perm-indices.h (vec_perm_builder): Change element type
16489 from HOST_WIDE_INT to poly_int64.
16490 (vec_perm_indices::element_type): Update accordingly.
16491 (vec_perm_indices::clamp): Handle polynomial element_types.
16492 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16493 (vec_perm_indices::all_in_range_p): Likewise.
16494 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16496 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16497 polynomial vec_perm_indices element types.
16498 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16499 * fold-const.c (fold_vec_perm): Likewise.
16500 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16501 * tree-vect-generic.c (lower_vec_perm): Likewise.
16502 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16503 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16504 element type to HOST_WIDE_INT.
16506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16507 Alan Hayward <alan.hayward@arm.com>
16508 David Sherwood <david.sherwood@arm.com>
16510 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16511 rather than an int. Use plus_constant.
16512 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16513 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16516 Alan Hayward <alan.hayward@arm.com>
16517 David Sherwood <david.sherwood@arm.com>
16519 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16520 a HOST_WIDE_INT to a poly_int64.
16522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16523 Alan Hayward <alan.hayward@arm.com>
16524 David Sherwood <david.sherwood@arm.com>
16526 * calls.c (load_register_parameters): Cope with polynomial
16527 mode sizes. Require a constant size for BLKmode parameters
16528 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16529 forces a parameter to be padded at the lsb end in order to
16530 fill a complete number of words, require the parameter size
16531 to be ordered wrt UNITS_PER_WORD.
16533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16534 Alan Hayward <alan.hayward@arm.com>
16535 David Sherwood <david.sherwood@arm.com>
16537 * reload1.c (spill_stack_slot_width): Change element type
16538 from unsigned int to poly_uint64_pod.
16539 (alter_reg): Treat mode sizes as polynomial.
16541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16542 Alan Hayward <alan.hayward@arm.com>
16543 David Sherwood <david.sherwood@arm.com>
16545 * reload.c (complex_word_subreg_p): New function.
16546 (reload_inner_reg_of_subreg, push_reload): Use it.
16548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16549 Alan Hayward <alan.hayward@arm.com>
16550 David Sherwood <david.sherwood@arm.com>
16552 * lra-constraints.c (process_alt_operands): Reject matched
16553 operands whose sizes aren't ordered.
16554 (match_reload): Refer to this check here.
16556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16557 Alan Hayward <alan.hayward@arm.com>
16558 David Sherwood <david.sherwood@arm.com>
16560 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16561 that the mode size is in the set {1, 2, 4, 8, 16}.
16563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16564 Alan Hayward <alan.hayward@arm.com>
16565 David Sherwood <david.sherwood@arm.com>
16567 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16568 Use plus_constant instead of gen_rtx_PLUS.
16570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16571 Alan Hayward <alan.hayward@arm.com>
16572 David Sherwood <david.sherwood@arm.com>
16574 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16575 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16576 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16577 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16578 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16579 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16580 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16581 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16582 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16583 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16585 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16586 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16587 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16588 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16589 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16590 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16591 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16592 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16593 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16594 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16596 * expr.c (emit_move_resolve_push): Treat the input and result
16597 of PUSH_ROUNDING as a poly_int64.
16598 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16599 (emit_push_insn): Likewise.
16600 * lra-eliminations.c (mark_not_eliminable): Likewise.
16601 * recog.c (push_operand): Likewise.
16602 * reload1.c (elimination_effects): Likewise.
16603 * rtlanal.c (nonzero_bits1): Likewise.
16604 * calls.c (store_one_arg): Likewise. Require the padding to be
16605 known at compile time.
16607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16608 Alan Hayward <alan.hayward@arm.com>
16609 David Sherwood <david.sherwood@arm.com>
16611 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16612 Use plus_constant instead of gen_rtx_PLUS.
16614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16615 Alan Hayward <alan.hayward@arm.com>
16616 David Sherwood <david.sherwood@arm.com>
16618 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16619 rather than an int.
16621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16622 Alan Hayward <alan.hayward@arm.com>
16623 David Sherwood <david.sherwood@arm.com>
16625 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16626 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16627 via stack temporaries. Treat the mode size as polynomial too.
16629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16630 Alan Hayward <alan.hayward@arm.com>
16631 David Sherwood <david.sherwood@arm.com>
16633 * expr.c (expand_expr_real_2): When handling conversions involving
16634 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16635 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16636 as a poly_uint64 too.
16638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16639 Alan Hayward <alan.hayward@arm.com>
16640 David Sherwood <david.sherwood@arm.com>
16642 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16645 Alan Hayward <alan.hayward@arm.com>
16646 David Sherwood <david.sherwood@arm.com>
16648 * combine.c (can_change_dest_mode): Handle polynomial
16649 REGMODE_NATURAL_SIZE.
16650 * expmed.c (store_bit_field_1): Likewise.
16651 * expr.c (store_constructor): Likewise.
16652 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16653 and polynomial REGMODE_NATURAL_SIZE.
16654 (gen_lowpart_common): Likewise.
16655 * reginfo.c (record_subregs_of_mode): Likewise.
16656 * rtlanal.c (read_modify_subreg_p): Likewise.
16658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16659 Alan Hayward <alan.hayward@arm.com>
16660 David Sherwood <david.sherwood@arm.com>
16662 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16663 numbers of elements.
16665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16666 Alan Hayward <alan.hayward@arm.com>
16667 David Sherwood <david.sherwood@arm.com>
16669 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16676 in a POINTER_PLUS_EXPR.
16678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16679 Alan Hayward <alan.hayward@arm.com>
16680 David Sherwood <david.sherwood@arm.com>
16682 * omp-simd-clone.c (simd_clone_subparts): New function.
16683 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16684 (ipa_simd_modify_function_body): Likewise.
16686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16687 Alan Hayward <alan.hayward@arm.com>
16688 David Sherwood <david.sherwood@arm.com>
16690 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16691 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16692 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16693 (expand_vector_condition, vector_element): Likewise.
16694 (subparts_gt): New function.
16695 (get_compute_type): Use subparts_gt.
16696 (count_type_subparts): Delete.
16697 (expand_vector_operations_1): Use subparts_gt instead of
16698 count_type_subparts.
16700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16701 Alan Hayward <alan.hayward@arm.com>
16702 David Sherwood <david.sherwood@arm.com>
16704 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16705 (vect_compile_time_alias): ...this new function. Do the calculation
16706 on poly_ints rather than trees.
16707 (vect_prune_runtime_alias_test_list): Update call accordingly.
16709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16710 Alan Hayward <alan.hayward@arm.com>
16711 David Sherwood <david.sherwood@arm.com>
16713 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16715 (vect_schedule_slp_instance): Likewise.
16717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16718 Alan Hayward <alan.hayward@arm.com>
16719 David Sherwood <david.sherwood@arm.com>
16721 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16722 constant and extern definitions for variable-length vectors.
16723 (vect_get_constant_vectors): Note that the number of units
16724 is known to be constant.
16726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16727 Alan Hayward <alan.hayward@arm.com>
16728 David Sherwood <david.sherwood@arm.com>
16730 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16731 of units as polynomial. Choose between WIDE and NARROW based
16734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16735 Alan Hayward <alan.hayward@arm.com>
16736 David Sherwood <david.sherwood@arm.com>
16738 * tree-vect-stmts.c (simd_clone_subparts): New function.
16739 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-vect-stmts.c (vectorizable_call): Treat the number of
16746 vectors as polynomial. Use build_index_vector for
16747 IFN_GOMP_SIMD_LANE.
16749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16750 Alan Hayward <alan.hayward@arm.com>
16751 David Sherwood <david.sherwood@arm.com>
16753 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16754 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16755 for variable-length vectors.
16756 (vectorizable_mask_load_store): Treat the number of units as
16757 polynomial, asserting that it is constant if the condition has
16758 already been enforced.
16759 (vectorizable_store, vectorizable_load): Likewise.
16761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16762 Alan Hayward <alan.hayward@arm.com>
16763 David Sherwood <david.sherwood@arm.com>
16765 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16766 of units as polynomial. Punt if we can't tell at compile time
16767 which vector contains the final result.
16769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16770 Alan Hayward <alan.hayward@arm.com>
16771 David Sherwood <david.sherwood@arm.com>
16773 * tree-vect-loop.c (vectorizable_induction): Treat the number
16774 of units as polynomial. Punt on SLP inductions. Use an integer
16775 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16776 cast of such a series for variable-length floating-point
16779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16780 Alan Hayward <alan.hayward@arm.com>
16781 David Sherwood <david.sherwood@arm.com>
16783 * tree.h (build_index_vector): Declare.
16784 * tree.c (build_index_vector): New function.
16785 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16786 of units as polynomial, forcibly converting it to a constant if
16787 vectorizable_reduction has already enforced the condition.
16788 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16789 to create a {1,2,3,...} vector.
16790 (vectorizable_reduction): Treat the number of units as polynomial.
16791 Choose vectype_in based on the largest scalar element size rather
16792 than the smallest number of units. Enforce the restrictions
16795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16796 Alan Hayward <alan.hayward@arm.com>
16797 David Sherwood <david.sherwood@arm.com>
16799 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16800 number of units as polynomial.
16802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16803 Alan Hayward <alan.hayward@arm.com>
16804 David Sherwood <david.sherwood@arm.com>
16806 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16807 * target.def (autovectorize_vector_sizes): Return the vector sizes
16808 by pointer, using vector_sizes rather than a bitmask.
16809 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16810 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16811 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16813 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16814 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16815 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16816 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16817 * omp-general.c (omp_max_vf): Likewise.
16818 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16819 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16820 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16821 * tree-vect-slp.c (vect_slp_bb): Likewise.
16822 * doc/tm.texi: Regenerate.
16823 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16825 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16826 the vector size as a poly_uint64 rather than an unsigned int.
16827 (current_vector_size): Change from an unsigned int to a poly_uint64.
16828 (get_vectype_for_scalar_type): Update accordingly.
16829 * tree.h (build_truth_vector_type): Take the size and number of
16830 units as a poly_uint64 rather than an unsigned int.
16831 (build_vector_type): Add a temporary overload that takes
16832 the number of units as a poly_uint64 rather than an unsigned int.
16833 * tree.c (make_vector_type): Likewise.
16834 (build_truth_vector_type): Take the number of units as a poly_uint64
16835 rather than an unsigned int.
16837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16838 Alan Hayward <alan.hayward@arm.com>
16839 David Sherwood <david.sherwood@arm.com>
16841 * target.def (get_mask_mode): Take the number of units and length
16842 as poly_uint64s rather than unsigned ints.
16843 * targhooks.h (default_get_mask_mode): Update accordingly.
16844 * targhooks.c (default_get_mask_mode): Likewise.
16845 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16846 * doc/tm.texi: Regenerate.
16848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16849 Alan Hayward <alan.hayward@arm.com>
16850 David Sherwood <david.sherwood@arm.com>
16852 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16853 * omp-general.c (omp_max_vf): Likewise.
16854 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16855 (expand_omp_simd): Handle polynomial safelen.
16856 * omp-low.c (omplow_simd_context): Add a default constructor.
16857 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16858 (lower_rec_simd_input_clauses): Update accordingly.
16859 (lower_rec_input_clauses): Likewise.
16861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16862 Alan Hayward <alan.hayward@arm.com>
16863 David Sherwood <david.sherwood@arm.com>
16865 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16866 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16867 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16868 (vect_analyze_slp_cost): Likewise.
16869 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16870 (vect_model_load_cost): Likewise.
16872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16873 Alan Hayward <alan.hayward@arm.com>
16874 David Sherwood <david.sherwood@arm.com>
16876 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16877 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16878 from an unsigned int * to a poly_uint64_pod *.
16879 (calculate_unrolling_factor): New function.
16880 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16883 Alan Hayward <alan.hayward@arm.com>
16884 David Sherwood <david.sherwood@arm.com>
16886 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16887 from an unsigned int to a poly_uint64.
16888 (_loop_vec_info::slp_unrolling_factor): Likewise.
16889 (_loop_vec_info::vectorization_factor): Change from an int
16891 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16892 (vect_get_num_vectors): New function.
16893 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16894 (vect_get_num_copies): Use vect_get_num_vectors.
16895 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16896 to an unsigned int *.
16897 (vect_analyze_data_refs): Change min_vf from an int * to a
16899 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16900 than an unsigned HOST_WIDE_INT.
16901 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16902 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16903 to an unsigned int *.
16904 (vect_analyze_data_ref_dependences): Likewise.
16905 (vect_compute_data_ref_alignment): Handle polynomial vf.
16906 (vect_enhance_data_refs_alignment): Likewise.
16907 (vect_prune_runtime_alias_test_list): Likewise.
16908 (vect_shift_permute_load_chain): Likewise.
16909 (vect_supportable_dr_alignment): Likewise.
16910 (dependence_distance_ge_vf): Take the vectorization factor as a
16911 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16912 (vect_analyze_data_refs): Change min_vf from an int * to a
16914 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16915 vfm1 as a poly_uint64 rather than an int. Make the same change
16916 for the returned bound_scalar.
16917 (vect_gen_vector_loop_niters): Handle polynomial vf.
16918 (vect_do_peeling): Likewise. Update call to
16919 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16920 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16922 * tree-vect-loop.c (vect_determine_vectorization_factor)
16923 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16924 (vect_get_known_peeling_cost): Likewise.
16925 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16926 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16927 (vect_transform_loop): Likewise. Use the lowest possible VF when
16928 updating the upper bounds of the loop.
16929 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16930 rather than an int.
16931 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16932 polynomial unroll factors.
16933 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16934 (vect_make_slp_decision): Likewise.
16935 (vect_supported_load_permutation_p): Likewise, and polynomial
16937 (vect_analyze_slp_cost): Handle polynomial vf.
16938 (vect_slp_analyze_node_operations): Likewise.
16939 (vect_slp_analyze_bb_1): Likewise.
16940 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16941 than an unsigned HOST_WIDE_INT.
16942 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16943 (vectorizable_load): Handle polynomial vf.
16944 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16946 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16949 Alan Hayward <alan.hayward@arm.com>
16950 David Sherwood <david.sherwood@arm.com>
16952 * match.pd: Handle bit operations involving three constants
16953 and try to fold one pair.
16955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16957 * tree-vect-loop-manip.c: Include gimple-fold.h.
16958 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16959 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16960 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16961 Add a path that uses a step of VF instead of 1, but disable it
16963 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16964 and niters_no_overflow parameters. Update calls to
16965 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16966 Create a new SSA name if the latter choses to use a ste other
16967 than zero, and return it via niters_vector_mult_vf_var.
16968 * tree-vect-loop.c (vect_transform_loop): Update calls to
16969 vect_do_peeling, vect_gen_vector_loop_niters and
16970 slpeel_make_loop_iterate_ntimes.
16971 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16972 (vect_gen_vector_loop_niters): Update declarations after above changes.
16974 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16976 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16977 128-bit round to integer instructions.
16978 (ceil<mode>2): Likewise.
16979 (btrunc<mode>2): Likewise.
16980 (round<mode>2): Likewise.
16982 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16984 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16985 unaligned VSX load/store on P8/P9.
16986 (expand_block_clear): Allow the use of unaligned VSX
16987 load/store on P8/P9.
16989 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16991 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16993 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16994 swap associated with both a load and a store.
16996 2018-01-02 Andrew Waterman <andrew@sifive.com>
16998 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16999 * config/riscv/riscv.md (clear_cache): Use it.
17001 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17003 * web.c: Remove out-of-date comment.
17005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17007 * expr.c (fixup_args_size_notes): Check that any existing
17008 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17009 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17010 (emit_single_push_insn): ...here.
17012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17014 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17015 (const_vector_encoded_nelts): New function.
17016 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17017 (const_vector_int_elt, const_vector_elt): Declare.
17018 * emit-rtl.c (const_vector_int_elt_1): New function.
17019 (const_vector_elt): Likewise.
17020 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17021 of CONST_VECTOR_ELT.
17023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17025 * expr.c: Include rtx-vector-builder.h.
17026 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17027 directly on the tree encoding.
17028 (const_vector_from_tree): Likewise.
17029 * optabs.c: Include rtx-vector-builder.h.
17030 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17031 sequence of "u" values.
17032 * vec-perm-indices.c: Include rtx-vector-builder.h.
17033 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17034 directly on the vec_perm_indices encoding.
17036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17038 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17039 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17040 * rtx-vector-builder.h: New file.
17041 * rtx-vector-builder.c: Likewise.
17042 * rtl.h (rtx_def::u2): Add a const_vector field.
17043 (CONST_VECTOR_NPATTERNS): New macro.
17044 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17045 (CONST_VECTOR_DUPLICATE_P): Likewise.
17046 (CONST_VECTOR_STEPPED_P): Likewise.
17047 (CONST_VECTOR_ENCODED_ELT): Likewise.
17048 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17049 (unwrap_const_vec_duplicate): Likewise.
17050 (const_vec_series_p): Check for a non-duplicated vector encoding.
17051 Say that the function only returns true for integer vectors.
17052 * emit-rtl.c: Include rtx-vector-builder.h.
17053 (gen_const_vec_duplicate_1): Delete.
17054 (gen_const_vector): Call gen_const_vec_duplicate instead of
17055 gen_const_vec_duplicate_1.
17056 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17057 (gen_const_vec_duplicate): Use rtx_vector_builder.
17058 (gen_const_vec_series): Likewise.
17059 (gen_rtx_CONST_VECTOR): Likewise.
17060 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17061 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17062 Build a new vector rather than modifying a CONST_VECTOR in-place.
17063 (handle_special_swappables): Update call accordingly.
17064 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17065 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17066 Build a new vector rather than modifying a CONST_VECTOR in-place.
17067 (handle_special_swappables): Update call accordingly.
17069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17071 * simplify-rtx.c (simplify_const_binary_operation): Use
17072 CONST_VECTOR_ELT instead of XVECEXP.
17074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17076 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17077 the selector elements to be different from the data elements
17078 if the selector is a VECTOR_CST.
17079 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17080 ssizetype for the selector.
17082 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17084 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17085 before testing each element individually.
17086 * tree-vect-generic.c (lower_vec_perm): Likewise.
17088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17090 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17091 * selftest-run-tests.c (selftest::run_tests): Call it.
17092 * vector-builder.h (vector_builder::operator ==): New function.
17093 (vector_builder::operator !=): Likewise.
17094 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17095 (vec_perm_indices::all_from_input_p): New function.
17096 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17097 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17098 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17099 instead of reading the VECTOR_CST directly. Detect whether both
17100 vector inputs are the same before constructing the vec_perm_indices,
17101 and update the number of inputs argument accordingly. Use the
17102 utility functions added above. Only construct sel2 if we need to.
17104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17106 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17107 the broadcast of the low byte.
17108 (expand_mult_highpart): Use an explicit encoding for the permutes.
17109 * optabs-query.c (can_mult_highpart_p): Likewise.
17110 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17111 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17112 (vectorizable_bswap): Likewise.
17113 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17114 explicit encoding for the power-of-2 permutes.
17115 (vect_permute_store_chain): Likewise.
17116 (vect_grouped_load_supported): Likewise.
17117 (vect_permute_load_chain): Likewise.
17119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17121 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17122 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17123 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17124 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17125 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17126 (vect_gen_perm_mask_any): Likewise.
17128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17130 * int-vector-builder.h: New file.
17131 * vec-perm-indices.h: Include int-vector-builder.h.
17132 (vec_perm_indices): Redefine as an int_vector_builder.
17133 (auto_vec_perm_indices): Delete.
17134 (vec_perm_builder): Redefine as a stand-alone class.
17135 (vec_perm_indices::vec_perm_indices): New function.
17136 (vec_perm_indices::clamp): Likewise.
17137 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17138 (vec_perm_indices::new_vector): New function.
17139 (vec_perm_indices::new_expanded_vector): Update for new
17140 vec_perm_indices class.
17141 (vec_perm_indices::rotate_inputs): New function.
17142 (vec_perm_indices::all_in_range_p): Operate directly on the
17143 encoded form, without computing elided elements.
17144 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17145 encoding. Update for new vec_perm_indices class.
17146 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17147 the given vec_perm_builder.
17148 (expand_vec_perm_var): Update vec_perm_builder constructor.
17149 (expand_mult_highpart): Use vec_perm_builder instead of
17150 auto_vec_perm_indices.
17151 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17152 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17153 or double series encoding as appropriate.
17154 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17155 vec_perm_indices instead of auto_vec_perm_indices.
17156 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17157 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17158 (vect_permute_store_chain): Likewise.
17159 (vect_grouped_load_supported): Likewise.
17160 (vect_permute_load_chain): Likewise.
17161 (vect_shift_permute_load_chain): Likewise.
17162 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17163 (vect_transform_slp_perm_load): Likewise.
17164 (vect_schedule_slp_instance): Likewise.
17165 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17166 (vectorizable_mask_load_store): Likewise.
17167 (vectorizable_bswap): Likewise.
17168 (vectorizable_store): Likewise.
17169 (vectorizable_load): Likewise.
17170 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17171 vec_perm_indices instead of auto_vec_perm_indices. Use
17172 tree_to_vec_perm_builder to read the vector from a tree.
17173 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17174 vec_perm_builder instead of a vec_perm_indices.
17175 (have_whole_vector_shift): Use vec_perm_builder and
17176 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17177 truncation to calc_vec_perm_mask_for_shift.
17178 (vect_create_epilog_for_reduction): Likewise.
17179 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17180 from auto_vec_perm_indices to vec_perm_indices.
17181 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17182 instead of changing individual elements.
17183 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17184 the vector in d.perm.
17185 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17186 from auto_vec_perm_indices to vec_perm_indices.
17187 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17188 instead of changing individual elements.
17189 (arm_vectorize_vec_perm_const): Use new_vector to install
17190 the vector in d.perm.
17191 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17192 Update vec_perm_builder constructor.
17193 (rs6000_expand_interleave): Likewise.
17194 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17195 (rs6000_expand_interleave): Likewise.
17197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17199 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17200 to qimode could truncate the indices.
17201 * optabs.c (expand_vec_perm_var): Likewise.
17203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17205 * Makefile.in (OBJS): Add vec-perm-indices.o.
17206 * vec-perm-indices.h: New file.
17207 * vec-perm-indices.c: Likewise.
17208 * target.h (vec_perm_indices): Replace with a forward class
17210 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17211 * optabs.h: Include vec-perm-indices.h.
17212 (expand_vec_perm): Delete.
17213 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17214 (expand_vec_perm_const): Declare.
17215 * target.def (vec_perm_const_ok): Replace with...
17216 (vec_perm_const): ...this new hook.
17217 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17218 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17219 * doc/tm.texi: Regenerate.
17220 * optabs.def (vec_perm_const): Delete.
17221 * doc/md.texi (vec_perm_const): Likewise.
17222 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17223 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17224 expand_vec_perm for constant permutation vectors. Assert that
17225 the mode of variable permutation vectors is the integer equivalent
17226 of the mode that is being permuted.
17227 * optabs-query.h (selector_fits_mode_p): Declare.
17228 * optabs-query.c: Include vec-perm-indices.h.
17229 (selector_fits_mode_p): New function.
17230 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17231 is defined, instead of checking whether the vec_perm_const_optab
17232 exists. Use targetm.vectorize.vec_perm_const instead of
17233 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17234 fit in the vector mode before using a variable permute.
17235 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17236 vec_perm_indices instead of an rtx.
17237 (expand_vec_perm): Replace with...
17238 (expand_vec_perm_const): ...this new function. Take the selector
17239 as a vec_perm_indices rather than an rtx. Also take the mode of
17240 the selector. Update call to shift_amt_for_vec_perm_mask.
17241 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17242 Use vec_perm_indices::new_expanded_vector to expand the original
17243 selector into bytes. Check whether the indices fit in the vector
17244 mode before using a variable permute.
17245 (expand_vec_perm_var): Make global.
17246 (expand_mult_highpart): Use expand_vec_perm_const.
17247 * fold-const.c: Includes vec-perm-indices.h.
17248 * tree-ssa-forwprop.c: Likewise.
17249 * tree-vect-data-refs.c: Likewise.
17250 * tree-vect-generic.c: Likewise.
17251 * tree-vect-loop.c: Likewise.
17252 * tree-vect-slp.c: Likewise.
17253 * tree-vect-stmts.c: Likewise.
17254 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17256 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17257 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17258 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17259 (aarch64_vectorize_vec_perm_const): ...this new function.
17260 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17261 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17262 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17263 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17264 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17265 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17266 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17268 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17269 check for NEON modes.
17270 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17271 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17272 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17273 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17275 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17276 the old VEC_PERM_CONST conditions.
17277 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17278 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17279 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17280 (ia64_vectorize_vec_perm_const_ok): Merge into...
17281 (ia64_vectorize_vec_perm_const): ...this new function.
17282 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17283 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17284 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17285 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17286 * config/mips/mips.c (mips_expand_vec_perm_const)
17287 (mips_vectorize_vec_perm_const_ok): Merge into...
17288 (mips_vectorize_vec_perm_const): ...this new function.
17289 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17290 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17291 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17292 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17293 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17294 (rs6000_expand_vec_perm_const): Delete.
17295 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17297 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17298 (altivec_expand_vec_perm_const_le): Take each operand individually.
17299 Operate on constant selectors rather than rtxes.
17300 (altivec_expand_vec_perm_const): Likewise. Update call to
17301 altivec_expand_vec_perm_const_le.
17302 (rs6000_expand_vec_perm_const): Delete.
17303 (rs6000_vectorize_vec_perm_const_ok): Delete.
17304 (rs6000_vectorize_vec_perm_const): New function.
17305 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17306 an element count and rtx array.
17307 (rs6000_expand_extract_even): Update call accordingly.
17308 (rs6000_expand_interleave): Likewise.
17309 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17310 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17311 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17312 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17313 (rs6000_expand_vec_perm_const): Delete.
17314 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17315 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17316 (altivec_expand_vec_perm_const_le): Take each operand individually.
17317 Operate on constant selectors rather than rtxes.
17318 (altivec_expand_vec_perm_const): Likewise. Update call to
17319 altivec_expand_vec_perm_const_le.
17320 (rs6000_expand_vec_perm_const): Delete.
17321 (rs6000_vectorize_vec_perm_const_ok): Delete.
17322 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17323 reference to the SPE evmerge intructions.
17324 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17325 an element count and rtx array.
17326 (rs6000_expand_extract_even): Update call accordingly.
17327 (rs6000_expand_interleave): Likewise.
17328 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17329 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17331 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17333 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17335 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17336 vector mode and that that mode matches the mode of the data
17338 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17339 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17340 directly using expand_vec_perm_1 when forcing selectors into
17342 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17346 * optabs-query.h (can_vec_perm_p): Delete.
17347 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17348 * optabs-query.c (can_vec_perm_p): Split into...
17349 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17350 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17351 particular selector is valid.
17352 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17353 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17354 (vect_grouped_load_supported): Likewise.
17355 (vect_shift_permute_load_chain): Likewise.
17356 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17357 (vect_transform_slp_perm_load): Likewise.
17358 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17359 (vectorizable_bswap): Likewise.
17360 (vect_gen_perm_mask_checked): Likewise.
17361 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17362 implementations of variable permutation vectors into account
17363 when deciding which selector to use.
17364 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17365 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17366 with a false third argument.
17367 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17368 to test whether the constant selector is valid and can_vec_perm_var_p
17369 to test whether a variable selector is valid.
17371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17373 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17374 * optabs-query.c (can_vec_perm_p): Likewise.
17375 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17376 instead of vec_perm_indices.
17377 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17378 (vect_gen_perm_mask_checked): Likewise,
17379 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17380 (vect_gen_perm_mask_checked): Likewise,
17382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17384 * optabs-query.h (qimode_for_vec_perm): Declare.
17385 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17386 (qimode_for_vec_perm): ...this new function.
17387 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17389 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17391 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17392 does not have a conditional at the top.
17394 2018-01-02 Richard Biener <rguenther@suse.de>
17396 * ipa-inline.c (big_speedup_p): Fix expression.
17398 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17401 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17404 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17408 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17409 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17410 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17411 cond_taken_branch_cost 3->4.
17413 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17415 PR tree-optimization/83581
17416 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17417 TODO_cleanup_cfg if any changes have been made.
17419 PR middle-end/83608
17420 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17421 convert_modes if target mode has the right side, but different mode
17424 PR middle-end/83609
17425 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17426 last argument when extracting from CONCAT. If either from_real or
17427 from_imag is NULL, use expansion through memory. If result is not
17428 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17429 the parts directly to inner mode, if even that fails, use expansion
17432 PR middle-end/83623
17433 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17434 check for bswap in mode rather than HImode and use that in expand_unop
17437 Copyright (C) 2018 Free Software Foundation, Inc.
17439 Copying and distribution of this file, with or without modification,
17440 are permitted in any medium without royalty provided the copyright
17441 notice and this notice are preserved.