1 2018-06-29 David Malcolm <dmalcolm@redhat.com>
3 * dumpfile.c (dump_loc): Add indentation based on scope depth.
4 (dump_scope_depth): New variable.
5 (get_dump_scope_depth): New function.
6 (dump_begin_scope): New function.
7 (dump_end_scope): New function.
8 * dumpfile.h (get_dump_scope_depth): New declaration.
9 (dump_begin_scope): New declaration.
10 (dump_end_scope): New declaration.
11 (class auto_dump_scope): New class.
12 (AUTO_DUMP_SCOPE): New macro.
13 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
16 2018-06-29 Richard Biener <rguenther@suse.de>
18 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
19 compute_all_dependences succeeds.
20 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
21 exceed --param loop-max-datarefs-for-datadeps.
23 2018-06-29 Jakub Jelinek <jakub@redhat.com>
25 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
27 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
30 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
31 alternative 0 in preferred_for_speed attribute.
33 2018-06-28 Paul Koning <ni1d@arrl.net>
35 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
36 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
37 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
38 * config/pdp11/pdp11.md: Correct "length" attribute calculation
39 for shift insn patterns.
41 2018-06-28 David Malcolm <dmalcolm@redhat.com>
43 * cgraph.c (cgraph_node::get_body): Replace assignments to
44 "dump_file" with calls to set_dump_file.
45 * dumpfile.c (alt_dump_file): Make static, and group with...
46 (alt_flags): ...this definition.
47 (dumps_are_enabled): New variable.
48 (refresh_dumps_are_enabled): New function.
49 (set_dump_file): New function.
50 (set_alt_dump_file): New function.
51 (gcc::dump_manager::dump_start): Replace assignments to
52 "dump_file" and "alt_dump_file" with calls to set_dump_file and
54 (gcc::dump_manager::dump_finish): Likewise.
55 * dumpfile.h (alt_dump_file): Delete decl.
56 (dumps_are_enabled): New variable decl.
57 (set_dump_file): New function decl.
58 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
60 * tree-nested.c (lower_nested_functions): Replace assignments to
61 "dump_file" with calls to set_dump_file.
63 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
65 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
66 goto_locus of each outgoing edge of each basic block.
68 2018-06-28 Richard Biener <rguenther@suse.de>
70 * dwarf2out.c (decl_scope_table): Remove.
71 (push_decl_scope): Likewise.
72 (pop_decl_scope): Likewise.
73 (gen_type_die_for_member): Do not call push/pop_decl_scope.
74 (gen_struct_or_union_type_die): Likewise.
75 (gen_tagged_type_die): Likewise.
76 (dwarf2out_init): Do not initialize decl_scope_table.
77 (dwarf2out_c_finalize): Do not free it.
79 2018-06-28 Richard Biener <rguenther@suse.de>
81 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
82 deciding whether to not re-use a DIE.
84 2018-06-28 Richard Biener <rguenther@suse.de>
86 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
87 DW_AT_abstract_origin attribute.
89 2018-06-28 Martin Liska <mliska@suse.cz>
91 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
92 Use newly introduced constants.
93 * tree-switch-conversion.h (struct jump_table_cluster):
94 Define max_ratio_for_size and max_ratio_for_speed.
96 2018-06-28 Martin Liska <mliska@suse.cz>
98 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
99 Add new checking assert to catch invalid state.
100 (jump_table_cluster::can_be_handled): Handle single case
102 (jump_table_cluster::is_beneficial): Bail out for such case.
103 (bit_test_cluster::find_bit_tests):
104 Add new checking assert to catch invalid state.
105 (bit_test_cluster::can_be_handled): Handle single case
107 (bit_test_cluster::is_beneficial): Bail out for such case.
108 (switch_decision_tree::analyze_switch_statement):
111 2018-06-28 Martin Liska <mliska@suse.cz>
113 * common.opt: Introduce -completion option.
114 * gcc.c (driver_handle_option): Handle it.
115 (driver::main): Print completions if completion
117 * opt-suggestions.c (option_proposer::get_completions):
119 (option_proposer::suggest_completion): Likewise.
120 (option_proposer::find_param_completions): Likewise.
121 (verify_autocompletions): Likewise.
122 (test_completion_valid_options): Likewise.
123 (test_completion_valid_params): Likewise.
124 (in_completion_p): Likewise.
125 (empty_completion_p): Likewise.
126 (test_completion_partial_match): Likewise.
127 (test_completion_garbage): Likewise.
128 (opt_proposer_c_tests): Likewise.
129 * opt-suggestions.h: Declare new functions.
130 * opts.c (common_handle_option): Handle OPT__completion_.
131 * selftest-run-tests.c (selftest::run_tests): Add
132 opt_proposer_c_tests.
133 * selftest.c (assert_str_startswith): New.
134 * selftest.h (assert_str_startswith): Likewise.
135 (opt_proposer_c_tests): New.
136 (ASSERT_STR_STARTSWITH): Likewise.
138 2018-06-28 Martin Liska <mliska@suse.cz>
140 * Makefile.in: Add opt-suggestions.o.
141 * gcc-main.c: Include opt-suggestions.h.
142 * gcc.c (driver::driver): Likewise.
143 (driver::~driver): Remove m_option_suggestions.
144 (driver::build_option_suggestions): Moved to option_proposer.
145 (driver::suggest_option): Likewise.
146 (driver::handle_unrecognized_options): Use option_proposer.
147 * gcc.h (class driver): Add new memver m_option_proposer.
148 * opt-suggestions.c: New file.
149 * opt-suggestions.h: New file.
151 2018-06-28 Martin Liska <mliska@suse.cz>
153 * vec.h (class auto_string_vec): New (moved from auto_argvec).
154 (auto_string_vec::~auto_string_vec): Likewise.
156 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
158 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
159 statement if id->reset_location is true.
160 (copy_edges_for_bb): Do not set goto_locus on the new edges if
161 id->reset_location is true.
162 (copy_phis_for_bb): Force input_location on the arguments if
163 id->reset_location is true.
164 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
165 is set on the function to be inlined.
166 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
167 prevent_decl_creation_for_types fields up and add reset_location field.
169 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
171 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
173 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
175 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
177 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
178 hard registers for the clobbered pseudo.
180 2018-06-27 Paul Koning <ni1d@arrl.net>
182 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
183 mutually exclusive options.
184 * config/pdp11/constraints.md (h): New constraint.
185 (O): Update definition to match shift code generation.
187 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
189 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
191 (output_jump): Change arguments.
192 (pdp11_fixed_cc_regs): New function.
193 (pdp11_cc_mode): Ditto.
194 (pdp11_expand_shift): Ditto.
195 (pdp11_assemble_shift): Ditto.
196 (pdp11_small_shift): Ditto.
197 (pdp11_branch_cost): Remove.
198 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
200 (pdp11_register_move_cost): Update for CC registers.
201 (pdp11_rtx_costs): Add case for LSHIFTRT.
202 (pdp11_output_jump): Add CCNZ mode conditional branches.
203 (notice_update_cc_on_set): Remove.
204 (pdp11_cc_mode): New function.
205 (simple_memory_operand): Correct pre/post decrement case.
206 (no_side_effect_operand): New function.
207 (pdp11_regno_reg_class): Add CC_REGS class.
208 (pdp11_fixed_cc_regs): New function.
209 (pdp11_small_shift): New function.
210 (pdp11_expand_shift): New function to expand shift insns.
211 (pdp11_assemble_shift): New function to output shifts.
212 (pdp11_branch_cost): Remove.
213 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
214 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
216 (PTRDIFF_TYPE): Ditto.
217 (ADJUST_INSN_LENGTH): New macro.
218 (FIXED_REGISTERS): Add CC registers.
219 (CALL_USED_REGISTERS): Ditto.
221 (REG_CLASS_NAMES): Ditto.
222 (REG_CLASS_CONTENTS): Ditto.
223 (SELECT_CC_MODE): Use new function.
224 (TARGET_FLAGS_REGNUM): New macro.
225 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
226 (cc0_reg_rtx): Remove.
227 (CC_STATUS_MDEP): Remove.
228 (CC_STATUS_MDEFP_INIT): Remove.
230 (NOTICE_UPDATE_CC): Remove.
231 (REGISTER_NAMES): Add CC registers.
232 (BRANCH_COST): Change to constant 1.
233 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
235 * config/pdp11/pdp11.opt (mbcopy): Remove.
236 (mbcopy-builtin): Remove.
237 (mbranch-cheap): Remove.
238 (mbranch-expensive): Remove.
239 * config/pdp11/predicates.md (expand_shift_operand): Update to
240 match shift code generation.
241 (ccnz_operator): New predicate.
242 * doc/invoke.texi (PDP-11 Options): Remove deleted options
243 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
244 Remove non-existent option -mabshi, -mno-abshi. Document mutually
246 * doc/md.texi (PDP-11): Document new D and h constraints. Update
247 description of O constraint.
249 2018-06-27 Jeff Law <law@redhat.com>
250 Austin Law <austinklaw@gmail.com>
252 * config/v850/v850.md (addsi3_set_flags): New pattern.
253 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
254 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
255 (zero_extendhisi2_v850_set_flags): Likewise.
256 (zero_extendqisi2_v850_set_flags): Likewise.
257 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
258 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
259 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
261 * config/v850/v850-protos.h (notice_update_cc): Remove.
262 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
263 (v850_print_operand): Handle 'D' and "d".
264 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
265 Add handling of arithmetic/logical operations compared against zero.
266 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
267 Do not look at v850_compare_op, instead get mode from last argument.
268 (v850_gen_compare): Remove
269 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
270 after reload for prologue insns.
271 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
273 (construct_save_jarl): Likewise.
274 (TARGET_FLAGS_REGNUM): Define.
275 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
276 (NOTICE_UPDATE_CC): Remove.
277 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
278 than cc0. Conditionalize on reload_completed.
279 (cmpsi_insn, setfcc_insn): Likewise.
280 (tst1 splitter): Turn into define_and_split which sets the flags
282 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
283 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
284 (cstoresf4, cstoredf4): Clobber the flags.
285 (cmpsi, cmpsf, cmpdf): Remove expanders.
286 (setf_insn): Remove pattern.
287 (addsi3): Turn into define_and_split which clobbers the flags after
288 reload and a suitable pattern (addsi3_clobber_flags) for use after
290 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
291 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
292 (ashrsi3, ashrsi3_v850e2): Likewise.
293 (bins): Clobber the flags.
294 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
295 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
296 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
297 (call_value_internal_short, call_value_internal_long): Likewise.
298 (callt_save_interrupt, callt_return_interrupt): Likewise.
299 (save_interrupt, return_interrupt): Likewise.
300 (callt_save_all_interrupt, save_all_interrupt): Likewise.
301 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
302 (restore_all_interrupt, _restore_all_interrupt): Likewise.
303 (All FP comparisons): Only allow after reload has completed.
305 (divh, divhu): Tweak output template.
306 (branch_z_normal, branch_z_invert): Remove
307 (branch_nz_normal, branch_nz_invert): Likewise.
308 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
310 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
311 * config/v850/v850.c (notice_update_cc): Remove.
312 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
313 (CC_NO_CARRY): Likewise.
314 (NOTICE_UPDATE_CC): Define to nothing.
315 * config/v850/v850.md: Remove block comment on cc0 handling
316 Remove "cc" attribute from all patterns. Remove cc_status handling
317 from all patterns. Minor formatting fixes.
319 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
321 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
322 (cortex-a76.cortex-a55): Likewise.
323 * config/aarch64/aarch64-tune.md: Regenerate.
324 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
325 cortex-a76.cortex-a55.
327 2018-06-27 Jeff Law <law@redhat.com>
329 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
330 (MULTILIB_DIRNAMES): Similarly.
332 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
334 * gimple.h (gimple_return_retbnd): Delete.
335 (gimple_return_set_retbnd): Likewise.
336 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
337 gimple_return_set_retbnd.
338 * gimple-pretty-print.c (dump_gimple_return): Remove call to
339 gimple_return_retbnd and adjust.
340 * tree-inline.h (struct copy_body_data): Remove retbnd field.
341 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
342 Explicitly return NULL in a couple more cases. Move assertion
343 on debug statements and remove unreachable code.
344 (reset_debug_binding): Do not test id->retbnd.
345 (expand_call_inline): Do not set it.
347 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
349 * configure.ac: Add --disable-gcov option.
350 * configure: Regenerate.
351 * Makefile.in: Honour @enable_gcov@.
352 * doc/install.texi: Document --disable-gcov.
354 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
356 * config/arm/arm-cpus.in (cortex-a76): New entry.
357 (cortex-a76.cortex-a55): Likewise.
358 * config/arm/arm-tables.opt: Regenerate.
359 * config/arm/arm-tune.md: Likewise.
360 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
361 * doc/invoke.texi (ARM Options): Document cortex-a76 and
362 cortex-a76.cortex-a55.
364 2018-06-27 Tamar Christina <tamar.christina@arm.com>
367 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
369 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
371 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
373 (EPILOGUE_USES): Likewise.
375 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
377 * tree-inline.c (remap_location): New function extracted from...
378 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
379 (copy_phis_for_bb): ...here. Call remap_location.
380 (copy_cfg_body): Adjust call to copy_edges_for_bb.
382 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
384 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
385 unaligned vsx for 16B memset.
387 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
390 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
391 ieee128_float_type_node to long_double_type_node unless
392 TARGET_LONG_DOUBLE_128 is set.
394 2018-06-26 David Malcolm <dmalcolm@redhat.com>
396 * cfgloop.c (get_loop_location): Convert return type from
397 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
398 by implicit construction from rtx_insn *, and using
399 dump_user_location_t::from_function_decl for the fallback case.
400 * cfgloop.h (get_loop_location): Convert return type from
401 location_t to dump_user_location_t.
402 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
403 dump_printf_loc to pass in a dump_location_t rather than a
404 location_t, via the gimple stmt.
405 * coverage.c (get_coverage_counts): Update calls to
406 dump_printf_loc to pass in dump_location_t rather than a
408 * doc/optinfo.texi (Dump types): Convert example of
409 dump_printf_loc from taking "locus" to taking "insn". Update
410 description of the "_loc" calls to cover dump_location_t.
411 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
413 (dump_user_location_t::dump_user_location_t): New constructors,
414 from gimple *stmt and rtx_insn *.
415 (dump_user_location_t::from_function_decl): New function.
416 (dump_loc): Make static.
417 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
418 const dump_location_t &.
419 (dump_generic_expr_loc): Delete.
420 (dump_printf_loc): Convert param "loc" from location_t to
421 const dump_location_t &.
422 (selftest::test_impl_location): New function.
423 (selftest::dumpfile_c_tests): New function.
424 * dumpfile.h: Include "profile-count.h".
425 (class dump_user_location_t): New class.
426 (struct dump_impl_location_t): New struct.
427 (class dump_location_t): New class.
428 (dump_printf_loc): Convert 2nd param from source_location to
429 const dump_location_t &.
430 (dump_generic_expr_loc): Delete.
431 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
432 const dump_location_t &.
433 * gimple-fold.c (fold_gimple_assign): Update call to
434 dump_printf_loc to pass in a dump_location_t rather than a
435 location_t, via the gimple stmt.
436 (gimple_fold_call): Likewise.
437 * gimple-loop-interchange.cc
438 (loop_cand::analyze_iloop_reduction_var): Update for change to
439 check_reduction_path.
440 (tree_loop_interchange::interchange): Update for change to
442 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
443 change in return-type of find_loop_location.
444 (graphite_regenerate_ast_isl): Likewise.
445 * graphite-optimize-isl.c (optimize_isl): Likewise.
446 * graphite.c (graphite_transform_loops): Likewise.
447 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
448 pass in a dump_location_t rather than a location_t, via the
450 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
451 * ipa.c (walk_polymorphic_call_targets): Likewise.
452 * loop-unroll.c (report_unroll): Convert "locus" param from
453 location_t to dump_location_t.
454 (decide_unrolling): Update for change to get_loop_location's
456 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
457 location_t to dump_user_location_t.
458 (grid_find_single_omp_among_assignments_1): Updates calls to
459 dump_printf_loc to pass in a dump_location_t rather than a
460 location_t, via the gimple stmt.
461 (grid_parallel_clauses_gridifiable): Convert "tloc" from
462 location_t to dump_location_t. Updates calls to dump_printf_loc
463 to pass in a dump_location_t rather than a location_t, via the
465 (grid_inner_loop_gridifiable_p): Likewise.
466 (grid_dist_follows_simple_pattern): Likewise.
467 (grid_gfor_follows_tiling_pattern): Likewise.
468 (grid_target_follows_gridifiable_pattern): Likewise.
469 (grid_attempt_target_gridification): Convert initialization
470 of local "grid" from memset to zero-initialization; FIXME: does
471 this require C++11? Update call to dump_printf_loc to pass in a
472 optinfo_location rather than a location_t, via the gimple stmt.
473 * profile.c (read_profile_edge_counts): Updates call to
474 dump_printf_loc to pass in a dump_location_t rather than a
476 (compute_branch_probabilities): Likewise.
477 * selftest-run-tests.c (selftest::run_tests): Call
479 * selftest.h (dumpfile_c_tests): New decl.
480 * tree-loop-distribution.c (pass_loop_distribution::execute):
481 Update for change in return type of find_loop_location.
482 * tree-parloops.c (parallelize_loops): Likewise.
483 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
484 "locus" from location_t to dump_user_location_t.
485 (canonicalize_loop_induction_variables): Likewise.
486 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
487 for change in return type of find_loop_location.
488 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
489 to dump_printf_loc to pass in a dump_location_t rather than a
490 location_t, via the stmt.
491 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
493 * tree-vect-loop-manip.c (find_loop_location): Convert return
494 type from source_location to dump_user_location_t.
495 (vect_do_peeling): Update for above change.
496 (vect_loop_versioning): Update for change in type of
498 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
499 from location_t to dump_user_location_t.
500 (vect_estimate_min_profitable_iters): Update for change in type
502 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
503 location_t to dump_location_t.
504 (vect_slp_bb): Update for change in type of vect_location.
505 * tree-vectorizer.c (vect_location): Convert from source_location
506 to dump_user_location_t.
507 (try_vectorize_loop_1): Update for change in vect_location's type.
508 (vectorize_loops): Likewise.
509 (increase_alignment): Likewise.
510 * tree-vectorizer.h (vect_location): Convert from source_location
511 to dump_user_location_t.
512 (find_loop_location): Convert return type from source_location to
513 dump_user_location_t.
514 (check_reduction_path): Convert 1st param from location_t to
515 dump_user_location_t.
516 * value-prof.c (check_counter): Update call to dump_printf_loc to
517 pass in a dump_user_location_t rather than a location_t; update
518 call to error_at for change in type of "locus".
519 (check_ic_target): Update call to dump_printf_loc to
520 pass in a dump_user_location_t rather than a location_t, via the
523 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
525 * config/s390/s390.h (enum processor_flags): Do not use
526 default tune parameter when -march was specified.
528 2018-06-26 Jakub Jelinek <jakub@redhat.com>
531 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
532 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
535 2018-06-26 Richard Biener <rguenther@suse.de>
537 PR tree-optimization/86287
539 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
540 (vect_analyze_loop): Initialize n_stmts.
542 2018-06-26 Richard Biener <rguenther@suse.de>
545 * fold-const.c (fold_convertible_p): Pointer extension
548 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
551 * dwarf2out.c (loc_list_has_views): Adjust comments.
552 (dw_loc_list): Split single cross-partition range with
555 2018-06-25 Jeff Law <law@redhat.com>
557 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
558 on -mbig-switch by default.
560 * config/v850/predicates.md (const_float_1_operand): Fix match_code
562 (const_float_0_operand): Remove unused predicate.
563 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
564 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
565 (recipsf2): New expander. Original pattern now called
567 (recipdf2, recipdf2_insn): Similarly.
568 (rsqrtsf2, rsqrtsf2_insn): Similarly
569 (rsqrtdf2, rsqrtdf2_insn): Similarly
571 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
573 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
574 Simplify logic for FreeBSD (twice).
576 2018-06-25 Martin Sebor <msebor@redhat.com>
578 PR tree-optimization/86204
579 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
580 a strnlen result if it's less than the length of the string.
582 2018-06-25 Martin Sebor <msebor@redhat.com>
584 PR tree-optimization/85700
585 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
586 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
587 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
589 2018-06-25 Martin Sebor <msebor@redhat.com>
591 * doc/extend.texi (Zero-length arrays): Update and clarify.
593 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
595 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
596 added IEEE/IBM long double multilib support on PowerPC little
597 endian Linux systems.
598 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
599 (MULTILIB_DEFAULTS): Likewise.
600 * config/rs6000/rs6000.c (rs6000_option_override_internal):
602 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
603 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
604 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
606 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
609 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
610 (REORDER_45): Likewise.
612 2018-06-25 Jeff Law <law@redhat.com>
614 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
615 dividend to 32 bits. Adjust length.
616 (udivmodhi4): Cleanup output template. Fix length.
618 2018-06-25 Carl Love <cel@us.ibm.com>
620 * config/rs6000/vsx.md: Change word selector to prefered location.
622 2018-06-25 Richard Biener <rguenther@suse.de>
624 PR tree-optimization/86304
625 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
626 epilogue-if-converted loops as well.
628 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
630 * lto-section-out.c (lto_begin_section): Do not print section
631 name for noaddr and unnumbered dumps.
633 2018-06-25 Richard Biener <rguenther@suse.de>
635 * tree-vectorizer.h (struct vec_info_shared): New structure
636 with parts split out from struct vec_info and loop_nest from
637 struct _loop_vec_info.
638 (struct vec_info): Adjust accordingly.
639 (struct _loop_vec_info): Likewise.
640 (LOOP_VINFO_LOOP_NEST): Adjust.
641 (LOOP_VINFO_DATAREFS): Likewise.
642 (LOOP_VINFO_DDRS): Likewise.
643 (struct _bb_vec_info): Likewise.
644 (BB_VINFO_DATAREFS): Likewise.
645 (BB_VINFO_DDRS): Likewise.
646 (struct _stmt_vec_info): Add dr_aux member.
647 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
648 (DR_MISALIGNMENT_UNINITIALIZED): New.
649 (set_dr_misalignment): Adjust.
650 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
651 (vect_analyze_loop): Adjust prototype.
652 (vect_analyze_loop_form): Likewise.
653 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
654 Compute dependences lazily.
655 (vect_record_base_alignments): Use shared datarefs/ddrs.
656 (vect_verify_datarefs_alignment): Likewise.
657 (vect_analyze_data_refs_alignment): Likewise.
658 (vect_analyze_data_ref_accesses): Likewise.
659 (vect_analyze_data_refs): Likewise.
660 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
661 constructor parameter for shared part.
662 (vect_analyze_loop_form): Pass in shared part and adjust.
663 (vect_analyze_loop_2): Pass in storage for the number of
664 stmts. Move loop nest finding to the caller. Compute
666 (vect_analyze_loop): Pass in shared part.
667 (vect_transform_loop): Verify shared datarefs are unchanged.
668 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
669 constructor parameter for shared part.
670 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
671 (vect_slp_bb): Verify shared datarefs are unchanged before
673 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
675 (new_stmt_vec_info): Initialize DR_AUX misalignment to
676 DR_MISALIGNMENT_UNINITIALIZED.
677 * tree-vectorizer.c (vec_info::vec_info): Add constructor
678 parameter for shared part.
679 (vec_info::~vec_info): Adjust.
680 (vec_info_shared::vec_info_shared): New.
681 (vec_info_shared::~vec_info_shared): Likewise.
682 (vec_info_shared::save_datarefs): Likewise.
683 (vec_info_shared::check_datarefs): Likewise.
684 (try_vectorize_loop_1): Construct shared part live for analyses
685 of a single loop for multiple vector sizes.
686 * tree-parloops.c (gather_scalar_reductions): Adjust.
688 2018-06-25 Richard Biener <rguenther@suse.de>
690 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
691 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
692 (vect_analyze_data_refs): Remove similar code from here and
693 simplify accordingly.
695 2018-06-25 Richard Biener <rguenther@suse.de>
697 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
698 for reverse storage order accesses rather than asserting
699 they cannot happen here.
701 2018-06-25 Tom de Vries <tdevries@suse.de>
704 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
705 Use data16 instead of .byte for insn prefix.
707 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
710 * parser.c (make_char_string_pack): Pass this literal chars
711 through cpp_interpret_string.
712 (cp_parser_userdef_numeric_literal): Check the result of
713 make_char_string_pack.
715 2018-06-24 Maya Rashish <coypu@sdf.org>
717 * ginclude/stddef.h: Simplify conditions around avoiding
718 re-definition of __size_t.
720 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
722 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
723 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
725 2018-06-22 Maya Rashish <coypu@sdf.org>
727 * doc/invoke.texi (mno-fancy-math-387): Update for changes
728 made to OpenBSD and NetBSD through the years.
730 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
732 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
733 behavior of vec_pack (vector double, vector double) to match
734 behavior of vec_float2 (vector double, vector double).
736 2018-06-22 Olivier Hainque <hainque@adacore.com>
738 * gimplify.c (gimplify_function_tree): Prevent creation
739 of a trampoline for the address of the current function
740 passed to entry/exit instrumentation hooks.
742 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
745 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
748 2018-06-22 Martin Liska <mliska@suse.cz>
750 PR tree-optimization/86263
751 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
752 Bail out if is_enabled is false.
753 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
755 (jump_table_cluster::is_enabled): New function.
757 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
759 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
760 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
761 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
762 (lto_input_ts_binfo_tree_pointers): Likewise.
763 * tree-streamer-out.c (streamer_write_tree_bitfields,
764 write_ts_binfo_tree_pointers): Likewise.
765 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
767 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
769 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
771 2018-06-22 Martin Liska <mliska@suse.cz>
773 * symbol-summary.h (get): Make it pure and inline move
774 functionality from ::get function.
775 (get): Remove and inline into ::get and ::get_create.
776 (get_create): Move code from ::get function.
778 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
781 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
782 -x assembler-with-cpp.
784 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
786 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
787 _FILE_OFFSET_BITS=64 for C++.
789 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
791 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
792 conversion insn that shows up when pr85657-3.c is compiled using
793 IEEE 128-bit long double.
794 (neg<mode>2_internal): Use the correct mode to check whether the
795 mode is IBM extended.
796 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
797 multiply and divide external functions from being created more
800 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
802 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
804 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
805 the edge can be forwarded.
806 (cfg_layout_merge_blocks): Likewise.
808 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
810 * except.c (finish_eh_generation): Commit edge insertions only after
811 the EH edges have been redirected from post-landing to landing pads.
813 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
815 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
816 create_tmp_var_for to create the FRAME decl.
817 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
819 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
821 * tree-inline.c (copy_edges_for_bb): Minor tweak.
822 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
823 debug statement when resetting its value.
824 (expand_call_inline): Copy the locus of the call onto the assignment
825 of the return value, if any. Use local variable in more cases.
827 2018-06-21 Martin Liska <mliska@suse.cz>
829 * ipa-pure-const.c (propagate_nothrow): Use
830 funct_state_summaries->get.
831 (dump_malloc_lattice): Likewise.
832 (propagate_malloc): Likewise.
834 2018-06-21 Richard Biener <rguenther@suse.de>
836 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
837 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
838 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
840 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
841 BLOCK_ABSTRACT_ORIGIN unconditionally.
843 2018-06-21 David Malcolm <dmalcolm@redhat.com>
845 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
847 * ipa-reference.c (ipa_reference_c_finalize): Delete
848 ipa_ref_opt_sum_summaries and set it to NULL.
850 2018-06-21 Tom de Vries <tdevries@suse.de>
852 PR tree-optimization/85859
853 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
854 test with comment from bb_no_side_effects_p.
856 2018-06-21 Richard Biener <rguenther@suse.de>
858 PR tree-optimization/86232
859 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
860 max for constant niter.
862 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
864 * config/aarch64/aarch64-simd.md
865 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
867 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
869 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
870 Make opernads of the unspec commutative.
872 2018-06-21 Richard Biener <rguenther@suse.de>
874 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
875 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
876 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
877 (vect_analyze_data_ref_dependence): Re-order checks to deal with
879 (vect_record_base_alignments): Do not record base alignment
880 for gathers or scatters.
881 (vect_compute_data_ref_alignment): Drop return value that is always
882 true. Bail out early for gathers or scatters.
883 (vect_enhance_data_refs_alignment): Bail out early for gathers
885 (vect_find_same_alignment_drs): Likewise.
886 (vect_analyze_data_refs_alignment): Remove dead code.
887 (vect_slp_analyze_and_verify_node_alignment): Likewise.
888 (vect_analyze_data_refs): For possible gathers or scatters do
889 not create an alternate DR, just check their possible validity
890 and mark them. Adjust DECL_NONALIASED handling to not rely
892 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
893 update inits of gathers or scatters.
894 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
895 Also copy gather/scatter flag to pattern vinfo.
897 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
899 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
900 behavior of vec_packsu (vector unsigned long long, vector unsigned
901 long long) to match behavior of vec_packs with same signature.
903 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
904 Thomas Schwinge <thomas@codesourcery.com>
905 Cesar Philippidis <cesar@codesourcery.com>
907 * gimplify.c (gimplify_scan_omp_clauses): Add support for
908 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
909 (gimplify_adjust_omp_clauses): Likewise.
910 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
911 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
912 (gimplify_omp_target_update): Update handling of acc update and
914 * omp-low.c (install_var_field): Remove unused parameter
915 base_pointers_restrict.
916 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
917 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
919 (omp_target_base_pointers_restrict_p): Delete.
920 (scan_omp_target): Update call to scan_sharing_clauses.
921 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
923 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
924 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
925 (convert_local_omp_clauses): Likewise.
926 * tree-pretty-print.c (dump_omp_clause): Likewise.
927 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
929 (omp_clause_code_name): Likewise.
931 2018-06-20 Jakub Jelinek <jakub@redhat.com>
934 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
937 PR tree-optimization/86231
938 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
939 anti-range don't overwrite *vr0min before using it to compute *vr0max.
941 2018-06-20 Tom de Vries <tdevries@suse.de>
943 PR tree-optimization/86097
944 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
945 iv type if signedness of iv type is not the same as that of *nit.
947 2018-06-20 Jakub Jelinek <jakub@redhat.com>
949 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
950 EDGE_EH edges, verify they are all EDGE_EH.
952 2018-06-20 Maya Rashish <coypu@sdf.org>
954 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
956 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
958 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
959 * config/aarch64/aarch64.c (xgene1_tunings): Add
960 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
961 (aarch64_mode_valid_for_sched_fusion_p):
963 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
964 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
966 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
968 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
969 * config/aarch64/iterators.md (VQ2): New mode iterator.
971 2018-06-20 Martin Liska <mliska@suse.cz>
973 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
974 Change default ratio from 10 to 8.
976 2018-06-20 Martin Liska <mliska@suse.cz>
978 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
980 (bit_test_cluster::find_bit_tests): Likewise.
981 (switch_decision_tree::analyze_switch_statement): Find clusters.
982 * tree-switch-conversion.h (struct jump_table_cluster): Document
985 2018-06-20 Martin Liska <mliska@suse.cz>
987 * tree-switch-conversion.c (switch_conversion::collect):
988 Record m_uniq property.
989 (switch_conversion::expand): Bail out for special conditions.
990 (group_cluster::~group_cluster): New.
991 (group_cluster::group_cluster): Likewise.
992 (group_cluster::dump): Likewise.
993 (jump_table_cluster::emit): New.
994 (switch_decision_tree::fix_phi_operands_for_edges): New.
995 (struct case_node): Remove struct.
996 (jump_table_cluster::can_be_handled): New.
997 (case_values_threshold): Moved to header.
998 (reset_out_edges_aux): Likewise.
999 (jump_table_cluster::is_beneficial): New.
1000 (bit_test_cluster::can_be_handled): Likewise.
1001 (add_case_node): Remove.
1002 (bit_test_cluster::is_beneficial): New.
1003 (case_bit_test::cmp): New.
1004 (bit_test_cluster::emit): New.
1005 (expand_switch_as_decision_tree_p): Remove.
1006 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1007 (fix_phi_operands_for_edge): Likewise.
1008 (switch_decision_tree::analyze_switch_statement): New.
1009 (compute_cases_per_edge): Move ...
1010 (switch_decision_tree::compute_cases_per_edge): ... here.
1011 (try_switch_expansion): Likewise.
1012 (switch_decision_tree::try_switch_expansion): Likewise.
1013 (record_phi_operand_mapping): Likewise.
1014 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1015 (emit_case_decision_tree): Likewise.
1016 (switch_decision_tree::emit): Likewise.
1017 (balance_case_nodes): Likewise.
1018 (switch_decision_tree::balance_case_nodes): Likewise.
1019 (dump_case_nodes): Likewise.
1020 (switch_decision_tree::dump_case_nodes): Likewise.
1021 (emit_jump): Likewise.
1022 (switch_decision_tree::emit_jump): Likewise.
1023 (emit_cmp_and_jump_insns): Likewise.
1024 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1025 (emit_case_nodes): Likewise.
1026 (switch_decision_tree::emit_case_nodes): Likewise.
1027 (conditional_probability): Remove.
1028 * tree-switch-conversion.h (enum cluster_type): New.
1030 (struct cluster): Likewise.
1031 (cluster::cluster): Likewise.
1032 (struct simple_cluster): Likewise.
1033 (simple_cluster::simple_cluster): Likewise.
1034 (struct group_cluster): Likewise.
1035 (struct jump_table_cluster): Likewise.
1036 (struct bit_test_cluster): Likewise.
1037 (struct min_cluster_item): Likewise.
1038 (struct case_tree_node): Likewise.
1039 (case_tree_node::case_tree_node): Likewise.
1040 (jump_table_cluster::case_values_threshold): Likewise.
1041 (struct case_bit_test): Likewise.
1042 (struct switch_decision_tree): Likewise.
1043 (struct switch_conversion): Likewise.
1044 (switch_decision_tree::reset_out_edges_aux): Likewise.
1046 2018-06-20 Martin Liska <mliska@suse.cz>
1048 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1049 (hoist_edge_and_branch_if_true): Likewise.
1050 (expand_switch_using_bit_tests_p): Likewise.
1051 (struct case_bit_test): Likewise.
1052 (case_bit_test_cmp): Likewise.
1053 (emit_case_bit_tests): Likewise.
1054 (switch_conversion::switch_conversion): New class.
1055 (struct switch_conv_info): Remove old struct.
1056 (collect_switch_conv_info): More to ...
1057 (switch_conversion::collect): ... this.
1058 (check_range): Likewise.
1059 (switch_conversion::check_range): Likewise.
1060 (check_all_empty_except_final): Likewise.
1061 (switch_conversion::check_all_empty_except_final): Likewise.
1062 (check_final_bb): Likewise.
1063 (switch_conversion::check_final_bb): Likewise.
1064 (create_temp_arrays): Likewise.
1065 (switch_conversion::create_temp_arrays): Likewise.
1066 (free_temp_arrays): Likewise.
1067 (gather_default_values): Likewise.
1068 (switch_conversion::gather_default_values): Likewise.
1069 (build_constructors): Likewise.
1070 (switch_conversion::build_constructors): Likewise.
1071 (constructor_contains_same_values_p): Likewise.
1072 (switch_conversion::contains_same_values_p): Likewise.
1073 (array_value_type): Likewise.
1074 (switch_conversion::array_value_type): Likewise.
1075 (build_one_array): Likewise.
1076 (switch_conversion::build_one_array): Likewise.
1077 (build_arrays): Likewise.
1078 (switch_conversion::build_arrays): Likewise.
1079 (gen_def_assigns): Likewise.
1080 (switch_conversion::gen_def_assigns): Likewise.
1081 (prune_bbs): Likewise.
1082 (switch_conversion::prune_bbs): Likewise.
1083 (fix_phi_nodes): Likewise.
1084 (switch_conversion::fix_phi_nodes): Likewise.
1085 (gen_inbound_check): Likewise.
1086 (switch_conversion::gen_inbound_check): Likewise.
1087 (process_switch): Use the newly created class.
1088 (switch_conversion::expand): New.
1089 (switch_conversion::~switch_conversion): New.
1090 * tree-switch-conversion.h: New file.
1092 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1094 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1095 tree-vect-patterns.c.
1096 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1097 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1098 (vect_recog_sad_pattern): Likewise.
1099 (vect_recog_widen_sum_pattern): Likewise.
1100 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1101 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1102 (vect_recog_rotate_pattern): Likewise.
1103 (vect_recog_mult_pattern): Likewise.
1104 (vect_recog_vector_vector_shift_pattern): Likewise.
1105 (vect_recog_divmod_pattern): Likewise.
1106 (vect_recog_mixed_size_cond_pattern): Likewise.
1107 (vect_recog_bool_pattern): Likewise.
1108 (vect_recog_mask_conversion_pattern): Likewise.
1109 (vect_try_gather_scatter_pattern): Likewise.
1110 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1111 (vect_recog_over_widening_pattern): Likewise.
1112 (vect_recog_gather_scatter_pattern): Likewise.
1113 (vect_recog_func_ptr): Move from tree-vectorizer.h
1114 (vect_vect_recog_func_ptrs): Move further down the file.
1115 (vect_recog_func): Likewise. Remove the third argument.
1116 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1117 (vect_pattern_recog_1): Expect the pattern function to do any
1118 necessary target tests. Also expect it to provide a vector type.
1119 Remove the type_in handling.
1121 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1123 * tree-vect-patterns.c (vect_pattern_detected): New function.
1124 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1125 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1126 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1127 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1128 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1129 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1130 (vect_recog_mask_conversion_pattern)
1131 (vect_try_gather_scatter_pattern): Likewise.
1133 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1135 * tree-vect-patterns.c (vect_get_internal_def): New function.
1136 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1137 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1138 (search_type_for_mask_1): Use it.
1140 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1142 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1143 redundant WIDEN_SUM_EXPR handling.
1144 (vect_recog_sad_pattern): Likewise.
1146 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1148 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1149 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1150 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1151 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1152 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1154 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1156 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1157 use the stmt_vec_info of the original bb statement for the
1158 new zero assignment, even if the call is part of a pattern.
1160 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1162 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1163 that the sequence is attached to the original statement rather
1164 than the pattern statement.
1165 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1166 PATTERN_DEF_SEQ from the original statement rather than
1167 the main pattern statement.
1168 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1169 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1170 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1172 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1174 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1175 definition statements before the early exit for statements that aren't
1177 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1179 (vect_transform_loop): ...here. Process pattern definition
1180 statements without first checking whether the main pattern
1181 statement is live or relevant.
1183 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1185 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1186 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1188 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1190 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1191 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1192 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1193 (expand_block_compare): Change select_block_compare_mode call.
1194 (expand_strncmp_align_check): Use new functions, fix comment.
1195 (emit_final_str_compare_gpr): New function.
1196 (expand_strn_compare): Refactor and clean up code.
1197 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1199 2018-06-19 Tony Reix <tony.reix@atos.com>
1200 Damien Bergamini <damien.bergamini@atos.com>
1201 David Edelsohn <dje.gcc@gmail.com>
1203 * collect2.c (static_obj): New variable.
1204 (static_libs): New variable.
1205 (is_in_list): Uncomment declaration.
1206 (main): Track AIX libraries linked statically.
1207 (is_in_list): Uncomment definition.
1208 (scan_prog_file): Don't add AIX shared libraries initializer
1209 to constructor list if linking statically.
1211 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1213 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1215 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1217 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1219 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1222 2018-06-19 Martin Liska <mliska@suse.cz>
1224 * config/i386/i386.c (ix86_can_inline_p): Do not use
1225 ipa_fn_summaries::get_create.
1226 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1228 (devirtualization_time_bonus): Likewise.
1229 (ipcp_propagate_stage): Likewise.
1230 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1231 (edge_set_predicate): Likewise.
1232 (evaluate_conditions_for_known_args): Likewise.
1233 (evaluate_properties_for_edge): Likewise.
1234 (ipa_call_summary::reset): Tranform to ...
1235 (ipa_call_summary::~ipa_call_summary): ... this.
1236 (ipa_fn_summary::reset): Transform to ...
1237 (ipa_fn_summary::~ipa_fn_summary): ... this.
1238 (ipa_fn_summary_t::remove): Rename to ...
1239 (ipa_fn_summary_t::remove_callees): ... this.
1240 (ipa_fn_summary_t::duplicate): Use placement new
1241 instead of memory copy.
1242 (ipa_call_summary_t::duplicate): Likewise.
1243 (ipa_call_summary_t::remove): Remove.
1244 (dump_ipa_call_summary): Change get_create to get.
1245 (ipa_dump_fn_summary): Dump only when summary exists.
1246 (analyze_function_body): Use symbol_summary::get instead
1248 (compute_fn_summary): Likewise.
1249 (estimate_edge_devirt_benefit): Likewise.
1250 (estimate_edge_size_and_time): Likewise.
1251 (inline_update_callee_summaries): Likewise.
1252 (remap_edge_change_prob): Likewise.
1253 (remap_edge_summaries): Likewise.
1254 (ipa_merge_fn_summary_after_inlining): Likewise.
1255 (write_ipa_call_summary): Likewise.
1256 (ipa_fn_summary_write): Likewise.
1257 (ipa_free_fn_summary): Likewise.
1258 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1259 (struct ipa_call_summary): Likewise.
1260 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1262 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1263 (estimate_size_after_inlining): Likewise.
1264 (estimate_growth): Likewise.
1265 (growth_likely_positive): Likewise.
1266 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1267 (inline_call): Likewise.
1268 * ipa-inline.c (caller_growth_limits): Likewise.
1269 (can_inline_edge_p): Likewise.
1270 (can_inline_edge_by_limits_p): Likewise.
1271 (compute_uninlined_call_time): Likewise.
1272 (compute_inlined_call_time): Likewise.
1273 (want_inline_small_function_p): Likewise.
1274 (edge_badness): Likewise.
1275 (update_caller_keys): Likewise.
1276 (update_callee_keys): Likewise.
1277 (inline_small_functions): Likewise.
1278 (inline_to_all_callers_1): Likewise.
1279 (dump_overall_stats): Likewise.
1280 (early_inline_small_functions): Likewise.
1281 (early_inliner): Likewise.
1282 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1283 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1284 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1285 * ipa-split.c (execute_split_functions): Likewise.
1286 * symbol-summary.h: Likewise.
1287 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1289 2018-06-19 Richard Biener <rguenther@suse.de>
1291 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1292 (vectorize_loops): ... here. Fix dbgcnt handling.
1293 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1295 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1298 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1299 ieee128 argument takes up only one (vector) register, not two (floating
1302 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1304 * gimplify.c (gimplify_init_constructor): Really never clear for an
1305 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1307 2018-06-19 Richard Biener <rguenther@suse.de>
1309 PR tree-optimization/86179
1310 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1311 after failed recognition.
1313 2018-06-18 Martin Sebor <msebor@redhat.com>
1316 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1317 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1318 Handle integer subtraction.
1319 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1320 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1322 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1324 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1325 param from rtx to rtx_insn *.
1326 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1328 (frv_ifcvt_modify_insn): Likwise.
1329 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1330 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1331 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1332 as_a <rtx_insn *> cast to local "unprotected_region" once
1333 it's been established that it's not NULL or pc_rtx.
1334 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1335 param "sethi" from rtx to rtx_insn *.
1336 (nds32_group_float_insns): Likewise for param "insn".
1337 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1339 (vax_output_int_subtract): Likewise.
1340 * config/vax/vax.c (vax_output_int_add): Likewise for param
1342 (vax_output_int_subtract): Likewise.
1343 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1344 (emit_pattern_after): Likewise for param "after".
1345 (emit_insn_after): Likewise.
1346 (emit_jump_insn_after): Likewise.
1347 (emit_call_insn_after): Likewise.
1348 (emit_debug_insn_after): Likewise.
1349 (emit_pattern_before): Likewise for param "before".
1350 (emit_insn_before): Likewise.
1351 (emit_jump_insn_before): Likewise.
1352 * final.c (get_insn_template): Likewise for param "insn", removing
1354 * output.h (get_insn_template): Likewise for 2nd param.
1355 * rtl.h (emit_insn_before): Likewise.
1356 (emit_jump_insn_before): Likewise.
1357 (emit_debug_insn_before_noloc): Likewise.
1358 (emit_insn_after): Likewise.
1359 (emit_jump_insn_after): Likewise.
1360 (emit_call_insn_after): Likewise.
1361 (emit_debug_insn_after): Likewise.
1362 (set_insn_deleted): Likewise for param.
1364 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1367 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1368 floating point modes, so that IFmode is numerically greater than
1369 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1370 to declare the ordering. This prevents IFmode from being
1371 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1372 machine. Include rs6000-modes.h to share the fractional values
1373 between genmodes* and the rest of the compiler.
1377 * config/rs6000/rs6000-modes.h: New file.
1378 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1379 meaning of rs6000_long_double_size so that 126..128 selects an
1380 appropriate 128-bit floating point type.
1381 (rs6000_option_override_internal): Likewise.
1382 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1383 (TARGET_LONG_DOUBLE_128): Change the meaning of
1384 rs6000_long_double_size so that 126..128 selects an appropriate
1385 128-bit floating point type.
1386 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1387 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1388 source and destination to match the standard usage.
1389 (truncifkf2): Likewise.
1390 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1391 ISA 2.07 to use an explicit clobber, instead of passing in a
1393 (copysign<mode>3_soft): Likewise.
1395 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1397 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1398 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1399 (vect_slp_analyze_instance_dependence): Likewise.
1400 (vect_enhance_data_refs_alignment): Likewise.
1401 (vect_analyze_data_refs_alignment): Likewise.
1402 (vect_slp_analyze_and_verify_instance_alignment
1403 (vect_analyze_data_ref_accesses): Likewise.
1404 (vect_prune_runtime_alias_test_list): Likewise.
1405 (vect_analyze_data_refs): Likewise.
1406 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1407 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1408 (vect_analyze_scalar_cycles_1): Likewise.
1409 (vect_get_loop_niters): Likewise.
1410 (vect_analyze_loop_form_1): Likewise.
1411 (vect_update_vf_for_slp): Likewise.
1412 (vect_analyze_loop_operations): Likewise.
1413 (vect_analyze_loop): Likewise.
1414 (vectorizable_induction): Likewise.
1415 (vect_transform_loop): Likewise.
1416 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1417 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1418 (vect_make_slp_decision): Likewise.
1419 (vect_detect_hybrid_slp): Likewise.
1420 (vect_slp_analyze_operations): Likewise.
1421 (vect_slp_bb): Likewise.
1422 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1423 (vectorizable_bswap): Likewise.
1424 (vectorizable_call): Likewise.
1425 (vectorizable_simd_clone_call): Likewise.
1426 (vectorizable_conversion): Likewise.
1427 (vectorizable_assignment): Likewise.
1428 (vectorizable_shift): Likewise.
1429 (vectorizable_operation): Likewise.
1430 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1432 2018-06-18 Martin Sebor <msebor@redhat.com>
1434 PR tree-optimization/81384
1435 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1436 * builtins.c (expand_builtin_strnlen): New function.
1437 (expand_builtin): Call it.
1438 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1439 * builtins.def (BUILT_IN_STRNLEN): New.
1440 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1441 Warn for bounds in excess of maximum object size.
1442 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1443 single-value ranges. Handle strnlen.
1444 (handle_builtin_strlen): Handle strnlen.
1445 (strlen_check_and_optimize_stmt): Same.
1446 * doc/extend.texi (Other Builtins): Document strnlen.
1448 2018-06-18 Maya Rashish <coypu@sdf.org>
1450 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1451 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1452 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1454 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1456 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1458 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1460 * tree.c (escaped_string::escape): Replace cast to char * by
1461 const_cast<char *> (unescaped).
1463 2018-06-18 Nick Clifton <nickc@redhat.com>
1466 * tree.c (escaped_string): New class. Converts an unescaped
1467 string into its escaped equivalent.
1468 (warn_deprecated_use): Use the new class to convert the
1469 deprecation message, if present.
1470 (test_escaped_strings): New self test.
1471 (test_c_tests): Add test_escaped_strings.
1472 * doc/extend.texi (deprecated): Add a note that the
1473 deprecation message is affected by the -fmessage-length
1474 option, and that control characters will be escaped.
1475 (#pragma GCC error): Document this pragma.
1476 (#pragma GCC warning): Likewise.
1477 * doc/invoke.texi (-fmessage-length): Document this option's
1478 effect on the #warning and #error preprocessor directives and
1479 the deprecated attribute.
1481 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1483 * tree.c (decl_value_expr_lookup): Revert latest change.
1484 (decl_value_expr_insert): Likewise.
1486 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1488 * gimplify.c (nonlocal_vlas): Delete.
1489 (nonlocal_vla_vars): Likewise.
1490 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1492 (gimplify_body): Do not create and destroy nonlocal_vlas.
1493 * tree-nested.c: Include diagnostic.h.
1494 (use_pointer_in_frame): Tweak.
1495 (lookup_field_for_decl): Add assertion and declare the transformation.
1496 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1497 internal error when the reference is in a wrong context. Do not
1498 create a debug decl by default.
1499 (note_nonlocal_block_vlas): Delete.
1500 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1501 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1502 create a debug decl by default.
1503 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1504 call is in a wrong context.
1505 (fixup_vla_decls): New function.
1506 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1507 debug variables were created.
1508 * tree.c (decl_value_expr_lookup): Add checking assertion.
1509 (decl_value_expr_insert): Likewise.
1511 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1514 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1515 * tree-scalar-evolution.c (interpret_expr): Likewise.
1516 (expression_expensive_p): Likewise.
1517 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1518 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1519 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1520 (ssa_defined_by_minus_one_stmt_p): New.
1522 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1525 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1526 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1527 * dojump.c (do_jump): Likewise.
1528 * expr.c (expand_expr_real_2): Check operand type's sign.
1529 * fold-const.c (const_unop): Handle ABSU_EXPR.
1530 (fold_abs_const): Likewise.
1531 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1532 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1533 (strip_sign_op_1): Likesise.
1534 * match.pd: Add new pattern to generate ABSU_EXPR.
1535 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1536 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1537 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1538 * tree-inline.c (estimate_operator_cost): Likewise.
1539 * tree-pretty-print.c (dump_generic_node): Likewise.
1540 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1541 * tree.def (ABSU_EXPR): New.
1543 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1546 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1547 documented as preserved for backward compatibility only.
1548 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1550 PR rtl-optimization/86108
1551 * bb-reorder.c (create_forwarder_block): Renamed to ...
1552 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1553 jump from new landing pad to the second part.
1554 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1557 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1560 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1561 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1562 Only call store_expr for halves if the mode is the same.
1565 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1566 Fix up comment formatting.
1568 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1570 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1571 (typed_splay_tree::closure,
1572 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1573 (typed_splay_tree::typed_splay_tree,
1574 typed_splay_tree::operator =): Declared private.
1575 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1576 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1577 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1578 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1579 typed_splay_tree::splay_tree_splay,
1580 typed_splay_tree::splay_tree_foreach_helper,
1581 typed_splay_tree::splay_tree_insert,
1582 typed_splay_tree::splay_tree_remove,
1583 typed_splay_tree::splay_tree_lookup,
1584 typed_splay_tree::splay_tree_predecessor,
1585 typed_splay_tree::splay_tree_successor,
1586 typed_splay_tree::splay_tree_min,
1587 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1588 (typed_splay_tree::root, typed_splay_tree::comp,
1589 typed_splay_tree::delete_key,
1590 typed_splay_tree::delete_value): New data members.
1591 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1592 typed_splay_tree::remove.
1594 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1596 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1597 -mginv and -mno-ginv to the assembler.
1598 * config/mips/mips.opt (-mcrc): New option.
1600 * doc/invoke.text (-mcrc): Document.
1603 2018-06-15 Nick Clifton <nickc@redhat.com>
1606 * tree.c (escaped_string): New class. Converts an unescaped
1607 string into its escaped equivalent.
1608 (warn_deprecated_use): Use the new class to convert the
1609 deprecation message, if present.
1610 (test_escaped_strings): New self test.
1611 (test_c_tests): Add test_escaped_strings.
1612 * doc/extend.texi (deprecated): Add a note that the
1613 deprecation message is affected by the -fmessage-length
1614 option, and that control characters will be escaped.
1615 (#pragma GCC error): Document this pragma.
1616 (#pragma GCC warning): Likewise.
1617 * doc/invoke.texi (-fmessage-length): Document this option's
1618 effect on the #warning and #error preprocessor directives and
1619 the deprecated attribute.
1621 2018-06-15 Richard Biener <rguenther@suse.de>
1623 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1624 here, also noting vector size used.
1625 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1626 size used in MSG_OPTIMIZED_LOCATIONS dump.
1627 (pass_slp_vectorize::execute): Adjust.
1629 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1632 * config/arc/arc.c (arc_return_address_register): Fix
1635 2018-06-15 Richard Biener <rguenther@suse.de>
1638 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1639 leave useless conversion stripping to force_gimple_operand_gsi.
1640 (gimplify_build2): Likewise.
1641 (gimplify_build1): Likewise.
1643 2018-06-15 Richard Biener <rguenther@suse.de>
1646 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1647 before adjusting their block.
1649 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1651 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1652 multilibs for *-*-rtems*.
1653 * config/riscv/t-rtems: New file.
1655 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1658 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1659 unsigned_type_for returns NULL.
1662 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1663 subregs of multi-word pseudos unless the float mode has word size.
1665 2018-06-14 Richard Biener <rguenther@suse.de>
1668 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1669 duplicate and harmful type_hash_canon.
1670 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1672 2018-06-14 Richard Biener <rguenther@suse.de>
1675 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1678 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1680 * config/rtems.h (STDINT_LONG32): Define.
1682 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1683 Prachi Godbole <prachi.godbole@imgtec.com>
1685 * config/mips/mips-cpus.def: Define P6600.
1686 * config/mips/mips-tables.opt: Regenerate.
1687 * config/mips/mips.c (mips_ucbranch_type): New enum.
1688 (mips_rtx_cost_data): Add support for P6600.
1689 (mips_issue_rate): Likewise.
1690 (mips_multipass_dfa_lookahead): Likewise.
1691 (mips_avoid_hazard): Likewise.
1692 (mips_reorg_process_insns): Likewise.
1693 (mips_classify_branch_p6600): New function.
1694 * config/mips/mips.h (TUNE_P6600): New define.
1695 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1696 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1697 * config/mips/mips.md: Include p6600.md.
1698 (processor): Add p6600.
1699 * config/mips/p6600.md: New file.
1700 * doc/invoke.texi: Add p6600 to supported architectures.
1702 2018-06-13 Martin Sebor <msebor@redhat.com>
1704 PR tree-optimization/86114
1705 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1707 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1709 2018-06-13 Richard Biener <rguenther@suse.de>
1711 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1712 Properly set vector type of the intermediate stmt.
1713 * tree-vect-stmts.c (vectorizable_operation): The destination
1714 var always has vectype_out type.
1716 2018-06-13 Jeff Law <law@redhat.com>
1718 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1719 integer 0 for argument to print_rtl_with_bb.
1720 (rl78_reorg): Likewise.
1722 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1724 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1725 from rtx to rtx_insn *.
1726 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1728 (add_sched_insns_for_speculation): Likewise for local "target",
1729 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1730 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1731 from rtx_insn ** to rtx_code_label **.
1732 (reorg_emit_nops): Likewise.
1733 (c6x_reorg): Likewise for local "call_labels".
1734 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1736 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1737 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1738 the loops over LABEL_REFS.
1739 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1741 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1742 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1743 (split_branches): Strengthen local "olabel" from rtx to
1744 rtx_insn *, adding a safe_as_a cast.
1745 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1747 (add_insn_after): Likewise for first two params.
1748 (add_insn_before): Likewise.
1749 (remove_insn): Likewise for param.
1750 (emit_pattern_before_noloc): Likewise for second and third params.
1751 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1752 (emit_call_insn_before_noloc): Likewise.
1753 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1755 (emit_barrier_before): Likewise.
1756 (emit_label_before): Strengthen "label" param from "rtx" to
1757 "rtx_code_label *". Strengthen "before" param from "rtx" to
1759 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1761 (emit_pattern_after_noloc): Likewise.
1762 (emit_insn_after_noloc): Likewise.
1763 (emit_jump_insn_after_noloc): Likewise.
1764 (emit_call_insn_after_noloc): Likewise.
1765 (emit_debug_insn_after_noloc): Likewise.
1766 (emit_barrier_after): Likewise.
1767 (emit_label_after): Likewise for both params.
1768 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1769 "loc" param from "int" to "location_t".
1770 (emit_insn_after_setloc): Likewise.
1771 (emit_jump_insn_after_setloc): Likewise.
1772 (emit_call_insn_after_setloc): Likewise.
1773 (emit_debug_insn_after_setloc): Likewise.
1774 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1775 "loc" param from "int" to "location_t".
1776 (emit_pattern_before): Convert NULL_RTX to NULL.
1777 (emit_insn_before_setloc): Convert "loc" param from "int" to
1779 (emit_jump_insn_before_setloc): Likewise.
1780 (emit_call_insn_before_setloc): Likewise.
1781 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1782 rtx_insn *. Convert "loc" param from "int" to "location_t".
1783 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1784 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1785 Convert 3rd param from "int" to "location_t".
1786 (emit_barrier_before, emit_barrier_after, next_real_insn):
1787 Strengthen param from rtx to rtx_insn *.
1788 (emit_label_before): Strengthen 1st param from "rtx" to
1789 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1791 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1792 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1793 Strengthen 2nd param from "rtx" to "rtx_insn *".
1794 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1795 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1796 Likewise. Convert 3rd param from "int" to "location_t".
1797 (emit_label_after): Strengthen 1st param from "rtx" to
1799 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1801 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1802 from "rtx" to "rtx_insn *".
1804 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1806 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1807 bodies streamed in with -Q.
1808 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1809 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1810 * gimple-streamer-out.c: Include gimple-pretty-print.h
1811 (output_bb): Dump stmts streamed.
1812 * lto-section-out.c: Include print-tree.h
1813 (lto_begin_section): Dump sections created.
1814 (lto_output_decl_index): Dump decl encoded.
1815 * lto-streamer-out.c: Include print-tree.h
1816 (create_output_block): Dump output block created.
1817 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1818 (output_function): Dump function output.
1819 (output_constructor): Dump constructor streamed.
1820 (write_global_stream): Output indexes encoded.
1821 (produce_asm_for_decls): Dump streams encoded.
1822 * lto-streamer.c (streamer_dump_file): New global var.
1823 * lto-streamer.h (streamer_dump_file): Declare.
1824 * passes.c (ipa_write_summaries): Initialize streamer dump.
1825 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1828 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1831 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1832 offsets for register save directives. Emit a second batch of save
1833 directives, if need be, when the function accesses prior frames.
1835 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1837 * config/arc/fpu.md (fmasf4): Force operand to register.
1838 (fnmasf4): Likewise.
1840 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1842 * config/arc/arc-protos.h (arc_pad_return): Remove.
1843 * config/arc/arc.c (machine_function): Remove force_short_suffix
1845 (arc_print_operand): Adjust printing of '&'.
1846 (arc_verify_short): Remove conditional printing of short suffix.
1847 (arc_final_prescan_insn): Remove reference to size_reason.
1848 (pad_return): New function.
1849 (arc_reorg): Call pad_return.
1850 (arc_pad_return): Remove.
1851 (arc_init_machine_status): Remove reference to force_short_suffix.
1852 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1853 (attr length): When attribute iscompact is true force to 2
1854 regardless; in the case of maybe check if we want to force the
1855 instruction to have 4 bytes length.
1856 (nopv): Change it to generate 4 byte long nop as well.
1857 (blockage): New pattern.
1858 (simple_return): Remove call to arc_pad_return.
1859 (p_return_i): Likewise.
1861 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1863 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1865 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1867 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1870 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1872 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1873 for ARC700 and ARCv2.
1875 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1878 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1879 operands[2] instead of operands[1].
1882 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1884 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1885 case, check whether the outer register overlaps an unallocatable
1886 register, not just whether it fits the required class.
1888 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1890 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1892 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1893 (memrefs_conflict_p): Likewise.
1894 (init_alias_analysis): Likewise.
1895 * cfgexpand.c (expand_debug_expr): Likewise.
1896 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1897 * cse.c (fold_rtx): Likewise.
1898 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1899 * expr.c (emit_block_move_hints): Likewise.
1900 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1901 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1902 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1903 (emit_group_store): Likewise.
1904 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1905 to read the PRE/POST_MODIFY increment.
1906 * calls.c (store_one_arg): Use strip_offset.
1907 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1909 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1911 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1912 (simplify_binary_operation_1): Extend CONST_INT handling to
1914 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1915 than a HOST_WIDE_INT.
1916 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1918 (adjust_mems, add_stores): Update accodingly.
1919 (vt_canonicalize_addr): Track polynomial offsets.
1920 (emit_note_insn_var_location): Likewise.
1921 (vt_add_function_parameter): Likewise.
1922 (vt_initialize): Likewise.
1924 2018-06-12 Jeff Law <law@redhat.com>
1926 * config.gcc (alpha*-*-freebsd*): Remove.
1927 * config/alpha/freebsd.h: Remove.
1929 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1932 * spellcheck-tree.c (levenshtein_distance): Rename to...
1933 (get_edit_distance): ...this, and update for underlying renaming.
1934 * spellcheck-tree.h (levenshtein_distance): Rename to...
1935 (get_edit_distance): ...this.
1936 * spellcheck.c (levenshtein_distance): Rename to...
1937 (get_edit_distance): ...this. Convert from Levenshtein distance
1938 to Damerau-Levenshtein distance by supporting transpositions of
1939 adjacent characters. Rename "v1" to "v_next" and "v0" to
1941 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1942 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1943 update for underlying renaming.
1944 (selftest::levenshtein_distance_unit_test): Rename to...
1945 (selftest::test_get_edit_distance_unit): ...this, and update for
1946 underlying renaming.
1947 (selftest::test_find_closest_string): Add example from PR 69968
1948 where transposition helps
1949 (selftest::test_metric_conditions): Update for renaming.
1950 (selftest::test_metric_conditions): Likewise.
1951 (selftest::spellcheck_c_tests): Likewise.
1952 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1953 (get_edit_distance): ...this.
1954 (best_match::consider): Update for renaming.
1956 2018-06-12 Martin Sebor <msebor@redhat.com>
1958 PR tree-optimization/85259
1959 * builtins.c (compute_objsize): Handle constant offsets.
1960 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1961 true iff a warning has been issued.
1962 * gimple.h (gimple_nonartificial_location): New function.
1963 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1964 gimple_nonartificial_location and handle -Wno-system-headers.
1965 (handle_builtin_stxncpy): Same.
1967 2018-06-12 Martin Sebor <msebor@redhat.com>
1970 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1972 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1974 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1975 BUILTIN_VEC_XST entries for pointer to double and long long.
1977 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1980 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1982 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1985 2018-06-12 Martin Liska <mliska@suse.cz>
1987 * doc/options.texi: Document IntegerRange.
1989 2018-06-12 Martin Liska <mliska@suse.cz>
1991 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1992 * opt-functions.awk: Handle Deprecated flag.
1993 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1995 (read_cmdline_option): Report warning for a deprecated option.
1996 * opts.h (struct cl_option): Add new field cl_deprecated.
1997 (CL_ERR_DEPRECATED): New.
1999 2018-06-12 Martin Liska <mliska@suse.cz>
2001 * doc/options.texi: Document Deprecated option flag.
2003 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2005 * config/arc/arc-arch.h (arc_extras): New enum.
2006 (arc_cpu_t):Add field extra.
2007 (arc_cpu_types): Consider the extras.
2008 * config/arc/arc-cpus.def: Add extras info.
2009 * config/arc/arc-opts.h (processor_type): Consider extra field.
2010 * config/arc/arc.c (arc_override_options): Handle extra field.
2012 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2014 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2015 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2017 * config/arc/arc.c (arc_init): Update pic warning.
2018 (irq_range): Update irq range parsing warnings.
2019 (arc_override_options): Update various warning messages.
2020 (arc_handle_aux_attribute): Likewise.
2022 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2024 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2026 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2028 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2029 compiler messages directives.
2031 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2033 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2034 * config/mips/mips-tables.opt: Regenerate.
2035 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2037 * doc/invoke.texi: Document -march=i6500.
2039 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2041 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2042 (i6400_gpmul): Add cpu_unit.
2043 (i6400_gpdiv): Likewise.
2044 (i6400_msa_add_d): Update reservations.
2045 (i6400_msa_int_add) Likewise.
2046 (i6400_msa_short_logic3) Likewise.
2047 (i6400_msa_short_logic2) Likewise.
2048 (i6400_msa_short_logic) Likewise.
2049 (i6400_msa_move) Likewise.
2050 (i6400_msa_cmp) Likewise.
2051 (i6400_msa_short_float2) Likewise.
2052 (i6400_msa_div_d) Likewise.
2053 (i6400_msa_long_logic1) Likewise.
2054 (i6400_msa_long_logic2) Likewise.
2055 (i6400_msa_mult) Likewise.
2056 (i6400_msa_long_float2) Likewise.
2057 (i6400_msa_long_float4) Likewise.
2058 (i6400_msa_long_float5) Likewise.
2059 (i6400_msa_long_float8) Likewise.
2060 (i6400_fpu_fadd): Include frint type.
2061 (i6400_fpu_store): New define_insn_reservation.
2062 (i6400_fpu_load): Likewise.
2063 (i6400_fpu_move): Likewise.
2064 (i6400_fpu_fcmp): Likewise.
2065 (i6400_fpu_fmadd): Likewise.
2066 (i6400_int_mult): Include imul3nc type and update reservation.
2067 (i6400_int_div): Include idiv3 type and update reservation.
2068 (i6400_int_load): Update to check type not move_type.
2069 (i6400_int_store): Likewise.
2070 (i6400_int_prefetch): Set zero latency.
2072 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2074 * gcc.c: Document new %@{...} sequence.
2075 (LINK_COMMAND_SPEC): Use it for the -L switches.
2076 (cpp_unique_options): Use it for the -I switches.
2077 (at_file_argbuf): New global variable.
2078 (in_at_file): Likewise.
2079 (alloc_args): Create at_file_argbuf.
2080 (clear_args): Truncate at_file_argbuf.
2081 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2082 (open_at_file): New function.
2083 (close_at_file): Likewise.
2084 (create_at_file): Delete.
2085 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2088 (validate_switches_from_spec): Deal with %@{...} sequence.
2089 (validate_switches): Likewise.
2090 (driver::finalize): Call clear_args.
2092 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2094 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2096 2018-06-11 Martin Sebor <msebor@redhat.com>
2098 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2099 (Wno-class-memaccess): Add @opindex.
2100 (Wno-templates, Wno-multiple-inheritance): Same.
2101 (Wno-virtual-inheritance, Wno-namespaces): Same.
2102 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2103 (Wno-format-overflow, Wno-format-truncation): Same.
2104 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2105 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2106 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2107 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2108 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2109 (Wno-misspelled-isr): Same.
2111 2018-06-11 Martin Sebor <msebor@redhat.com>
2113 * PR tree-optimization/86083
2114 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2116 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2118 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2120 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2123 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2124 on the correct operand.
2125 (*movdi_internal64): Ditto.
2127 2018-06-11 Martin Liska <mliska@suse.cz>
2129 PR tree-optimization/86089
2130 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2132 2018-06-11 Julia Koval <julia.koval@intel.com>
2134 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2135 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2136 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2138 2018-06-11 Olivier Hainque <hainque@adacore.com>
2140 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2141 for Ada with strict dwarf2.
2143 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2146 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2149 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2151 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2153 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2155 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2158 2018-06-08 Martin Liska <mliska@suse.cz>
2160 * tree-cfg.h (debug_function): Fix argument type to match
2163 2018-06-08 Martin Liska <mliska@suse.cz>
2165 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2166 Remove usage of MPX-related (and removed) fields.
2167 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2169 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2171 * cfg.c (debug): Use TDF_NONE rather than 0.
2172 * cfghooks.c (debug): Likewise.
2173 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2174 (struct dump_option_value_info): Convert to...
2175 (struct kv_pair): ...this template type.
2176 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2178 (optinfo_verbosity_options): Likewise.
2179 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2181 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2182 than int for "optgroup_flags" param.
2183 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2185 (dump_dec): Likewise.
2186 (dump_finish): Use TDF_NONE rather than 0.
2187 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2188 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2189 than 0. Update for change to option_ptr.
2190 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2191 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2192 0. Update for changes to optinfo_verbosity_options and
2194 (opt_info_switch_p): Convert optgroup_flags from int to
2196 (dump_basic_block): Use dump_flags_t rather than int
2197 for "dump_kind" param.
2198 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2199 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2200 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2201 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2202 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2203 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2204 TDF_NONE): Convert from macros to...
2205 (enum dump_flag): ...this new enum.
2206 (dump_flags_t): Update to use enum.
2207 (operator|, operator&, operator~, operator|=, operator&=):
2208 Implement for dump_flags_t.
2209 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2210 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2211 Convert from macros to...
2212 (enum optgroup_flag): ...this new enum.
2213 (optgroup_flags_t): New typedef.
2214 (operator|, operator|=): Implement for optgroup_flags_t.
2215 (struct dump_file_info): Convert field "alt_flags" to
2216 dump_flags_t. Convert field "optgroup_flags" to
2218 (dump_basic_block): Use dump_flags_t rather than int for param.
2219 (dump_generic_expr_loc): Likewise.
2220 (dump_dec): Likewise.
2221 (dump_register): Convert param "optgroup_flags" to
2223 (opt_info_enable_passes): Likewise.
2224 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2226 * gimple-pretty-print.c (debug): Likewise.
2227 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2228 (merged_store_group::apply_stores): Likewise.
2229 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2230 * gimple.c (verify_gimple_pp): Likewise.
2231 * graphite-poly.c (print_pbb_body): Likewise.
2232 * passes.c (pass_manager::register_one_dump_file): Convert
2233 local "optgroup_flags" to optgroup_flags_t.
2234 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2236 (debug_body): Likewise.
2237 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2238 to optgroup_flags_t.
2239 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2241 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2242 (convert_mult_to_fma): Likewise.
2243 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2244 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2245 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2246 "dump_kind" to dump_flags_t.
2248 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2250 * config/rs6000/rs6000.c (min, max): Delete.
2252 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2254 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2255 -mabi=spe and -mabi=no-spe.
2257 2018-06-08 Martin Liska <mliska@suse.cz>
2259 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2260 where we expect an existing summary.
2262 2018-06-08 Martin Liska <mliska@suse.cz>
2264 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2265 * ipa-inline.h (estimate_edge_growth): Likewise.
2267 2018-06-08 Martin Liska <mliska@suse.cz>
2269 * cgraph.c (function_version_hasher::hash): Use
2270 cgraph_node::get_uid ().
2271 (function_version_hasher::equal):
2272 * cgraph.h (cgraph_node::get_uid): New method.
2273 * ipa-inline.c (update_caller_keys): Use
2274 cgraph_node::get_uid ().
2275 (update_callee_keys): Likewise.
2276 * ipa-utils.c (searchc): Likewise.
2277 (ipa_reduced_postorder): Likewise.
2278 * lto-cgraph.c (input_node): Likewise.
2279 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2280 * symbol-summary.h (symtab_insertion): Likewise.
2281 (symtab_removal): Likewise.
2282 (symtab_duplication): Likewise.
2283 * tree-pretty-print.c (dump_function_header): Likewise.
2284 * tree-sra.c (convert_callers_for_node): Likewise.
2286 2018-06-08 Martin Liska <mliska@suse.cz>
2288 * cgraph.c (symbol_table::create_edge): Always assign a new
2290 (symbol_table::free_edge): Do not recycle numbers.
2291 * cgraph.h (cgraph_edge::get): New method.
2292 * symbol-summary.h (symtab_removal): Use it.
2293 (symtab_duplication): Likewise.
2294 (call_summary::hashable_uid): Remove.
2296 2018-06-08 Martin Liska <mliska@suse.cz>
2298 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2299 (initialize_growth_caches): Remove.
2300 (free_growth_caches): Likewise.
2301 (do_estimate_edge_time): Use edge_growth_cache.
2302 (do_estimate_edge_size): Likewise.
2303 (do_estimate_edge_hints): Likewise.
2304 * ipa-inline.c (reset_edge_caches): Likewise.
2305 (recursive_inlining): Likewise.
2306 (inline_small_functions): Likewise.
2307 * ipa-inline.h (initialize_growth_caches): Remove.
2308 (estimate_edge_size): Likewise.
2309 (estimate_edge_time): Likewise.
2310 (estimate_edge_hints): Likewise.
2311 (reset_edge_growth_cache): Likewise.
2312 * symbol-summary.h (call_summary::remove): New method.
2314 2018-06-08 Martin Liska <mliska@suse.cz>
2316 * ipa-cp.c (class edge_clone_summary): New summary.
2317 (grow_edge_clone_vectors): Remove.
2318 (ipcp_edge_duplication_hook): Remove.
2319 (class edge_clone_summary_t): New call_summary class.
2320 (ipcp_edge_removal_hook): Remove.
2321 (edge_clone_summary_t::duplicate): New function.
2322 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2323 (create_specialized_node): Likewise.
2324 (ipcp_driver): Initialize edge_clone_summaries and do not
2327 2018-06-08 Martin Liska <mliska@suse.cz>
2329 * symbol-summary.h (get): New function.
2330 (call_summary::m_initialize_when_cloning): New class member.
2332 2018-06-08 Martin Liska <mliska@suse.cz>
2334 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2335 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2336 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2337 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2338 (remove_cgraph_node_from_order): Use the removed_nodes set.
2339 (do_per_function_toporder): Likwise.
2340 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2341 instead of summary_uid.
2342 (symtab_removal): Likewise.
2343 (symtab_duplication): Likewise.
2345 2018-06-08 Martin Liska <mliska@suse.cz>
2347 * ipa-cp.c (ipcp_store_bits_results): Use
2348 ipcp_transformation_sum.
2349 (ipcp_store_vr_results): Likewise.
2350 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2352 (ipcp_transformation_initialize): ... this.
2353 (ipa_set_node_agg_value_chain):
2354 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2355 (write_ipcp_transformation_info): Likewise.
2356 (read_ipcp_transformation_info): Likewise.
2357 (ipcp_update_bits): Likewise.
2358 (ipcp_update_vr): Likewise.
2359 (ipcp_transform_function): Likewise.
2360 * ipa-prop.h: Rename ipcp_transformation_summary to
2361 ipcp_transformation.
2362 (class ipcp_transformation_t): New function summary.
2363 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2364 (ipa_get_agg_replacements_for_node): Likewise.
2366 2018-06-08 Martin Liska <mliska@suse.cz>
2368 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2370 (class funct_state_summary_t): New function_summary class.
2371 (has_function_state): Remove.
2372 (get_function_state): Likewise.
2373 (set_function_state): Likewise.
2374 (add_new_function): Likewise.
2375 (funct_state_summary_t::insert): New function.
2376 (duplicate_node_data): Remove.
2377 (remove_node_data): Remove.
2378 (funct_state_summary_t::duplicate): New function.
2379 (register_hooks): Create new funct_state_summaries.
2380 (pure_const_generate_summary): Use it.
2381 (pure_const_write_summary): Likewise.
2382 (pure_const_read_summary): Likewise.
2383 (propagate_pure_const): Likewise.
2384 (propagate_nothrow): Likewise.
2385 (dump_malloc_lattice): Likewise.
2386 (propagate_malloc): Likewise.
2387 (execute): Do not register hooks, just remove summary
2389 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2392 2018-06-08 Martin Liska <mliska@suse.cz>
2394 * ipa-reference.c (remove_node_data): Remove.
2395 (duplicate_node_data): Likewise.
2396 (class ipa_ref_var_info_summary_t): New class.
2397 (class ipa_ref_opt_summary_t): Likewise.
2398 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2399 (get_reference_optimization_summary): Use
2400 ipa_ref_opt_sum_summaries.
2401 (set_reference_vars_info): Remove.
2402 (set_reference_optimization_summary): Likewise.
2403 (ipa_init): Create summaries.
2404 (init_function_info): Use function summary.
2405 (ipa_ref_opt_summary_t::duplicate): New function.
2406 (ipa_ref_opt_summary_t::remove): New function.
2407 (get_read_write_all_from_node): Fix GNU coding style.
2408 (propagate): Use function summary.
2409 (write_node_summary_p): Fix GNU coding style.
2410 (stream_out_bitmap): Likewise.
2411 (ipa_reference_read_optimization_summary): Use function summary.
2412 (ipa_reference_c_finalize): Do not release hooks.
2414 2018-06-08 Martin Liska <mliska@suse.cz>
2416 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2417 (analyze_function_body): Extract multiple calls of get_create.
2418 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2419 * ipa-inline.c (recursive_inlining): Use ::get method.
2420 * ipa-inline.h (estimate_edge_growth): Likewise.
2422 2018-06-08 Martin Liska <mliska@suse.cz>
2424 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2426 (hsa_function_summary::hsa_function_summary): Use the new enum
2428 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2429 * hsa-gen.c (hsa_get_host_function): Likewise.
2430 (get_brig_function_name): Likewise.
2431 * ipa-hsa.c (process_hsa_functions): Likewise.
2432 (ipa_hsa_write_summary): Likewise.
2433 * symbol-summary.h (symtab_duplication): Use ::get function/
2434 (get): New function.
2436 2018-06-08 Martin Liska <mliska@suse.cz>
2438 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2440 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2441 (hsa_register_kernel): Likewise.
2442 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2443 * hsa-gen.c (hsa_get_host_function): Likewise.
2444 (get_brig_function_name): Likewise.
2445 (generate_hsa): Likewise.
2446 (pass_gen_hsail::execute): Likewise.
2447 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2448 (devirtualization_time_bonus): Likewise.
2449 (ipcp_propagate_stage): Likewise.
2450 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2451 (edge_set_predicate): Likewise.
2452 (evaluate_conditions_for_known_args): Likewise.
2453 (evaluate_properties_for_edge): Likewise.
2454 (ipa_fn_summary::reset): Likewise.
2455 (ipa_fn_summary_t::duplicate): Likewise.
2456 (dump_ipa_call_summary): Likewise.
2457 (ipa_dump_fn_summary): Likewise.
2458 (analyze_function_body): Likewise.
2459 (compute_fn_summary): Likewise.
2460 (estimate_edge_devirt_benefit): Likewise.
2461 (estimate_edge_size_and_time): Likewise.
2462 (estimate_calls_size_and_time): Likewise.
2463 (estimate_node_size_and_time): Likewise.
2464 (inline_update_callee_summaries): Likewise.
2465 (remap_edge_change_prob): Likewise.
2466 (remap_edge_summaries): Likewise.
2467 (ipa_merge_fn_summary_after_inlining): Likewise.
2468 (ipa_update_overall_fn_summary): Likewise.
2469 (read_ipa_call_summary): Likewise.
2470 (inline_read_section): Likewise.
2471 (write_ipa_call_summary): Likewise.
2472 (ipa_fn_summary_write): Likewise.
2473 (ipa_free_fn_summary): Likewise.
2474 * ipa-hsa.c (process_hsa_functions): Likewise.
2475 (ipa_hsa_write_summary): Likewise.
2476 (ipa_hsa_read_section): Likewise.
2477 * ipa-icf.c (sem_function::merge): Likewise.
2478 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2479 (do_estimate_edge_time): Likewise.
2480 (estimate_size_after_inlining): Likewise.
2481 (estimate_growth): Likewise.
2482 (growth_likely_positive): Likewise.
2483 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2484 (inline_call): Likewise.
2485 * ipa-inline.c (caller_growth_limits): Likewise.
2486 (can_inline_edge_p): Likewise.
2487 (can_inline_edge_by_limits_p): Likewise.
2488 (compute_uninlined_call_time): Likewise.
2489 (compute_inlined_call_time): Likewise.
2490 (want_inline_small_function_p): Likewise.
2491 (edge_badness): Likewise.
2492 (update_caller_keys): Likewise.
2493 (update_callee_keys): Likewise.
2494 (recursive_inlining): Likewise.
2495 (inline_small_functions): Likewise.
2496 (inline_to_all_callers_1): Likewise.
2497 (dump_overall_stats): Likewise.
2498 (early_inline_small_functions): Likewise.
2499 (early_inliner): Likewise.
2500 * ipa-inline.h (estimate_edge_growth): Likewise.
2501 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2502 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2503 * ipa-prop.h (IPA_NODE_REF): Likewise.
2504 (IPA_EDGE_REF): Likewise.
2505 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2506 (propagate_malloc): Likewise.
2507 * ipa-split.c (execute_split_functions): Likewise.
2508 * symbol-summary.h: Rename get to get_create.
2510 (get_create): Likewise.
2511 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2513 2018-06-08 Martin Liska <mliska@suse.cz>
2515 * symbol-summary.h (release): Move definition out of class
2517 (symtab_removal): Likewise.
2518 (symtab_duplication): Likewise.
2520 2018-06-08 Martin Liska <mliska@suse.cz>
2522 * symbol-summary.h (function_summary): Move constructor
2523 implementation out of class declaration.
2524 (release): Likewise.
2525 (symtab_insertion): Likewise.
2526 (symtab_removal): Likewise.
2527 (symtab_duplication): Likewise.
2530 2018-06-08 Martin Liska <mliska@suse.cz>
2532 * Makefile.in: Remove support for MPX (macros, related functions,
2533 fields in cgraph_node, ...).
2534 * builtin-types.def (BT_BND): Likewise.
2535 (BT_FN_BND_CONST_PTR): Likewise.
2536 (BT_FN_CONST_PTR_BND): Likewise.
2537 (BT_FN_VOID_PTR_BND): Likewise.
2538 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2539 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2540 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2541 (expand_builtin_mempcpy_with_bounds): Likewise.
2542 (expand_builtin_memset_with_bounds): Likewise.
2543 (expand_builtin_memset_args): Likewise.
2544 (std_expand_builtin_va_start): Likewise.
2545 (expand_builtin): Likewise.
2546 (expand_builtin_with_bounds): Likewise.
2547 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2548 (DEF_LIB_BUILTIN_CHKP): Likewise.
2549 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2550 (DEF_CHKP_BUILTIN): Likewise.
2551 (BUILT_IN_MEMCPY): Likewise.
2552 (BUILT_IN_MEMMOVE): Likewise.
2553 (BUILT_IN_MEMPCPY): Likewise.
2554 (BUILT_IN_MEMSET): Likewise.
2555 (BUILT_IN_STPCPY): Likewise.
2556 (BUILT_IN_STRCAT): Likewise.
2557 (BUILT_IN_STRCHR): Likewise.
2558 (BUILT_IN_STRCPY): Likewise.
2559 (BUILT_IN_STRLEN): Likewise.
2560 (BUILT_IN_MEMCPY_CHK): Likewise.
2561 (BUILT_IN_MEMMOVE_CHK): Likewise.
2562 (BUILT_IN_MEMPCPY_CHK): Likewise.
2563 (BUILT_IN_MEMSET_CHK): Likewise.
2564 (BUILT_IN_STPCPY_CHK): Likewise.
2565 (BUILT_IN_STRCAT_CHK): Likewise.
2566 (BUILT_IN_STRCPY_CHK): Likewise.
2567 * calls.c (store_bounds): Likewise.
2568 (emit_call_1): Likewise.
2569 (special_function_p): Likewise.
2570 (maybe_warn_nonstring_arg): Likewise.
2571 (initialize_argument_information): Likewise.
2572 (finalize_must_preallocate): Likewise.
2573 (compute_argument_addresses): Likewise.
2574 (expand_call): Likewise.
2575 * cfgexpand.c (expand_call_stmt): Likewise.
2576 (expand_return): Likewise.
2577 (expand_gimple_stmt_1): Likewise.
2578 (pass_expand::execute): Likewise.
2579 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2580 (cgraph_node::remove): Likewise.
2581 (cgraph_node::dump): Likewise.
2582 (cgraph_node::verify_node): Likewise.
2583 * cgraph.h (chkp_function_instrumented_p): Likewise.
2584 (symtab_node::get_alias_target): Likewise.
2585 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2586 (cgraph_local_p): Likewise.
2587 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2588 (cgraph_edge::rebuild_references): Likewise.
2589 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2590 (walk_polymorphic_call_targets): Likewise.
2591 (cgraph_node::expand_thunk): Likewise.
2592 (symbol_table::output_weakrefs): Likewise.
2593 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2594 (ix86_handle_option): Likewise.
2595 * config/i386/constraints.md: Likewise.
2596 * config/i386/i386-builtin-types.def (BND): Likewise.
2600 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2601 (BDESC_FIRST): Likewise.
2603 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2604 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2605 * config/i386/i386.c (enum reg_class): Likewise.
2606 (ix86_target_string): Likewise.
2607 (ix86_option_override_internal): Likewise.
2608 (ix86_conditional_register_usage): Likewise.
2609 (ix86_valid_target_attribute_inner_p): Likewise.
2610 (ix86_set_indirect_branch_type): Likewise.
2611 (ix86_set_current_function): Likewise.
2612 (ix86_function_arg_regno_p): Likewise.
2613 (init_cumulative_args): Likewise.
2614 (ix86_function_arg_advance): Likewise.
2615 (ix86_function_arg): Likewise.
2616 (ix86_pass_by_reference): Likewise.
2617 (ix86_function_value_regno_p): Likewise.
2618 (ix86_function_value_1): Likewise.
2619 (ix86_function_value_bounds): Likewise.
2620 (ix86_return_in_memory): Likewise.
2621 (ix86_setup_incoming_vararg_bounds): Likewise.
2622 (ix86_va_start): Likewise.
2623 (indirect_thunk_need_prefix): Likewise.
2624 (print_reg): Likewise.
2625 (ix86_print_operand): Likewise.
2626 (ix86_expand_call): Likewise.
2627 (ix86_output_function_return): Likewise.
2628 (reg_encoded_number): Likewise.
2629 (BDESC_VERIFYS): Likewise.
2630 (ix86_init_mpx_builtins): Likewise.
2631 (ix86_init_builtins): Likewise.
2632 (ix86_emit_cmove): Likewise.
2633 (ix86_emit_move_max): Likewise.
2634 (ix86_expand_builtin): Likewise.
2635 (ix86_builtin_mpx_function): Likewise.
2636 (ix86_get_arg_address_for_bt): Likewise.
2637 (ix86_load_bounds): Likewise.
2638 (ix86_store_bounds): Likewise.
2639 (ix86_load_returned_bounds): Likewise.
2640 (ix86_store_returned_bounds): Likewise.
2641 (ix86_class_likely_spilled_p): Likewise.
2642 (ix86_hard_regno_mode_ok): Likewise.
2643 (x86_order_regs_for_local_alloc): Likewise.
2644 (ix86_mitigate_rop): Likewise.
2645 (ix86_bnd_prefixed_insn_p): Likewise.
2646 (ix86_mpx_bound_mode): Likewise.
2647 (ix86_make_bounds_constant): Likewise.
2648 (ix86_initialize_bounds): Likewise.
2649 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2650 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2651 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2652 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2653 (TARGET_CHKP_BOUND_MODE): Likewise.
2654 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2655 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2656 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2657 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2658 * config/i386/i386.h (TARGET_MPX): Likewise.
2659 (TARGET_MPX_P): Likewise.
2660 (VALID_BND_REG_MODE): Likewise.
2661 (FIRST_BND_REG): Likewise.
2662 (LAST_BND_REG): Likewise.
2663 (enum reg_class): Likewise.
2664 (BND_REG_P): Likewise.
2665 (BND_REGNO_P): Likewise.
2666 (BNDmode): Likewise.
2667 (ADJUST_INSN_LENGTH): Likewise.
2668 * config/i386/i386.md: Likewise.
2669 * config/i386/i386.opt: Likewise.
2670 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2671 (defined): Likewise.
2672 (LINK_MPX): Likewise.
2673 (MPX_SPEC): Likewise.
2674 (LIBMPX_SPEC): Likewise.
2675 (LIBMPXWRAPPERS_SPEC): Likewise.
2676 (CHKP_SPEC): Likewise.
2677 * config/i386/predicates.md: Likewise.
2678 * dbxout.c (dbxout_type): Likewise.
2679 * doc/extend.texi: Likewise.
2680 * doc/invoke.texi: Likewise.
2681 * doc/md.texi: Likewise.
2682 * doc/tm.texi: Likewise.
2683 * doc/tm.texi.in: Likewise.
2684 * dwarf2out.c (is_base_type): Likewise.
2685 (gen_formal_types_die): Likewise.
2686 (gen_subprogram_die): Likewise.
2687 (gen_type_die_with_usage): Likewise.
2688 (gen_decl_die): Likewise.
2689 (dwarf2out_late_global_decl): Likewise.
2690 * expr.c (expand_assignment): Likewise.
2691 (emit_storent_insn): Likewise.
2692 (store_expr_with_bounds): Likewise.
2693 (store_expr): Likewise.
2694 (expand_expr_real_1): Likewise.
2695 * expr.h (store_expr_with_bounds): Likewise.
2696 * function.c (use_register_for_decl): Likewise.
2697 (struct bounds_parm_data): Likewise.
2698 (assign_parms_augmented_arg_list): Likewise.
2699 (assign_parm_find_entry_rtl): Likewise.
2700 (assign_parm_is_stack_parm): Likewise.
2701 (assign_parm_load_bounds): Likewise.
2702 (assign_bounds): Likewise.
2703 (assign_parms): Likewise.
2704 (expand_function_start): Likewise.
2705 * gcc.c (CHKP_SPEC): Likewise.
2706 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2707 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2708 (wrestrict_dom_walker::check_call): Likewise.
2709 * gimple.c (gimple_build_call_from_tree): Likewise.
2710 * gimple.h (enum gf_mask): Likewise.
2711 (gimple_call_with_bounds_p): Likewise.
2712 (gimple_call_set_with_bounds): Likewise.
2713 * gimplify.c (gimplify_init_constructor): Likewise.
2714 * ipa-cp.c (initialize_node_lattices): Likewise.
2715 (propagate_constants_across_call): Likewise.
2716 (find_more_scalar_values_for_callers_subset): Likewise.
2717 * ipa-hsa.c (process_hsa_functions): Likewise.
2718 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2719 * ipa-icf.c (sem_function::merge): Likewise.
2720 * ipa-inline.c (early_inliner): Likewise.
2721 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2722 (warn_function_cold): Likewise.
2723 (propagate_pure_const): Likewise.
2724 * ipa-ref.h (enum GTY): Likewise.
2725 * ipa-split.c (find_retbnd): Likewise.
2726 (consider_split): Likewise.
2727 (split_function): Likewise.
2728 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2729 * ipa.c (walk_polymorphic_call_targets): Likewise.
2730 (symbol_table::remove_unreachable_nodes): Likewise.
2731 (process_references): Likewise.
2732 (cgraph_build_static_cdtor_1): Likewise.
2733 * lto-cgraph.c (lto_output_node): Likewise.
2734 (output_refs): Likewise.
2735 (compute_ltrans_boundary): Likewise.
2736 (input_overwrite_node): Likewise.
2737 (input_node): Likewise.
2738 (input_cgraph_1): Likewise.
2739 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2740 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2741 (class pass_chkp_instrumentation_passes): Likewise.
2742 (make_pass_chkp_instrumentation_passes): Likewise.
2743 * passes.def: Likewise.
2744 * rtl.h (struct GTY): Likewise.
2745 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2746 * stor-layout.c (layout_type): Likewise.
2747 * symtab.c: Likewise.
2748 * target.def: Likewise.
2749 * targhooks.c (default_chkp_bound_type): Likewise.
2750 (default_chkp_bound_mode): Likewise.
2751 (default_builtin_chkp_function): Likewise.
2752 (default_chkp_function_value_bounds): Likewise.
2753 (default_chkp_make_bounds_constant): Likewise.
2754 (default_chkp_initialize_bounds): Likewise.
2755 * targhooks.h (default_chkp_bound_type): Likewise.
2756 (default_chkp_bound_mode): Likewise.
2757 (default_builtin_chkp_function): Likewise.
2758 (default_chkp_function_value_bounds): Likewise.
2759 (default_chkp_make_bounds_constant): Likewise.
2760 (default_chkp_initialize_bounds): Likewise.
2761 * toplev.c (compile_file): Likewise.
2762 (process_options): Likewise.
2763 * tree-core.h (DEF_BUILTIN): Likewise.
2764 (DEF_BUILTIN_CHKP): Likewise.
2765 * tree-inline.c (declare_return_variable): Likewise.
2766 (remap_gimple_stmt): Likewise.
2767 (copy_bb): Likewise.
2768 (initialize_inlined_parameters): Likewise.
2769 (expand_call_inline): Likewise.
2770 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2771 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2772 (make_pass_ipa_chkp_produce_thunks): Likewise.
2773 (make_pass_chkp): Likewise.
2774 (make_pass_chkp_opt): Likewise.
2775 (make_pass_chkp_instrumentation_passes): Likewise.
2776 * tree-pretty-print.c (dump_generic_node): Likewise.
2777 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2778 * tree-ssa-dce.c (propagate_necessity): Likewise.
2779 (eliminate_unnecessary_stmts): Likewise.
2780 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2781 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2782 * tree-ssa-sccvn.h: Likewise.
2783 * tree-ssa-strlen.c (get_string_length): Likewise.
2784 (valid_builtin_call): Likewise.
2785 (adjust_last_stmt): Likewise.
2786 (handle_builtin_strchr): Likewise.
2787 (handle_builtin_strcpy): Likewise.
2788 (handle_builtin_stxncpy): Likewise.
2789 (handle_builtin_memcpy): Likewise.
2790 (handle_builtin_strcat): Likewise.
2791 (strlen_check_and_optimize_stmt): Likewise.
2792 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2793 * tree-streamer-in.c: Likewise.
2794 * tree-streamer.c (record_common_node): Likewise.
2795 * tree.c (tree_code_size): Likewise.
2796 (wide_int_to_tree_1): Likewise.
2797 (type_contains_placeholder_1): Likewise.
2798 (build_common_tree_nodes): Likewise.
2799 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2800 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2801 (POINTER_BOUNDS_P): Likewise.
2802 (BOUNDED_TYPE_P): Likewise.
2803 (BOUNDED_P): Likewise.
2804 (CALL_WITH_BOUNDS_P): Likewise.
2805 (pointer_bounds_type_node): Likewise.
2806 * value-prof.c (gimple_ic): Likewise.
2807 * var-tracking.c (vt_add_function_parameters): Likewise.
2808 * varasm.c (make_decl_rtl): Likewise.
2809 (assemble_start_function): Likewise.
2810 (output_constant): Likewise.
2811 (maybe_assemble_visibility): Likewise.
2812 * varpool.c (ctor_for_folding): Likewise.
2813 * chkp-builtins.def: Remove.
2814 * ipa-chkp.c: Remove.
2815 * ipa-chkp.h: Remove.
2816 * rtl-chkp.c: Remove.
2817 * rtl-chkp.h: Remove.
2818 * tree-chkp-opt.c: Remove.
2819 * tree-chkp.c: Remove.
2820 * tree-chkp.h: Remove.
2822 2018-06-07 Carl Love <cel@us.ibm.com>
2824 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2825 vextract_fp_from_shortl): Add BE support.
2827 2018-06-07 Paul Koning <ni1d@arrl.net>
2829 * compare-elim.c (try_merge_compare): Don't merge compare if
2830 address contains a side effect.
2831 (try_eliminate_compare): Likewise.
2833 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2835 * config.gcc: Support "tremont".
2836 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2837 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2839 * config/i386/i386.c (m_TREMONT): Define.
2840 (processor_target_table): Add "tremont".
2841 (PTA_TREMONT): Define.
2842 (ix86_lea_outperforms): Add TARGET_TREMONT.
2843 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2844 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2845 and M_INTEL_GOLDMONT_PLUS.
2846 (fold_builtin_cpu): Add "tremont".
2847 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2848 (ix86_option_override_internal): Add "tremont".
2849 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2850 (processor_type): Add PROCESSOR_TREMONT.
2851 * config/i386/x86-tune.def: Add m_TREMONT.
2852 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2854 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2856 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2857 symbol defined for msp430i* devices to be lower case.
2859 gcc/testsuite/gcc.target/msp430/
2860 * msp430i-device-symbol.c: New test.
2861 * msp430f-device-symbol.c: New test.
2862 * msp430.h: New test header file.
2864 2018-06-07 Richard Biener <rguenther@suse.de>
2866 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2867 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2868 Properly wrap signed arithmetic if overflow wraps.
2870 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2872 PR tree-optimization/69615
2873 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2874 of a cast from a same precision integral SSA_NAME in a bb dominated
2875 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2876 cast to utype if rhs2 has already a compatible type.
2878 2018-06-07 Richard Biener <rguenther@suse.de>
2880 PR tree-optimization/85935
2881 * graphite-scop-detection.c (find_params_in_bb): Analyze
2882 condition operands with respect to the correct loop. Assert
2883 the analysis doesn't fail.
2885 2018-06-04 Carl Love <cel@us.ibm.com>
2887 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2888 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2889 as it is slightly cheaper.
2890 (first_match_or_eos_index_<mode>):
2891 Calculate index using natural element order.
2892 (first_match_index_<mode>):
2893 Calculate index using natural element order.
2894 (first_match_or_eos_index_<mode>):
2895 Calculate index using natural order.
2896 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2897 for BE and LE modes.
2898 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2899 P9V_BUILTIN_VCLZLSBB_V16QI.
2900 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2903 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2905 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2906 indentation and line wrap for many prototypes. Add missing
2907 @smallexample directives around block of prototypes for vec_xl and
2910 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2912 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2913 track if we pass or return IEEE 128-bit floating point.
2914 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2915 C++ mangling that is compatible with GCC 8.1.
2916 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2917 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2918 floating point types.
2919 (rs6000_function_arg_advance_1): Likewise.
2920 (rs6000_mangle_type): Optionally generate mangled names that match
2921 what GCC 8.1 generated for IEEE 128-bit floating point types.
2922 (rs6000_globalize_decl_name): If we have an external function that
2923 passes or returns IEEE 128-bit types, generate a weak reference
2924 from the mangled name used in GCC 8.1 to the current mangled
2926 (rs6000_init_builtins): Make __ibm128 use the long double type if
2927 long double is IBM extended double. Make __float128 use the long
2928 double type if long double is IEEE 128-bit.
2931 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2932 macro for __ibm128 built-in functions.
2933 (PACK_IF): Add __ibm128 pack/unpack functions.
2934 (UNPACK_IF): Likewise.
2935 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2936 enable long double built-in functions if long double is IEEE
2937 128-bit floating point.
2938 (rs6000_invalid_builtin): Update long double built-in function
2940 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2941 functions, adjust the built-in function to use the long double
2942 built-in function if __ibm128 and long double are the same type.
2943 * doc/extend.texi (PowerPC builtins): Update documention for
2944 __builtin_{,un}pack_longdouble. Add documentation for
2945 __builtin_{,un}pack_ibm128.
2947 2018-06-06 Jim Wilson <jimw@sifive.com>
2949 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2950 (struct machine_function): New field interrupt_mode.
2951 (riscv_handle_type_attribute): New function. Add forward declaration.
2952 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2953 (riscv_expand_epilogue): Check interrupt_mode field.
2954 (riscv_set_current_function): Check interrupt attribute args and
2955 set interrupt_mode field.
2956 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2957 (riscv_sret, riscv_uret): New.
2958 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2959 new arguments to interrupt attribute.
2961 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2964 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2965 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2967 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2969 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2970 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2971 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2972 several redundant entries.
2974 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2976 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2977 type from "rtx" to "rtx_insn *".
2978 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2979 for local "call_insn", removing cast.
2980 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2982 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2984 PR tree-optimization/86066
2985 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2986 for BIT_INSERT_EXPR stores.
2988 2018-06-06 Richard Biener <rguenther@suse.de>
2990 PR tree-optimization/86062
2991 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2992 component refs ontop
2993 of to be offsetted base.
2995 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2997 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2998 to be static and remove check on interrupt attribute name.
3000 gcc/testsuite/gcc.target/msp430/
3001 * function-attributes-4.c: New test.
3002 * static-interrupts.c: New test.
3004 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3006 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3007 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3009 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3012 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3014 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3015 Remove second argument, change how error is called.
3016 (aarch64_layout_arg): Remove second argument from
3017 aarch64_err_no_fpadvsimd call.
3018 (aarch64_init_cumulative_args): Ditto.
3019 (aarch64_gimplify_va_arg_expr): Ditto.
3020 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3022 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3024 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3025 (*simple_return_indirect_internal<mode>): Rename from
3026 simple_return_indirect_internal. Use W mode iterator.
3027 (rstorssp): New expander.
3028 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3029 (clrssbsy): New expander.
3030 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3032 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3034 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3036 (cmse_check_pointed_object): Likewise.
3038 2018-06-05 Martin Liska <mliska@suse.cz>
3040 PR gcov-profile/47618
3041 * doc/invoke.texi: Document how -fprofile-dir format
3044 2018-06-05 Richard Biener <rguenther@suse.de>
3046 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3047 removal pretend DOM info isn't available so we do not update
3048 it and only remove edges, not dominated blocks. Actually free
3049 DOM info in case we removed something. Remove unreachable blocks.
3050 (mfb_keep_latches): Work with either DOM info or marked backedges.
3051 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3052 first. Mark backedges if DOM info isn't available.
3053 (Re-)compute DOM info after cleanup_control_flow_pre.
3055 2018-06-05 Richard Biener <rguenther@suse.de>
3057 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3058 (locus_discrim_hasher::hash): Adjust.
3059 (locus_discrim_hasher::equal): Likewise.
3060 (next_discriminator_for_locus): Work on line directly.
3061 (same_line_p): Pass in expanded locus1 as well.
3062 (assign_discriminators): Avoid redundant location expansions.
3064 2018-06-05 Richard Biener <rguenther@suse.de>
3066 PR tree-optimization/86046
3067 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3068 if required after clearing TREE_ADDRESSABLE.
3070 2018-06-05 Richard Biener <rguenther@suse.de>
3072 PR tree-optimization/86047
3073 * tree-ssa-loop.c (for_each_index): Glob handling of all
3074 decls and constants and really handle all of them.
3076 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3079 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3080 qualifier_void_pointer and qualifier_const_void_pointer.
3081 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3082 (arm_init_builtins): Handle the above.
3083 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3084 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3085 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3088 2018-06-05 Martin Liska <mliska@suse.cz>
3090 * auto-profile.c (read_autofdo_file): Do not use
3091 gcov_ctr_summary struct.
3092 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3093 * coverage.c (struct counts_entry): Likewise.
3094 (read_counts_file): Read just single summary entry.
3095 (get_coverage_counts): Use gcov_summary struct.
3096 * coverage.h (get_coverage_counts): Likewise.
3097 * gcov-dump.c (dump_working_sets): Likewise.
3098 (tag_summary): Dump just single summary.
3099 * gcov-io.c (gcov_write_summary): Write just histogram
3101 (gcov_read_summary): Read just single summary.
3102 (compute_working_sets): Use gcov_summary struct.
3103 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3104 of GCOV_COUNTERS_SUMMABLE.
3105 (GCOV_COUNTERS_SUMMABLE): Remove.
3106 (GCOV_FIRST_VALUE_COUNTER): Replace with
3107 GCOV_COUNTER_V_INTERVAL.
3108 (struct gcov_ctr_summary): Remove.
3109 (struct gcov_summary): Directly use fields of former
3111 (compute_working_sets): Use gcov_summary struct.
3112 * gcov.c (read_count_file): Do not use ctrs fields.
3113 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3115 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3117 * profile.c: Likewise.
3118 * profile.h: Likewise.
3120 2018-06-05 Martin Liska <mliska@suse.cz>
3122 PR gcov-profile/84846
3123 * gcov.c (output_lines): Print working directory only
3124 in intermediate format.
3126 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3128 * config/s390/s390-builtin-types.def: Add void function type.
3129 * config/s390/s390-builtins.def: Use the function type for the
3132 2018-06-04 Jim Wilson <jimw@sifive.com>
3134 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3136 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3137 and maybe_eh_return. Change regno to unsigned int. Use new args to
3138 handle EH_RETURN_DATA_REGNO registers properly.
3139 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3140 (riscv_expand_epilogue): Update comment. Change argument name and
3141 type. Update code to use new name and type. Pass new args to
3142 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3144 * config/riscv/riscv.md (NORMAL_RETURN): New.
3145 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3146 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3147 (eh_return): Call gen_eh_return_internal and emit barrier.
3148 (eh_return_internal): Call riscv_expand_epilogue.
3150 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3152 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3153 bit_insertion field and declare can_be_merged_into method.
3154 (merged_store_group::can_be_merged_into): New method.
3155 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3156 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3157 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3159 2018-06-04 Richard Biener <rguenther@suse.de>
3161 PR tree-optimization/85955
3162 * builtins.c (fold_builtin_sincos): Convert pointers to
3163 destination to appropriate type before dereferencing.
3165 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3167 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3169 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3171 * expr.c (expand_expr_real_1): Force the operand into memory if
3172 its TYPE_MODE is BLKmode and if there is no integer mode for
3173 the number of bits being extracted.
3175 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3179 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3180 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3182 2018-06-04 Richard Biener <rguenther@suse.de>
3184 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3185 (cleanup_tree_cfg_noloop): ... single caller. Do
3186 start_recording_case_labels later.
3188 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3190 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3191 to _IMMINTRIN_H_INCLUDED.
3192 * config/i386/pconfigintrin.h: Ditto.
3193 * config/i386/waitpkgintrin.h: Ditto.
3194 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3195 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3196 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3197 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3198 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3199 waitpkgintrin.h and cldemoteintrin.h.
3201 2018-06-04 Richard Biener <rguenther@suse.de>
3203 PR tree-optimization/86038
3204 * tracer.c (find_best_successor): Check probability for
3205 being initialized, bail out if not.
3207 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3210 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3211 of bits to ignore when comparing architectures.
3213 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3215 PR tree-optimization/69615
3216 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3217 maximum or minimum of the type, try to merge it also as if
3218 range1 is + [-, x - 1] or + [x + 1, -].
3221 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3223 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3225 PR tree-optimization/86034
3226 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3227 the unsigned bitfield type in a bit insertion sequence if it does not
3228 have a larger precision than the bitfield size.
3229 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3231 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3233 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3235 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3237 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3238 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3239 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3240 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3242 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3244 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3245 Disable -fdelete-null-pointer-checks for ELF toolchain.
3247 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3248 Kito Cheng <kito.cheng@gmail.com>
3250 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3251 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3252 (nds32*-*-*): Add glibc and uclibc conditions.
3253 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3254 (TARGET_EXCEPT_UNWIND_INFO): Define.
3255 * config/nds32/elf.h: New file.
3256 * config/nds32/linux.h: New file.
3257 * config/nds32/nds32-elf.opt: New file.
3258 * config/nds32/nds32-linux.opt: New file.
3259 * config/nds32/nds32-fp-as-gp.c
3260 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3261 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3263 (nds32_asm_file_end): Ditto.
3264 (nds32_print_operand): Ditto.
3265 (nds32_insert_attributes): Ditto.
3266 (nds32_init_libfuncs): New function.
3267 (TARGET_HAVE_TLS): Define.
3268 (TARGET_INIT_LIBFUNCS): Define.
3269 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3271 (TARGET_ELF): Apply different mcmodel setting.
3272 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3273 been migrated into elf.h and linux.h files.
3274 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3275 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3276 (mcmodel): The content has been migrated into nds32-elf.opt and
3277 nds32-linux.opt files.
3278 * config/nds32/t-elf: New file.
3279 * config/nds32/t-linux: New file.
3281 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3282 Shiva Chen <shiva0217@gmail.com>
3284 * config/nds32/constants.md (unspec_volatile_element): Add
3285 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3286 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3288 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3289 (make_pass_nds32_fp_as_gp): Declare.
3290 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3292 (nds32_asm_function_end_prologue): Remove unused asm output.
3293 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3294 (nds32_asm_file_start): Output necessary fp_as_gp information.
3295 (nds32_option_override): Adjust register usage.
3296 (nds32_expand_prologue): Consider fp_as_gp situation.
3297 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3298 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3301 (simple_return): Ditto.
3302 (omit_fp_begin): Output special directive for fp_as_gp.
3303 (omit_fp_end): Output special directive for fp_as_gp.
3304 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3305 mforbid-fp-as-gp): New options.
3307 2018-06-01 Mark Wielaard <mark@klomp.org>
3309 * dwarf2out.c (dwarf2out_finish): Remove generation of
3310 DW_AT_loclists_base.
3312 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3314 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3315 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3316 (struct merged_store_group): Add bit_insertion field.
3317 (dump_char_array): Use standard hexadecimal format.
3318 (merged_store_group::merged_store_group): Set bit_insertion to false.
3319 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3320 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3321 also print the mask in the dump file.
3322 (pass_store_merging::gate): Minor tweak.
3323 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3324 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3325 stores with INTEGER_CST stores.
3326 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3327 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3328 and use it throughout. Generate bit insertion sequences if need be.
3329 (pass_store_merging::process_store): Remove redundant condition.
3330 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3332 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3334 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3335 the 128-bit floating point types. Fix function comment.
3337 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3339 * config/aarch64/aarch64-simd.md
3340 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3342 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3345 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3347 PR tree-optimization/85989
3348 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3350 (backprop::intersect_uses): Check it when deciding whether this
3351 is a backedge reference.
3352 (backprop::process_block): Add each phi to m_visited_phis
3353 after visiting it, then clear it at the end.
3355 2018-06-01 Richard Biener <rguenther@suse.de>
3357 * tree-vectorizer.h (vect_dr_stmt): New function.
3358 (vect_get_load_cost): Adjust.
3359 (vect_get_store_cost): Likewise.
3360 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3361 Use vect_dr_stmt instead of DR_SMTT.
3362 (vect_record_base_alignments): Likewise.
3363 (vect_calculate_target_alignment): Likewise.
3364 (vect_compute_data_ref_alignment): Likewise and make static.
3365 (vect_update_misalignment_for_peel): Likewise.
3366 (vect_verify_datarefs_alignment): Likewise.
3367 (vector_alignment_reachable_p): Likewise.
3368 (vect_get_data_access_cost): Likewise. Pass down
3369 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3370 (vect_get_peeling_costs_all_drs): Likewise.
3371 (vect_peeling_hash_get_lowest_cost): Likewise.
3372 (vect_enhance_data_refs_alignment): Likewise.
3373 (vect_find_same_alignment_drs): Likewise.
3374 (vect_analyze_data_refs_alignment): Likewise.
3375 (vect_analyze_group_access_1): Likewise.
3376 (vect_analyze_group_access): Likewise.
3377 (vect_analyze_data_ref_access): Likewise.
3378 (vect_analyze_data_ref_accesses): Likewise.
3379 (vect_vfa_segment_size): Likewise.
3380 (vect_small_gap_p): Likewise.
3381 (vectorizable_with_step_bound_p): Likewise.
3382 (vect_prune_runtime_alias_test_list): Likewise.
3383 (vect_analyze_data_refs): Likewise.
3384 (vect_supportable_dr_alignment): Likewise.
3385 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3386 (vect_gen_prolog_loop_niters): Likewise.
3387 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3388 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3390 (vect_recog_mask_conversion_pattern): Likewise.
3391 (vect_try_gather_scatter_pattern): Likewise.
3392 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3393 to vect_get_store_cost.
3394 (vect_get_store_cost): Get stmt_info instead of DR.
3395 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3396 (vect_get_load_cost): Get stmt_info instead of DR.
3398 2018-06-01 Richard Biener <rguenther@suse.de>
3401 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3402 vars wrapped in MEM_REFs.
3404 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3406 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3407 Fix subreg tests so that we only return a choice between
3408 GENERAL_REGS and FP_REGS if the original classes included both.
3410 2018-06-01 Richard Biener <rguenther@suse.de>
3413 * tree-ssa-structalias.c (get_function_part_constraint):
3414 Handle NULL fi->decl.
3415 (find_func_aliases_for_call): Properly handle indirect
3416 fi from direct call.
3417 (find_func_clobbers): Likewise.
3418 (ipa_pta_execute): Likewise.
3419 (create_variable_info_for): For functions that are ifunc_resolver
3420 resolve to a varinfo that contains the result of the resolver call.
3421 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3424 2018-05-31 Michael Collison <michael.collison@arm.com>
3426 * config/aarch64/aarch64.md:
3427 (*fix_to_zero_extenddfdi2): New pattern.
3428 * gcc.target/aarch64/fix_extend1.c: New testcase.
3430 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3434 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3435 and BUILT_IN_STRNCMP_EQ.
3436 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3437 BUILT_IN_STRNCMP_EQ.
3438 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3439 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3440 (gimple_fold_builtin): Likewise.
3441 * tree-ssa-strlen.c (compute_string_length): New function.
3442 (determine_min_obsize): New function.
3443 (handle_builtin_string_cmp): New function to handle calls to
3444 string compare functions.
3445 (strlen_optimize_stmt): Add handling to builtin string compare
3447 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3448 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3449 * tree.c (build_common_builtin_nodes): Add new defines of
3450 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3452 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3455 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3456 functions with naked attribute.
3458 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3460 * config/i386/sse.md (avx_vec_concat<mode>):
3461 Substitute concat_tg_mode mode attribute with xtg_mode.
3462 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3463 (concat_tg_mode): Remove mode attribute.
3465 2018-05-31 Martin Sebor <msebor@redhat.com>
3468 * calls.c (alloc_max_size): Correct a logic error/typo.
3469 Treat excessive arguments as infinite. Warn for invalid arguments.
3470 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3472 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3475 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3476 and movx for Haswell.
3478 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3479 Cesar Philippidis <cesar@codesourcery.com>
3482 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3483 when emitting error on private/firstprivate reductions.
3484 * omp-low.c (lower_omp_target): Avoid reference-type processing
3485 on pointers for firstprivate clause.
3487 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3489 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3492 * config/aarch64/aarch64-simd.md
3493 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3494 (aarch64_ld1_x3_<mode>): Likewise
3495 (aarch64_st1x2<VALLDIF:mode>): Likewise
3496 (aarch64_st1_x2_<mode>): Likewise
3497 (aarch64_st1x3<VALLDIF:mode>): Likewise
3498 (aarch64_st1_x3_<mode>): Likewise
3499 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3500 (vld1_s8_x3): Likewise.
3501 (vld1_u16_x3): Likewise.
3502 (vld1_s16_x3): Likewise.
3503 (vld1_u32_x3): Likewise.
3504 (vld1_s32_x3): Likewise.
3505 (vld1_u64_x3): Likewise.
3506 (vld1_s64_x3): Likewise.
3507 (vld1_f16_x3): Likewise.
3508 (vld1_f32_x3): Likewise.
3509 (vld1_f64_x3): Likewise.
3510 (vld1_p8_x3): Likewise.
3511 (vld1_p16_x3): Likewise.
3512 (vld1_p64_x3): Likewise.
3513 (vld1q_u8_x3): Likewise.
3514 (vld1q_s8_x3): Likewise.
3515 (vld1q_u16_x3): Likewise.
3516 (vld1q_s16_x3): Likewise.
3517 (vld1q_u32_x3): Likewise.
3518 (vld1q_s32_x3): Likewise.
3519 (vld1q_u64_x3): Likewise.
3520 (vld1q_s64_x3): Likewise.
3521 (vld1q_f16_x3): Likewise.
3522 (vld1q_f32_x3): Likewise.
3523 (vld1q_f64_x3): Likewise.
3524 (vld1q_p8_x3): Likewise.
3525 (vld1q_p16_x3): Likewise.
3526 (vld1q_p64_x3): Likewise.
3527 (vst1_s64_x2): Likewise.
3528 (vst1_u64_x2): Likewise.
3529 (vst1_f64_x2): Likewise.
3530 (vst1_s8_x2): Likewise.
3531 (vst1_p8_x2): Likewise.
3532 (vst1_s16_x2): Likewise.
3533 (vst1_p16_x2): Likewise.
3534 (vst1_s32_x2): Likewise.
3535 (vst1_u8_x2): Likewise.
3536 (vst1_u16_x2): Likewise.
3537 (vst1_u32_x2): Likewise.
3538 (vst1_f16_x2): Likewise.
3539 (vst1_f32_x2): Likewise.
3540 (vst1_p64_x2): Likewise.
3541 (vst1q_s8_x2): Likewise.
3542 (vst1q_p8_x2): Likewise.
3543 (vst1q_s16_x2): Likewise.
3544 (vst1q_p16_x2): Likewise.
3545 (vst1q_s32_x2): Likewise.
3546 (vst1q_s64_x2): Likewise.
3547 (vst1q_u8_x2): Likewise.
3548 (vst1q_u16_x2): Likewise.
3549 (vst1q_u32_x2): Likewise.
3550 (vst1q_u64_x2): Likewise.
3551 (vst1q_f16_x2): Likewise.
3552 (vst1q_f32_x2): Likewise.
3553 (vst1q_f64_x2): Likewise.
3554 (vst1q_p64_x2): Likewise.
3555 (vst1_s64_x3): Likewise.
3556 (vst1_u64_x3): Likewise.
3557 (vst1_f64_x3): Likewise.
3558 (vst1_s8_x3): Likewise.
3559 (vst1_p8_x3): Likewise.
3560 (vst1_s16_x3): Likewise.
3561 (vst1_p16_x3): Likewise.
3562 (vst1_s32_x3): Likewise.
3563 (vst1_u8_x3): Likewise.
3564 (vst1_u16_x3): Likewise.
3565 (vst1_u32_x3): Likewise.
3566 (vst1_f16_x3): Likewise.
3567 (vst1_f32_x3): Likewise.
3568 (vst1_p64_x3): Likewise.
3569 (vst1q_s8_x3): Likewise.
3570 (vst1q_p8_x3): Likewise.
3571 (vst1q_s16_x3): Likewise.
3572 (vst1q_p16_x3): Likewise.
3573 (vst1q_s32_x3): Likewise.
3574 (vst1q_s64_x3): Likewise.
3575 (vst1q_u8_x3): Likewise.
3576 (vst1q_u16_x3): Likewise.
3577 (vst1q_u32_x3): Likewise.
3578 (vst1q_u64_x3): Likewise.
3579 (vst1q_f16_x3): Likewise.
3580 (vst1q_f32_x3): Likewise.
3581 (vst1q_f64_x3): Likewise.
3582 (vst1q_p64_x3): Likewise.
3584 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3586 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3587 user_label_prefix to name.
3589 * tree-core.h: Update comment about the format of NAME string
3590 passed to handler in attribute_spec.
3592 * config/msp430/msp430.md: Remove erroneous subreg expression from
3593 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3594 zero_extend{q,h}isi2.
3596 2018-05-30 Borislav Petkov <bp@suse.de>
3598 * doc/extend.texi: Document some architecture specific
3599 constraints and sort entries.
3601 2018-05-30 Martin Sebor <msebor@redhat.com>
3604 * builtins.c (expand_builtin_stpcpy_1): New function.
3605 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3606 only if the former succeeds.
3608 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3610 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3613 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3615 * doc/invoke.texi (-flinker-output): Document
3617 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3619 * passes.c (ipa_write_summaries): Only modify statements if body
3621 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3622 incrementally linking.
3623 (ipa_passes): Likewise.
3624 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3625 pass down resolution info.
3626 * common.opt (flag_incremental_link): Update info.
3627 * gcc.c (plugin specs): Turn flinker-output=* to
3628 -plugin-opt=-linker-output-known
3629 * toplev.c (compile_file): Also cut compilation when doing incremental
3631 * flag-types. (enum lto_partition_model): Add
3632 LTO_LINKER_OUTPUT_NOLTOREL.
3633 (invoke.texi): Add -flinker-output docs.
3634 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3635 link same way as WPA; do not stream in dead initializers.
3637 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3638 darf2out_register_external_decl): Support incremental link.
3640 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3642 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3644 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3646 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3647 it down to simple_object_copy_lto_debug_sections.
3648 (run_gcc): Determine incremental LTO link time and configure
3649 lto1 into non-wpa mode, disable renaming of debug sections.
3651 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3653 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3654 descriptions of various incorrectly documented functions.
3656 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3659 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3660 address check not strict.
3662 2018-05-30 Richard Biener <rguenther@suse.de>
3664 PR tree-optimization/85964
3665 * tracer.c (better_p): Drop initialized count check, we only
3666 call the function with initialized counts now.
3667 (find_best_successor): Do find a best edge if one
3668 has uninitialized count.
3669 (find_best_predecessor): Likewise. Do BB frequency check only
3670 if count is initialized.
3672 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3674 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3675 (aarch64_ldrstr_offset_compare): New.
3676 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3677 load/store orderings.
3678 (aarch64_gen_adjusted_ldpstp): Likewise.
3680 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3682 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3683 Check for subset of GENERAL_REGS and FP_REGS.
3684 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3687 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3689 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3690 and wi::to_poly_offset. Add the current offset and then check
3691 whether the sum fits, rather than using an unchecked addition of
3692 a checked term. Check for a shwi rather than a uhwi.
3693 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3694 (store_constructor): Use poly_int_tree_p.
3695 (expand_expr_real_1): Likewise.
3696 * function.c (assign_temp): Likewise.
3697 * fold-const.c (const_binop): Use poly_int_tree_p and
3699 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3701 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3702 to_poly_offset for MEM offsets.
3703 * ipa-icf.c (sem_variable::equals): Likewise.
3704 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3705 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3706 wi::to_poly_offset for BIT_FIELD_REF offsets.
3707 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3709 * var-tracking.c (emit_note_insn_var_location): Use
3710 tree_to_poly_uint64.
3712 2018-05-29 Jim Wilson <jimw@sifive.com>
3714 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3716 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3719 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3720 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3722 (sse4_1_round<mode>2): Use nonimmediate_operand
3723 for operand 1 predicate.
3725 2018-05-29 Martin Sebor <msebor@redhat.com>
3726 Richard Biener <rguenther@suse.de>
3729 * calls.c (get_size_range): Call determine_value_range instead
3730 of get_value_range..
3731 * tree-vrp.h (determine_value_range): Declared new function.
3732 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3734 2018-05-29 Richard Biener <rguenther@suse.de>
3736 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3737 sure to use non-pattern stmts for get_earlier_stmt arguments.
3738 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3739 called on pattern stmts.
3740 (get_later_stmt): Likewise.
3742 2018-05-29 Martin Liska <mliska@suse.cz>
3744 PR gcov-profile/85759
3745 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3748 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3750 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3752 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3756 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3757 VEC_PACK_FLOAT_EXPR): New tree codes.
3758 * tree-pretty-print.c (op_code_prio): Handle
3759 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3760 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3761 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3762 * tree-inline.c (estimate_operator_cost): Likewise.
3763 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3764 * fold-const.c (const_binop): Likewise.
3765 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3766 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3767 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3768 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3769 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3770 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3771 * expr.c (expand_expr_real_2): Likewise.
3772 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3773 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3774 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3776 * optabs.c (expand_widen_pattern_expr): For
3777 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3778 sign from result type rather than operand's type.
3779 (expand_binop_directly): For vec_packu_float_optab and
3780 vec_packs_float_optab allow result type to be different from operand's
3782 * optabs-tree.c (optab_for_tree_code): Handle
3783 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3784 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3785 * tree-vect-generic.c (expand_vector_operations_1): Handle
3786 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3787 VEC_PACK_FLOAT_EXPR.
3788 * tree-vect-stmts.c (supportable_widening_operation): Handle
3790 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3791 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3792 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3793 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3794 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3796 (vec_pack<floatprefix>_float_<mode>): New expander.
3797 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3799 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3800 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3801 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3802 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3803 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3805 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3806 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3807 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3808 VEC_PACK_FLOAT_EXPR): Document.
3810 2018-05-29 Richard Biener <rguenther@suse.de>
3812 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3814 (stmt_vec_info_vec): Make pointer.
3815 (init_stmt_vec_info_vec): Remove.
3816 (free_stmt_vec_info_vec): Likewise.
3817 (set_stmt_vec_info_vec): New function.
3818 (free_stmt_vec_infos): Likewise.
3819 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3820 (set_vinfo_for_stmt): Likewise.
3821 (get_earlier_stmt): Likewise.
3822 (get_later_stmt): Likewise.
3823 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3824 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3825 (vec_info::~vec_info): Free stmt_vec_infos.
3826 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3827 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3828 (pass_slp_vectorize::execute): Likewise.
3829 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3830 (free_stmt_vec_info_vec): Likewise.
3831 (set_stmt_vec_info_vec): New function.
3832 (free_stmt_vec_infos): Likewise.
3833 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3834 the global stmt_vec_info_vec.
3835 * tree-parloops.c (gather_scalar_reductions): Use
3836 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3839 2018-05-29 Richard Biener <rguenther@suse.de>
3841 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3843 2018-05-29 Martin Liska <mliska@suse.cz>
3844 David Malcolm <dmalcolm@redhat.com>
3846 * vec.c (test_reverse): New.
3847 (vec_c_tests): Add new test.
3848 * vec.h (vl_ptr>::reverse): New function.
3850 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3852 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3854 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3857 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3859 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3861 2018-05-28 Richard Biener <rguenther@suse.de>
3863 PR tree-optimization/85933
3864 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3865 look at stmts marked as vectorizable.
3867 2018-05-28 Richard Biener <rguenther@suse.de>
3869 PR tree-optimization/85934
3870 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3871 vector boolean check before scalar optimization.
3873 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3875 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3878 2018-05-28 Mark Wielaard <mark@klomp.org>
3880 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3881 if it is an expression containing a minus sign.
3883 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3885 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3887 2018-05-27 Paul Koning <ni1d@arrl.net>
3889 * config/pdp11/pdp11.md (truncsihi2): Remove.
3891 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3892 Chung-Ju Wu <jasonwucj@gmail.com>
3894 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3896 (unaligned_store_dw): Ditto.
3897 * config/nds32/nds32-memory-manipulation.c
3898 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3899 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3900 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3901 (emit_setmem_word_loop): Rename to ...
3902 (emit_setmem_doubleword_loop): ... this.
3903 (nds32_gen_dup_4_byte_to_word_value): New function.
3904 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3905 (nds32_expand_setmem_loop): Refine implementation.
3906 (nds32_expand_setmem_loop_v3m): Ditto.
3907 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3910 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3912 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3914 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3916 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3917 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3918 machine->attr_no_prologue_p.
3919 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3920 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3921 (nds32_expand_epilogue): Consider attr_naked_p.
3922 (nds32_expand_epilogue_v3pop): Likewise.
3923 (nds32_can_use_return_insn): Likewise.
3924 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3925 attr_no_prologue_p fields.
3926 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3928 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3931 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3933 * config/i386/sse.md
3934 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3936 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3938 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3940 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3942 (*<floatsuffix>floatv2div2sf2): Rename to ...
3943 (*float<floatunssuffix>v2div2sf2): ... this.
3944 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3945 (float<floatunssuffix>v2div2sf2_mask): ... this.
3946 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3947 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3948 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3950 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3952 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3954 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3956 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3958 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3960 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3961 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3962 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3963 gen_ufix_truncv8dfv8si2.
3964 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3965 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3966 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3967 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3968 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3969 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3970 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3971 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3973 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3977 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3979 2018-05-25 Jim Wilson <jimw@sifive.com>
3981 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3982 * config/riscv/riscv.c (struct machine_function): Add
3983 interrupt_handler_p and attribute_checked_p fields.
3984 (riscv_attribute_table): Add interrupt.
3985 (riscv_interrupt_type_p): New.
3986 (riscv_save_reg_p): Save extra regs for interrupt handler.
3987 (riscv_use_save_libcall): Return false for interrupt handler.
3988 (riscv_first_stack_step): Add forward declaration.
3989 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3990 for interrupt handler with large frame. Use it for saved reg list.
3991 (riscv_expand_prologue): Move flag_stack_usage_info support to
3992 eliminate duplication.
3993 (riscv_expand_epilogue): Generate mret for interrupt handler.
3994 (riscv_epilogue_uses): New.
3995 (riscv_can_use_return_insn): Return false for interrupt handler.
3996 (riscv_function_ok_for_sibcall): Likewise.
3997 (riscv_set_current_function): Add interrupt handler support.
3998 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3999 * config/riscv/riscv.md (UNSPECV_MRET): New.
4001 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4003 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4005 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4007 PR tree-optimization/85712
4008 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4009 this candidate has already been replaced in-situ by a copy.
4011 2018-05-25 Jason Merrill <jason@redhat.com>
4013 PR c++/80485 - inline function non-zero address.
4014 * symtab.c (nonzero_address): Check DECL_COMDAT.
4016 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4019 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4020 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4021 extension of SImode operation. Use const123_operand predicate.
4022 (*saddsi_1): Remove.
4023 (*saddl_se_1): Ditto.
4025 (*ssubl_se_1): Ditto.
4026 * config/alpha/predicates.md (const123_operand): New predicate.
4027 * config/alpha/constraints.md (P): Use IN_RANGE.
4029 2018-05-25 Richard Biener <rguenther@suse.de>
4031 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4033 (ref_maybe_used_by_stmt_p): Likewise.
4034 (stmt_may_clobber_ref_p): Likewise.
4035 (stmt_may_clobber_ref_p_1): Likewise.
4036 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4038 (ref_maybe_used_by_stmt_p): Likewise.
4039 (stmt_may_clobber_ref_p): Likewise.
4040 (stmt_may_clobber_ref_p_1): Likewise.
4041 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4042 the alias oracle to disambiguate DRs with stmts DR analysis
4044 (vect_analyze_data_refs): Do not give up on not analyzable
4045 DRs for BB vectorization. Remove code truncating the dataref
4048 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4051 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4052 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4053 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4055 2018-05-25 Richard Biener <rguenther@suse.de>
4057 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4058 function, combining stmt data ref gathering and fatal analysis
4060 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4061 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4062 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4063 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4064 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4066 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4068 PR tree-optimization/85720
4069 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4070 SCC if all partitions are builtins.
4071 (version_loop_by_alias_check): New parameter. Generate cancelable
4072 runtime alias check if all partitions are builtins.
4073 (distribute_loop): Update call to above function.
4075 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4077 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4078 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4079 (parm_default_def_partition_arg): Ditto.
4080 (set_parm_default_def_partition): Ditto.
4081 (get_parm_default_def_partitions): Ditto and make it static.
4082 (get_undefined_value_partitions): Ditto and make it static.
4083 (remove_ssa_form): Refactor call to init_var_map here.
4084 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4085 computation for loop region.
4086 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4087 (register_default_def): Delete.
4088 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4089 (parm_default_def_partition_arg): Ditto.
4090 (set_parm_default_def_partition): Ditto.
4091 (get_parm_default_def_partitions): Ditto and make it static.
4092 (get_undefined_value_partitions): Ditto and make it static.
4093 (coalesce_with_default, coalesce_with_default): Update comment.
4094 (create_coalesce_list_for_region): New func factored out from
4095 create_outofssa_var_map.
4096 (populate_coalesce_list_for_outofssa): New func factored out from
4097 create_outofssa_var_map and coalesce_ssa_name.
4098 (create_outofssa_var_map): Delete.
4099 (coalesce_ssa_name): Refactor to support live range computation.
4100 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4101 (get_parm_default_def_partitions): Delete.
4102 (get_undefined_value_partitions): Ditto.
4103 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4104 computation for loop region.
4105 (new_tree_live_info, loe_visit_block): Ditto.
4106 (live_worklist, set_var_live_on_entry): Ditto.
4107 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4108 * tree-ssa-live.h (struct _var_map): New fields.
4109 (init_var_map): Change decl.
4110 (region_contains_p): New.
4112 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4114 * tree-ssa-live.h (live_merge_and_clear): Delete.
4116 2018-05-25 Richard Biener <rguenther@suse.de>
4119 * tree-dump.c (dequeue_and_dump): Remove access to removed
4120 operand 2 of a SWITCH_EXPR.
4122 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4124 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4125 multiplication and division.
4126 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4127 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4128 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4129 (cond_udiv_optab, cond_umod_optab): New optabs.
4130 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4131 (IFN_COND_RDIV): New internal functions.
4132 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4133 TRUNC_MOD_EXPR and RDIV_EXPR.
4134 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4135 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4137 (SVE_INT_BINARY): Include mult.
4138 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4139 (optab, sve_int_op): Handle mult.
4140 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4142 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4143 for SVE_INT_BINARY_SD.
4145 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4147 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4148 (optab, sve_int_op): Handle div and udiv.
4149 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4150 for SVE_INT_BINARY_SD.
4151 (*<optab><mode>3): New insn for the same.
4153 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4155 * tree-vect-patterns.c: Include predict.h.
4156 (vect_recog_divmod_pattern): Restrict check for division support
4157 to when optimizing for size.
4159 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4161 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4162 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4163 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4164 (gimple_match_op::set_op): Likewise.
4165 (gimple_resimplify4): Declare.
4166 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4167 (expr::gen_transform): Likewise.
4168 (decision_tree::gen): Generate a simplification routine for 4 operands.
4169 * gimple-match-head.c (gimple_simplify): Add an overload for
4170 4 operands. In the top-level function, handle up to 4 call
4171 arguments and call gimple_resimplify4.
4172 (gimple_resimplify4): New function.
4173 (build_call_internal): Pass a fourth operand.
4174 (maybe_push_to_seq): Likewise.
4175 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4176 Fold VEC_COND_EXPRs of an operation and a default value into
4177 an IFN_COND_* function if possible.
4178 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4180 (SVE_COND_FP_BINARY): Include them.
4181 (optab, sve_fp_op): Handle them.
4182 (SVE_INT_BINARY_REV): New code iterator.
4183 (SVE_COND_FP_BINARY_REV): New int iterator.
4184 (commutative): New int attribute.
4185 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4187 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4189 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4190 (*cond_<optab><mode>): New patterns for reversed operands.
4192 2018-05-25 Richard Biener <rguenther@suse.de>
4194 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4195 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4196 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4197 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4198 * tree-vect-data-refs.c (everywhere): Adjust users.
4199 * tree-vect-loop.c (everywhere): Likewise.
4200 * tree-vect-slp.c (everywhere): Likewise.
4201 * tree-vect-stmts.c (everywhere): Likewise.
4202 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4204 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4206 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4208 (gcc_cv_as_section_exclude): ... this.
4209 Try Solaris as #exclude syntax.
4210 * configure: Regenerate.
4211 * config.in: Regenerate.
4212 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4214 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4215 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4217 * varasm.c (default_elf_asm_named_section): Don't check if
4218 HAVE_GAS_SECTION_EXCLUDE is defined.
4220 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4222 * doc/md.texi: Update the documentation of the cond_* optabs
4223 to mention the new final operand. Fix GET_MODE_NUNITS call.
4224 Describe the scalar case too.
4225 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4226 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4228 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4229 (get_conditional_internal_fn): Update comment.
4230 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4231 accumulator value as a final argument to conditional functions.
4232 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4233 a define_expand and add an "else" operand. Assert for now that
4234 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4235 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4236 (*cond_<optab><mode>): New patterns.
4237 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4238 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4239 (UNSPEC_COND_EOR): Delete.
4240 (optab): Remove associated mappings.
4241 (SVE_INT_BINARY): New code iterator.
4242 (sve_int_op): Remove int attribute and add "minus" to the code
4244 (SVE_COND_INT_OP): Delete.
4245 (SVE_COND_FP_OP): Rename to...
4246 (SVE_COND_FP_BINARY): ...this.
4248 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4250 * optabs.c (can_reuse_operands_p): New function.
4251 (maybe_legitimize_operands): Try to reuse the results for
4254 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4256 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4257 Add {q} suffix to insn mnemonic.
4259 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4261 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4262 (msp430_warn_func_return): New.
4264 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4266 * fold-const.c (tree_nonzero_bits): New function.
4267 * fold-const.h (tree_nonzero_bits): Likewise.
4268 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4269 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4271 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4275 * varasm.c (assemble_alias): Check ifunc_resolver only on
4278 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4281 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4282 when memory input operand is handled.
4284 2018-05-24 Luis Machado <luis.machado@linaro.org>
4286 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4288 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4290 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4292 * match.pd: Delay FMA folds until after vectorization.
4294 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4297 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4298 address check not strict.
4300 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4302 * gimple-match.h (gimple_match_op): New class.
4303 (mprts_hook): Replace parameters with a gimple_match_op *.
4304 (maybe_build_generic_op): Likewise.
4305 (gimple_simplified_result_is_gimple_val): Replace parameters with
4306 a const gimple_match_op *.
4307 (gimple_simplify): Replace code_helper * and tree * parameters with
4308 a gimple_match_op * parameter.
4309 (gimple_resimplify1): Replace code_helper *, tree and tree *
4310 parameters with a gimple_match_op * parameter.
4311 (gimple_resimplify2): Likewise.
4312 (gimple_resimplify3): Likewise.
4313 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4314 parameters with a gimple_match_op * parameter.
4315 * gimple-match-head.c (gimple_simplify): Change prototypes of
4316 auto-generated functions to take a gimple_match_op * instead of
4317 separate code_helper * and tree * parameters. Make the same
4318 change in the top-level overload and update calls to the
4319 gimple_resimplify routines. Update calls to the auto-generated
4320 functions and to maybe_push_res_to_seq in the publicly-facing
4321 operation-specific gimple_simplify overloads.
4322 (gimple_match_op::MAX_NUM_OPS): Define.
4323 (gimple_resimplify1): Replace rcode and ops with a single res_op
4324 parameter. Update call to gimple_simplify.
4325 (gimple_resimplify2): Likewise.
4326 (gimple_resimplify3): Likewise.
4327 (mprts_hook): Replace parameters with a gimple_match_op *.
4328 (maybe_build_generic_op): Likewise.
4329 (build_call_internal): Replace type, nargs and ops with
4330 a gimple_match_op *.
4331 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4332 with a single gimple_match_op *. Update calls to mprts_hook,
4333 build_call_internal and gimple_simplified_result_is_gimple_val.
4334 Factor out code that is common to the tree_code and combined_fn cases.
4335 * genmatch.c (expr::gen_transform): Replace tem_code and
4336 tem_ops with a gimple_match_op called tem_op. Update calls
4337 to the gimple_resimplify functions and maybe_push_res_to_seq.
4338 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4339 res_ops. Update call to the gimple_resimplify functions.
4340 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4341 (decision_tree::gen): Make the functions take a gimple_match_op *
4342 called res_op instead of separate res_code and res_ops parameters.
4343 Update call accordingly.
4344 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4345 and ops with a single res_op parameter. Update calls to
4346 maybe_build_generic_op and maybe_push_res_to_seq.
4347 (fold_stmt_1): Update calls to gimple_simplify and
4348 replace_stmt_with_simplification.
4349 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4350 and gimple_simplified_result_is_gimple_val.
4351 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4353 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4354 with a gimple_match_op *.
4355 (vn_nary_build_or_lookup): Likewise. Update call to
4356 vn_nary_build_or_lookup_1.
4357 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4358 gimple_match_op *. Update calls to the gimple_resimplify routines
4359 and to gimple_simplified_result_is_gimple_val.
4360 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4361 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4362 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4363 (visit_nary_op): Likewise.
4364 (visit_reference_op_load): Likewise.
4366 2018-05-23 Luis Machado <luis.machado@linaro.org>
4368 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4369 modifier for printing the step amount.
4371 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4374 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4377 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4379 * doc/sourcebuild.texi (Endianness): New subsubsection.
4381 2018-05-23 Luis Machado <luis.machado@linaro.org>
4383 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4384 <prefetch_dynamic_strides>: New const bool field.
4385 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4386 prefetch_dynamic_strides.
4387 (exynosm1_prefetch_tune): Likewise.
4388 (thunderxt88_prefetch_tune): Likewise.
4389 (thunderx_prefetch_tune): Likewise.
4390 (thunderx2t99_prefetch_tune): Likewise.
4391 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4393 (aarch64_override_options_internal): Update to set
4394 PARAM_PREFETCH_DYNAMIC_STRIDES.
4395 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4396 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4397 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4398 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4399 prefetch-dynamic-strides setting.
4401 2018-05-23 Luis Machado <luis.machado@linaro.org>
4403 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4404 <minimum_stride>: New const int field.
4405 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4406 minimum_stride field defaulting to -1.
4407 (exynosm1_prefetch_tune): Likewise.
4408 (thunderxt88_prefetch_tune): Likewise.
4409 (thunderx_prefetch_tune): Likewise.
4410 (thunderx2t99_prefetch_tune): Likewise.
4411 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4412 <default_opt_level>: Set to 3.
4413 (aarch64_override_options_internal): Update to set
4414 PARAM_PREFETCH_MINIMUM_STRIDE.
4415 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4416 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4417 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4418 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4419 stride is constant and is below the minimum stride threshold.
4421 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4423 * config/arm/arm-cpus.in (mode26): Delete.
4424 (armv4): Delete mode26 reference.
4425 * config/arm/arm.c (arm_configure_build_target): Delete use of
4428 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4430 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4432 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4433 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4435 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4436 (floatunsdidf2): Ditto.
4438 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4440 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4441 (fixuns_trunc<mode>si2_avx512f): Ditto.
4442 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4443 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4444 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4446 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4448 PR rtl-optimization/79985
4449 * df-scan.c (df_insn_refs_collect): Remove special case for
4450 global registers and asm statements.
4452 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4454 * extend.texi (Global Register Variables): Rewrite the bullet list.
4455 Note that the register is available for allocation. Note that access
4456 via inline asm must use constraints. Add note about async-signal
4457 handlers. Remove paragraph about automagic register selection.
4459 2018-05-23 Richard Biener <rguenther@suse.de>
4461 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4462 of fixed offset from memset VN.
4464 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4466 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4468 (alloc_cand_and_find_basis): Initialize first_interp field.
4469 (slsr_process_mul): Modify first_interp field.
4470 (slsr_process_add): Likewise.
4471 (slsr_process_cast): Modify first_interp field for each new
4473 (slsr_process_copy): Likewise.
4474 (dump_candidate): Dump first_interp field.
4475 (replace_mult_candidate): Process all interpretations, not just
4477 (replace_rhs_if_not_dup): Likewise.
4478 (replace_one_candidate): Likewise.
4480 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4482 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4484 (aarch64_needs_frame_chain): New function.
4485 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4487 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4490 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4491 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4492 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4493 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4494 as true for strict-align.
4495 (aarch64_can_inline_p): Perform checks even when callee has no
4496 attributes to check for strict alignment.
4497 * doc/extend.texi (AArch64 Function Attributes): Document
4499 * doc/invoke.texi: (AArch64 Options): Likewise.
4501 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4503 PR tree-optimization/85853
4504 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4505 the handling of the root of the node to...
4506 (vect_slp_analyze_node_operations_1): ...this new function,
4507 and run the whole thing with the child nodes' def types
4508 set according to their SLP node's def type.
4510 2018-05-23 Richard Biener <rguenther@suse.de>
4513 * tree-data-ref.c (create_runtime_alias_checks): Defer
4514 and ignore overflow warnings.
4516 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4518 PR tree-optimization/85822
4519 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4522 2018-05-23 Richard Biener <rguenther@suse.de>
4524 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4525 memset constants via native_interpret_expr.
4527 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4530 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4532 (cgraph_node::create_alias): Likewise.
4533 (cgraph_node::get_availability): Check ifunc_resolver instead
4534 of looking up ifunc attribute.
4535 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4536 * varasm.c (do_assemble_alias): Likewise.
4537 (assemble_alias): Likewise.
4538 (default_binds_local_p_3): Likewise.
4539 * cgraph.h (cgraph_node): Add ifunc_resolver.
4540 (cgraph_node::only_called_directly_or_aliased_p): Return false
4542 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4544 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4545 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4546 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4547 instead of looking up ifunc attribute.
4549 2018-05-22 Luis Machado <luis.machado@linaro.org>
4551 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4553 2018-05-22 Martin Sebor <msebor@redhat.com>
4556 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4557 only when expasion succeeds.
4558 (expand_builtin_strcmp): Same.
4559 (expand_builtin_strncmp): Same.
4561 2018-05-22 Martin Sebor <msebor@redhat.com>
4563 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4565 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4566 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4568 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4569 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4571 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4572 Add check for legitimate_address.
4573 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4574 (aarch64_swap_ldrstr_operands): New.
4575 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4578 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4579 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4581 * config/aarch64/aarch64.md: New patterns to generate stp
4583 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4584 single words and double words.
4585 (load_pair_sw, load_pair_dw): Likewise.
4586 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4588 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4590 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4591 for different mode ldpstp and add peephole for merged zero stores.
4593 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4595 (aarch64_gen_store_pair): Rename calls to match new patterns.
4596 (aarch64_gen_load_pair): Rename calls to match new patterns.
4597 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4598 (load_pair<DREG:mode><DREG2:mode>): ... This.
4599 (store_pair<mode>): Rename to...
4600 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4601 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4603 (V_INT_EQUIV): Handle SImode.
4604 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4607 2018-05-22 Martin Sebor <msebor@redhat.com>
4610 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4611 or ajust the presumed bound on an operation to avoid unnecessary
4614 2018-05-22 Martin Sebor <msebor@redhat.com>
4616 PR tree-optimization/85826
4617 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4618 assuming that a DECL necesarily has a constant size.
4620 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4623 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4625 2018-05-22 Richard Biener <rguenther@suse.de>
4627 PR tree-optimization/85834
4628 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4629 non-constant and non-zero memset arguments.
4631 2018-05-22 Martin Liska <mliska@suse.cz>
4634 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4636 2018-05-22 Richard Biener <rguenther@suse.de>
4638 PR tree-optimization/85863
4639 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4640 comparisons when vectype is specified.
4641 (vectorizable_condition): Do not specify vectype for
4642 vect_is_simple_cond when SLP vectorizing.
4644 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4647 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4648 define __ibm128 as long double.
4649 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4650 as a distinct type when IEEE 128-bit support is enabled.
4651 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4652 128-bit types to use the correct functions.
4653 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4654 convert between 128-bit floating point types that have different
4655 modes but the same representation, instead of using gen_lowpart to
4657 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4659 (IFKF_reg): New attributes to give the register constraints for
4661 (extend<mode>tf2_internal): New insns to mark an explicit
4662 conversion between 128-bit floating point types that have a
4663 different mode but share the same representation.
4665 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4667 PR tree-optimization/85814
4668 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4669 a null return from get_strinfo when unsharing the next
4670 strinfo in the chain.
4672 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4675 * varasm.c (weak_finish): Clean up weak_decls.
4677 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4679 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4680 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4681 UNSPEC_UADALP values.
4682 * config/aarch64/iterators.md (ABAL): New int iterator.
4685 (sur): Add mappings for the above.
4686 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4688 (aarch64_<sur>abal<mode>_4): Likewise.
4689 (aarch64_<sur>adalp<mode>_3): Likewise.
4690 (<sur>sadv16qi): New define_expand.
4692 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4694 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4695 (*movdf_internal): Ditto.
4696 (*rcpsf2_sse): Ditto.
4697 (*rsqrtsf2_sse): Ditto.
4698 (*sqrt<mode>2_sse): Ditto.
4700 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4702 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4704 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4705 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4706 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4707 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4709 * config/aarch64/arm_neon.h: Likewise.
4710 * config/aarch64/iterators.md (VQ_I): New.
4712 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4714 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4716 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4719 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4721 * config/nds32/constraints.md (S): New constraint.
4722 * config/nds32/nds32.md (call_internal): Use constraint S.
4723 (call_value_internal): Likewise.
4724 (sibcall_internal): Likewise.
4725 (sibcall_value_internal): Likewise.
4727 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4728 Chung-Ju Wu <jasonwucj@gmail.com>
4730 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4733 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4734 Chung-Ju Wu <jasonwucj@gmail.com>
4736 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4737 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4738 (nds32_rtx_costs_impl): Simplify.
4739 (nds32_address_cost_impl): Simplify.
4740 (nds32_init_rtx_costs): New function.
4741 (nds32_rtx_costs_speed_prefer): Likewise.
4742 (nds32_rtx_costs_size_prefer): Likewise.
4743 (nds32_address_cost_speed_prefer): Likewise.
4744 (nds32_address_cost_speed_fwprop): Likewise.
4745 (nds32_address_cost_size_prefer): Likewise.
4746 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4747 * config/nds32/nds32.c (nds32_option_override): Use
4748 nds32_init_rtx_costs function.
4750 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4752 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4753 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4754 (TARGET_PIPELINE_N8): Likewise.
4755 (TARGET_PIPELINE_N10): Likewise.
4756 (TARGET_PIPELINE_N13): Likewise.
4757 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4759 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4761 * config/nds32/nds32-fpu.md: Update copyright year.
4763 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4765 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4767 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4769 * config/nds32/nds32.c
4770 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4771 * config/nds32/nds32.opt (minline-asm-r15): New option.
4773 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4775 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4777 * config/nds32/nds32.md (abssi2): New pattern.
4779 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4781 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4782 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4783 Merge insn pattern from sse_cvtsi2ss<round_name> and
4784 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4785 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4786 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4787 using SWI48 mode iterator.
4788 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4789 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4790 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4791 pattern from sse_cvttss2si<round_saeonly_name>
4792 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4793 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4794 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4795 using SWI48 mode iterator.
4796 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4797 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4798 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4799 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4800 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4801 using SWI48 mode iterator.
4802 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4803 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4804 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4805 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4806 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4807 SWI48 mode iterator.
4808 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4809 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4810 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4811 pattern from sse_cvttsd2si<round_saeonly_name>
4812 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4814 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32-md-auxiliary.c
4817 (nds32_valid_smw_lwm_base_p): Refine.
4818 (nds32_output_smw_single_word): Refine.
4819 (nds32_output_smw_double_word): New.
4820 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4822 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4824 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4825 (nds32_output_stack_pop): Refine.
4826 (nds32_expand_unaligned_load): Refine.
4827 (nds32_expand_unaligned_store): Refine.
4829 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4830 Chung-Ju Wu <jasonwucj@gmail.com>
4832 * config/nds32/constants.md: Add TP_REGNUM constant.
4833 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4834 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4836 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4837 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4838 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4839 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4840 and PIC code generation.
4841 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4843 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4845 * config/nds32/nds32.md: Support TLS and PIC.
4846 * config/nds32/nds32.c: Support TLS and PIC.
4847 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4848 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4851 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4853 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4854 mode with E_ prefix.
4856 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4857 Chung-Ju Wu <jasonwucj@gmail.com>
4859 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4860 * config/nds32/nds32-md-auxiliary.c
4861 (symbolic_reference_mentioned_p): New.
4862 (nds32_legitimize_ict_address): New.
4863 (nds32_expand_ict_move): New.
4864 (nds32_indirect_call_referenced_p): New.
4865 (nds32_symbol_binds_local_p): Delete.
4866 (nds32_long_call_p): Modify.
4867 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4868 * config/nds32/nds32-protos.h
4869 (symbolic_reference_mentioned_p): Declare.
4870 (nds32_legitimize_ict_address): Declare.
4871 (nds32_expand_ict_move): Declare.
4872 (nds32_indirect_call_referenced_p): Declare.
4873 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4874 (nds32_relax_group): Use nds32_ict_const_p as condition.
4875 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4876 (nds32_asm_file_start): Output ict_model directive in asm code.
4877 (nds32_legitimate_address_p): Consider indirect call.
4878 (nds32_print_operand): Consider indirect call.
4879 (nds32_print_operand_address): Consider indirect call.
4880 (nds32_insert_attributes): Handle "indirect_call" attribute.
4881 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4882 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4883 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4884 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4885 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4886 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4887 (TARGET_ICT_MODEL_SMALL): Define.
4888 (TARGET_ICT_MODEL_LARGE): Define.
4889 * config/nds32/nds32.md (movsi): Consider ict model.
4890 (call, call_value): Consider ict model.
4891 (sibcall, sibcall_value): Consider ict model.
4892 * config/nds32/nds32.opt (mict-model): New option.
4893 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4896 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4897 Monk Chiang <sh.chiang04@gmail.com>
4898 Jim Wilson <jimw@sifive.com>
4900 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4901 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4902 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4903 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4904 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4905 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4906 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4907 compute save_libcall_adjustment properly.
4908 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4909 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4910 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4911 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4912 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4913 (ABI_SPEC): Handle mabi=ilp32e.
4914 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4915 (RVE): Add RVE mask.
4916 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4917 <-march>: Add rv32e as an example.
4919 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4922 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4923 (intra_create_variable_infos): Handle C++ constructors.
4925 2018-05-18 Martin Liska <mliska@suse.cz>
4927 * passes.def: Remove a redundant pass.
4929 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4932 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4934 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4936 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4938 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4940 (armv2, armv2a, armv3, armv3m): Delete architectures.
4941 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4942 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4943 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4945 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4946 (*mulsidi3adddi): Likewise.
4947 (mulsidi3): Likewise.
4948 (*mulsidi3_nov6): Likewise.
4949 (umulsidi3): Likewise.
4950 (umulsidi3_nov6): Likewise.
4951 (umaddsidi4): Likewise.
4952 (*umulsidi3adddi): Likewise.
4953 (smulsi3_highpart): Likewise.
4954 (*smulsi3_highpart_nov6): Likewise.
4955 (umulsi3_highpart): Likewise.
4956 (*umulsi3_highpart_nov6): Likewise.
4957 * config/arm/arm.h (arm_arch3m): Delete.
4958 * config/arm/arm.c (arm_arch3m): Delete.
4959 (arm_option_override_internal): Update armv3-related comment.
4960 (arm_configure_build_target): Delete use of isa_bit_mode32.
4961 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4962 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4963 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4965 (mulusa3): Likewise.
4966 * config/arm/arm-protos.h (arm_arch3m): Delete.
4967 * config/arm/arm-tables.opt: Regenerate.
4968 * config/arm/arm-tune.md: Likewise.
4969 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4970 deleted architectures.
4972 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4974 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4975 (armv5t, armv5te): New features.
4976 (ARMv5, ARMv5e): Delete fgroups.
4977 (ARMv5t, ARMv5te): Adjust for above changes.
4979 (armv5, armv5e): Delete arches.
4980 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4982 (*call_reg_arm): Likewise.
4983 (*call_value_reg_armv5): Likewise.
4984 (*call_value_reg_arm): Likewise.
4985 (*call_symbol): Likewise.
4986 (*call_value_symbol): Likewise.
4987 (*sibcall_insn): Likewise.
4988 (*sibcall_value_insn): Likewise.
4990 (prefetch): Likewise.
4991 (define_split and define_peephole2 dependent on arm_arch5):
4993 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4995 (TARGET_ARM_QBIT): Likewise.
4996 (TARGET_DSP_MULTIPLY): Likewise.
4997 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4998 (arm_arch5, arm_arch5e): Delete.
4999 (arm_arch5t, arm_arch5te): Declare.
5000 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5001 (arm_arch5t): Declare.
5002 (arm_option_reconfigure_globals): Update for the above.
5003 (arm_options_perform_arch_sanity_checks): Update comment, replace
5004 use of arm_arch5 with arm_arch5t.
5005 (use_return_insn): Likewise.
5006 (arm_emit_call_insn): Likewise.
5007 (output_return_instruction): Likewise.
5008 (arm_final_prescan_insn): Likewise.
5009 (arm_coproc_builtin_available): Likewise.
5010 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5011 arm_arch5e with arm_arch5t and arm_arch5te.
5012 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5013 (arm_arch5t, arm_arch5te): Declare.
5014 * config/arm/arm-tables.opt: Regenerate.
5015 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5016 * config/arm/t-multilib: Likewise.
5017 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5018 instead of arm_arch5.
5019 (*call_reg_thumb1): Likewise.
5020 (*call_value_reg_thumb1_v5): Likewise.
5021 (*call_value_reg_thumb1): Likewise.
5022 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5024 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5026 2018-05-18 Martin Liska <mliska@suse.cz>
5028 PR gcov-profile/84846
5029 * doc/gcov.texi: Document -t option of gcov tool.
5031 2018-05-18 Martin Liska <mliska@suse.cz>
5033 PR gcov-profile/84846
5034 * gcov.c (print_usage): Add new -t option.
5035 (process_args): Handle the option.
5036 (generate_results): Use stdout as output when requested by
5039 2018-05-18 Martin Liska <mliska@suse.cz>
5041 PR gcov-profile/84846
5042 * coverage.c (coverage_init): Write PWD to .gcno file.
5043 * doc/gcov.texi: Document how working directory is printed.
5044 * gcov-dump.c (dump_gcov_file): Print PWD.
5045 * gcov.c (output_intermediate_file): Likewise.
5046 (read_graph_file): Read PWD string.
5047 (output_lines): Print PWD.
5049 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5052 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5053 for retval and return false if all args to phi are zero.
5055 2018-05-18 Richard Biener <rguenther@suse.de>
5057 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5059 (evrp_dom_walker::before_dom_children): Call it.
5061 2018-05-18 Richard Biener <rguenther@suse.de>
5063 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5064 results when processing array refs with variable index.
5066 2018-05-18 Toon Moene <toon@moene.org>
5068 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5069 directly after that of -floop-interchange. Indicate that both
5070 options are enabled by default when specifying -O3.
5072 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5074 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5075 iterator. Delete separate integer-mode vec_set<mode> expander.
5076 (aarch64_simd_vec_setv2di): Delete.
5077 (vec_setv2di): Delete.
5078 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5079 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5080 the "w, r" alternative.
5082 2018-05-18 Martin Liska <mliska@suse.cz>
5084 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5085 * tree-pass.h (make_pass_lower_switch_O0): New function.
5086 * tree-switch-conversion.c (node_has_low_bound): Remove.
5087 (node_has_high_bound): Likewise.
5088 (node_is_bounded): Likewise.
5089 (class pass_lower_switch): Make it a template type and create
5091 (pass_lower_switch::execute): Add template argument.
5092 (make_pass_lower_switch): New function.
5093 (make_pass_lower_switch_O0): New function.
5094 (do_jump_if_equal): Remove.
5095 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5096 all the hard work to tree optimization passes.
5098 2018-05-18 Martin Liska <mliska@suse.cz>
5100 * dbgcnt.c (limit_low): Renamed from limit.
5101 (limit_high): New variable.
5102 (dbg_cnt_is_enabled): Check for upper limit.
5103 (dbg_cnt): Adjust dumping.
5104 (dbg_cnt_set_limit_by_index): Add new argument for high
5106 (dbg_cnt_set_limit_by_name): Likewise.
5107 (dbg_cnt_process_single_pair): Parse new format.
5108 (dbg_cnt_process_opt): Use strtok.
5109 (dbg_cnt_list_all_counters): Remove 'value' and add
5111 * doc/invoke.texi: Document changes.
5113 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5115 * doc/sourcebuild.texi (scalar_all_fma): Document.
5116 * tree.def (FMA_EXPR): Delete.
5117 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5118 * internal-fn.c (ternary_direct): New macro.
5119 (expand_ternary_optab_fn): Likewise.
5120 (direct_ternary_optab_supported_p): Likewise.
5121 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5122 * builtins.c (fold_builtin_fma): Delete.
5123 (fold_builtin_3): Don't call it.
5124 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5125 * expr.c (expand_expr_real_2): Likewise.
5126 * fold-const.c (operand_equal_p): Likewise.
5127 (fold_ternary_loc): Likewise.
5128 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5129 * gimple.c (DEFTREECODE): Likewise.
5130 * gimplify.c (gimplify_expr): Likewise.
5131 * optabs-tree.c (optab_for_tree_code): Likewise.
5132 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5133 * tree-eh.c (operation_could_trap_p): Likewise.
5134 (stmt_could_throw_1_p): Likewise.
5135 * tree-inline.c (estimate_operator_cost): Likewise.
5136 * tree-pretty-print.c (dump_generic_node): Likewise.
5137 (op_code_prio): Likewise.
5138 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5139 * tree-ssa-operands.c (get_expr_operands): Likewise.
5140 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5141 * fold-const-call.h (fold_fma): Delete.
5142 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5143 CFN_FNMA and CFN_FNMS.
5145 * genmatch.c (combined_fn): New enum.
5146 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5147 (commutative_op): New function.
5148 (commutate): Use it. Handle more than 2 operands.
5149 (dt_operand::gen_gimple_expr): Use commutative_op.
5150 (parser::parse_expr): Allow :c to be used with non-binary
5151 operators if the commutative operand is known.
5152 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5153 CFN_FMS, CFN_FNMA and CFN_FNMS.
5154 (backprop::process_assign_use): Remove FMA_EXPR handling.
5155 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5156 (gen_hsa_fma): New function.
5157 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5158 IFN_FNMA and IFN_FNMS.
5159 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5160 * gimple-fold.h (follow_all_ssa_edges): Declare.
5161 * gimple-fold.c (follow_all_ssa_edges): New function.
5162 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5163 gimple_build interface and use follow_all_ssa_edges to fold the result.
5164 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5165 instead of checking for optabs directly.
5166 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5167 rather than FMA_EXPRs.
5168 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5169 call to IFN_FMA instead of an FMA_EXPR.
5171 2018-05-17 Jim Wilson <jimw@sifive.com>
5173 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5174 sign extended promoted subreg, and the range does not have the sign bit
5175 set, then do a sign extend.
5177 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5178 test, check for sign extended subreg and/or constant operands, and
5179 do a sign extend in that case.
5181 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5183 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5184 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5186 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5187 Change logics_shift_reg to logics_shift_imm.
5188 (thunderx2t99_fp_loadpair_basic): Delete.
5189 (thunderx2t99_fp_storepair_basic): Delete.
5190 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5191 (thunderx2t99_asimd_polynomial): Delete.
5192 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5193 and neon_fp_mul_d_scalar_q.
5194 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5195 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5196 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5197 (thunderx2t99_asimd_lut): Add missing tbl types.
5198 (thunderx2t99_asimd_ext): Delete.
5199 (thunderx2t99_asimd_load1_1_mult): Delete.
5200 (thunderx2t99_asimd_load1_2_mult): Delete.
5201 (thunderx2t99_asimd_load1_ldp): New.
5202 (thunderx2t99_asimd_load1): New.
5203 (thunderx2t99_asimd_load2): Add missing *load2* types.
5204 (thunderx2t99_asimd_load3): New.
5205 (thunderx2t99_asimd_load4): New.
5206 (thunderx2t99_asimd_store1_1_mult): Delete.
5207 (thunderx2t99_asimd_store1_2_mult): Delete.
5208 (thunderx2t99_asimd_store2_mult): Delete.
5209 (thunderx2t99_asimd_store2_onelane): Delete.
5210 (thunderx2t99_asimd_store_stp): New.
5211 (thunderx2t99_asimd_store1): New.
5212 (thunderx2t99_asimd_store2): New.
5213 (thunderx2t99_asimd_store3): New.
5214 (thunderx2t99_asimd_store4): New.
5216 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5218 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5219 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5221 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5222 Segher Boessenkool <segher@kernel.crashing.org>
5225 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5228 2018-05-17 Richard Biener <rguenther@suse.de>
5230 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5231 for pruning loop and prune defs feeding only already visited PHIs.
5233 2018-05-17 Richard Biener <rguenther@suse.de>
5235 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5237 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5238 Richard Biener <rguenther@suse.de>
5240 PR tree-optimization/85793
5241 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5242 for VMAT_ELEMENTWISE.
5244 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5246 * internal-fn.h (lookup_internal_fn): Declare
5247 * internal-fn.c (lookup_internal_fn): New function.
5248 * gimple.c (gimple_build_call_from_tree): Handle calls to
5250 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5251 internal function names.
5252 * tree-pretty-print.c (dump_generic_node): Likewise.
5253 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5255 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5257 * gimple-fold.h (gimple_build): Make the function forms take
5258 combined_fn rather than built_in_function.
5259 (gimple_simplify): Likewise.
5260 * gimple-match-head.c (gimple_simplify): Likewise.
5261 * gimple-fold.c (gimple_build): Likewise.
5262 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5263 rather than gimple_build_call_internal.
5264 (get_initial_defs_for_reduction): Likewise.
5265 (vect_create_epilog_for_reduction): Likewise.
5266 (vectorizable_live_operation): Likewise.
5268 2018-05-17 Martin Liska <mliska@suse.cz>
5270 * gimple-ssa-sprintf.c (format_directive): Do not use
5271 space in between 'G_' and '('.
5273 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5276 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5277 even if the mask is not all ones.
5280 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5282 (ix86_gimple_fold_builtin): Likewise.
5285 * config/i386/i386.c: Include tree-vector-builder.h.
5286 (ix86_vector_shift_count): New function.
5287 (ix86_fold_builtin): Fold shift builtins by scalar count.
5288 (ix86_gimple_fold_builtin): Likewise.
5290 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5291 _mm512_setzero): New intrinsics.
5293 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5294 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5296 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5297 code generation for cases where splatting a value is not useful.
5298 * simplify-rtx.c (simplify_ternary_operation): Simplify
5299 vec_merge across a vec_duplicate and a paradoxical subreg forming
5300 a vector mode to a vec_concat.
5302 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5304 * config.gcc: Support "goldmont-plus".
5305 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5307 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5308 PROCESSOR_GOLDMONT_PLUS.
5309 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5310 (processor_target_table): Add "goldmont-plus".
5311 (PTA_GOLDMONT_PLUS): Define.
5312 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5313 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5314 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5315 (fold_builtin_cpu): Add "goldmont-plus".
5316 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5317 (ix86_option_override_internal): Add "goldmont-plus".
5318 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5319 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5320 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5321 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5323 2018-05-17 Richard Biener <rguenther@suse.de>
5325 PR tree-optimization/85757
5326 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5327 remove defs that only feed that PHI from further processing.
5329 2018-05-16 Jim Wilson <jimw@sifive.com>
5331 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5333 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5335 2018-05-16 Mark Wielaard <mark@klomp.org>
5337 * dwarf2out.c (count_index_strings): New function.
5338 (output_indirect_strings): Call count_index_strings and generate
5339 header for dwarf_version >= 5.
5341 2018-05-16 Mark Wielaard <mark@klomp.org>
5343 * dwarf2out.c (dwarf_FORM): New function.
5344 (set_indirect_string): Use dwarf_FORM.
5345 (reset_indirect_string): Likewise.
5346 (size_of_die): Likewise.
5347 (value_format): Likewise.
5348 (output_die): Likewise.
5349 (add_skeleton_AT_string): Likewise.
5350 (output_macinfo_op): Likewise.
5351 (index_string): Likewise.
5352 (output_index_string_offset): Likewise.
5353 (output_index_string): Likewise.
5354 (count_index_strings): Likewise.
5356 2018-05-16 Carl Love <cel@us.ibm.com>
5358 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5359 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5361 2018-05-16 Martin Jambor <mjambor@suse.cz>
5363 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5364 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5366 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5368 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5369 (fnma<mode>4): Likewise.
5370 (fms<mode>4): Likewise.
5371 (fnms<mode>4): Likewise.
5372 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5373 (aarch64_fnma<mode>4): Likewise.
5374 (aarch64_fms<mode>4): Likewise.
5375 (aarch64_fnms<mode>4): Likewise.
5376 (aarch64_fnmadd<mode>4): Likewise.
5378 2018-05-16 Jason Merrill <jason@redhat.com>
5380 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5382 2018-05-16 Richard Biener <rguenther@suse.de>
5384 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5385 (dump_stmt_cost): Declare.
5386 (add_stmt_cost): Dump cost we add.
5387 (add_stmt_costs): New function.
5388 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5390 (vect_analyze_stmt): Adjust prototype.
5391 (vectorizable_condition): Likewise.
5392 (vectorizable_live_operation): Likewise.
5393 (vectorizable_reduction): Likewise.
5394 (vectorizable_induction): Likewise.
5395 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5396 cost vector to pass to vectorizable_ and record afterwards.
5397 (vect_model_reduction_cost): Take cost vector argument and adjust.
5398 (vect_model_induction_cost): Likewise.
5399 (vectorizable_reduction): Likewise.
5400 (vectorizable_induction): Likewise.
5401 (vectorizable_live_operation): Likewise.
5402 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5403 SLP_TREE_NUMBER_OF_VEC_STMTS.
5404 (vect_analyze_slp_cost_1): Remove.
5405 (vect_analyze_slp_cost): Likewise.
5406 (vect_slp_analyze_node_operations): Take visited args and
5407 a target cost vector. Avoid processing already visited stmt sets.
5408 (vect_slp_analyze_operations): Use a local cost vector to gather
5409 costs and register those of non-discarded instances.
5410 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5411 (vect_schedule_slp_instance): Remove copying of
5412 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5414 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5415 adding cost. Record cost entry location.
5416 (vect_prologue_cost_for_slp_op): Function to compute cost of
5417 a constant or invariant generated for SLP vect in the prologue,
5418 split out from vect_analyze_slp_cost_1.
5419 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5420 (vect_model_promotion_demotion_cost): Likewise.
5421 (vect_model_store_cost): Likewise, make static.
5422 (vect_model_load_cost): Likewise.
5423 (vectorizable_bswap): Add cost vector arg and adjust.
5424 (vectorizable_call): Likewise.
5425 (vectorizable_simd_clone_call): Likewise.
5426 (vectorizable_conversion): Likewise.
5427 (vectorizable_assignment): Likewise.
5428 (vectorizable_shift): Likewise.
5429 (vectorizable_operation): Likewise.
5430 (vectorizable_store): Likewise.
5431 (vectorizable_load): Likewise.
5432 (vectorizable_condition): Likewise.
5433 (vectorizable_comparison): Likewise.
5434 (can_vectorize_live_stmts): Likewise.
5435 (vect_analyze_stmt): Likewise.
5436 (vect_transform_stmt): Adjust calls to vectorizable_*.
5437 * tree-vectorizer.c: Include gimple-pretty-print.h.
5438 (dump_stmt_cost): New function.
5440 2018-05-16 Richard Biener <rguenther@suse.de>
5442 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5443 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5444 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5445 (check_name): New callback.
5446 (dse_classify_store): Track cycles via a visited bitmap of PHI
5447 defs and simplify handling of in-loop and across loop dead stores
5448 and properly fail for loop-variant refs. Handle byte-tracking with
5449 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5452 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5454 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5455 (vect_get_mask_type_for_stmt): Likewise.
5456 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5458 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5459 to determine the statement's vector type and the vector type that
5460 should be used for calculating nunits. Deal with cases in which
5461 the type has to be deferred.
5462 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5463 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5464 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5465 (vect_determine_vf_for_stmt): New functions, split out from...
5466 (vect_determine_vectorization_factor): ...here.
5467 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5468 (vect_get_mask_type_for_stmt): New functions, split out from
5469 vect_determine_vectorization_factor.
5471 2018-05-16 Richard Biener <rguenther@suse.de>
5473 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5474 verify the [VEC_]COND_EXPR embedded comparison.
5476 2018-05-15 Martin Sebor <msebor@redhat.com>
5478 PR tree-optimization/85753
5479 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5480 RECORD_TYPE in addition to ARRAY_TYPE.
5482 2018-05-15 Martin Sebor <msebor@redhat.com>
5485 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5487 2018-05-15 Richard Biener <rguenther@suse.de>
5489 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5490 add by_clobber_p one. Change algorithm to collect all defs
5491 representing uses we need to walk and try reducing them to
5492 a single one before failing.
5493 (dse_dom_walker::dse_optimize_stmt): Adjust.
5495 2018-05-13 Mark Wielaard <mark@klomp.org>
5497 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5498 (size_of_loc_descr): Likewise.
5499 (output_loc_operands): Likewise.
5500 (output_loc_operands_raw): Likewise.
5501 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5502 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5503 (hash_loc_operands): Likewise.
5504 (compare_loc_operands): Likewise.
5506 2018-05-14 Mark Wielaard <mark@klomp.org>
5508 * dwarf2out.c (count_index_addrs): New function.
5509 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5511 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5513 PR tree-optimization/83648
5514 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5515 return value as malloc candidate.
5517 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5520 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5521 param as true in call to suggest_attribute.
5523 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5525 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5526 -mreadonly-in-sdata.
5528 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5530 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5532 (aarch64_crypto_aesd_fused): Likewise.
5534 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5536 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5537 (movsi_aarch64): Likewise.
5538 (load_pairsi): Likewise.
5539 (load_pairdi): Likewise.
5540 (store_pairsi): Likewise.
5541 (store_pairdi): Likewise.
5542 (load_pairsf): Likewise.
5543 (load_pairdf): Likewise.
5544 (store_pairsf): Likewise.
5545 (store_pairdf): Likewise.
5546 (zero_extend): Likewise.
5547 (trunc): Swap alternatives.
5548 (fcvt_target): Add '?' to prefer w over r.
5550 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5553 * config/i386/i386.md: Disallow non-commutative arithmetics in
5554 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5555 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5556 in the peephole2 before it.
5558 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5560 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5561 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5562 (ix86_handle_option): Handle -mcldemote.
5563 * config.gcc: New header.
5564 * config/i386/cldemoteintrin.h: New file.
5565 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5566 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5568 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5569 OPTION_MASK_ISA_CLDEMOTE.
5570 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5571 (ix86_valid_target_attribute_inner_p): Ditto.
5572 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5573 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5574 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5575 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5576 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5578 * config/i386/i386.opt: Add -mcldemote.
5579 * config/i386/x86intrin.h: New header.
5580 * doc/invoke.texi: Add -mcldemote.
5582 2018-05-14 Richard Biener <rguenther@suse.de>
5584 * doc/match-and-simplify.texi: Adjust :s documentation.
5586 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5588 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5589 intended memcpy size.
5590 (REORDER_45): Likewise.
5592 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5594 * sort.cc: New file.
5595 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5596 * vec.c (qsort_chk): Use gcc_qsort.
5597 * Makefile.in (OBJS-libcommon): Add sort.o.
5598 (build/sort.o): New target. Use it...
5599 (BUILD_RTL): ... here, and...
5600 (build/gencfn-macros): ... here, and...
5601 (build/genmatch): ... here.
5603 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5604 Chung-Ju Wu <jasonwucj@gmail.com>
5606 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5607 * config/nds32/nds32-graywolf.md: New file.
5608 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5609 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5611 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5612 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5613 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5614 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5615 * config/nds32/pipelines.md: Include n15 settings.
5617 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5618 Chung-Ju Wu <jasonwucj@gmail.com>
5620 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5621 * config/nds32/nds32-n13.md: New file.
5622 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5623 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5625 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5626 * config/nds32/nds32.md (pipeline_model): Add n13.
5627 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5628 * config/nds32/pipelines.md: Include n13 settings.
5630 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5631 Chung-Ju Wu <jasonwucj@gmail.com>
5633 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5634 * config/nds32/nds32-n10.md: New file.
5635 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5636 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5638 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5639 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5640 * config/nds32/nds32.md (pipeline_model): Add n10.
5641 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5642 * config/nds32/pipelines.md: Include n10 settings.
5644 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5645 Kito Cheng <kito.cheng@gmail.com>
5646 Chung-Ju Wu <jasonwucj@gmail.com>
5648 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5649 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5650 Add enum values for DSP extension instructions.
5651 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5653 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5654 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5656 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5657 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5658 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5659 * config/nds32/nds32-intrinsic.md: Likewise.
5660 * config/nds32/nds32_intrinsic.h: Likewise.
5661 * config/nds32/nds32-md-auxiliary.c: Likewise.
5662 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5663 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5664 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5665 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5666 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5667 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5668 TYPE_DMAC in switch statement.
5669 * config/nds32/nds32.c: New checking and implementation for DSP
5670 extension instructions.
5671 * config/nds32/nds32.h: Likewise.
5672 * config/nds32/nds32.md: Likewise.
5673 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5674 * config/nds32/predicates.md: Implement new predicates for DSP
5677 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5679 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5680 Reformat alternatives and attributes so it is easier to identify
5681 which constraints/attributes go with which instruction.
5682 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5683 (mov<mode>_softfloat32, FMOVE64): Likewise.
5684 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5685 (mov<mode>_softfloat64, FMOVE64): Likewise.
5687 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5689 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5691 (Basic PowerPC Built-in Functions): The new name of the
5692 subsection previously known as "PowerPC Built-in Functions".
5693 (Basic PowerPC Built-in Functions Available on all Configurations):
5695 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5696 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5697 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5698 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5700 2018-05-11 Martin Jambor <mjambor@suse.cz>
5703 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5706 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5709 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5711 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5713 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5714 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5715 (ix86_handle_option): Handle -mwaitpkg.
5716 * config.gcc: New header.
5717 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5718 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5719 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5721 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5722 OPTION_MASK_ISA_WAITPKG.
5723 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5724 (ix86_option_override_internal): Add PTA_WAITPKG.
5725 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5726 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5727 IX86_BUILTIN_TPAUSE.
5728 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5729 __builtin_ia32_umwait and __builtin_ia32_tpause.
5730 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5731 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5732 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5733 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5734 UNSPECV_TPAUSE): New.
5735 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5736 * config/i386/i386.opt: Add -mwaitpkg.
5737 * config/i386/waitpkgintrin.h: New file.
5738 * config/i386/x86intrin.h: New header.
5739 * doc/invoke.texi: Add -mwaitpkg.
5741 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5744 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5746 (cortex-m0): Use armv6s-m isa.
5747 (cortex-m0plus): Likewise.
5748 (cortex-m1): Likewise.
5749 (cortex-m0.small-multiply): Likewise.
5750 (cortex-m0plus.small-multiply): Likewise.
5751 (cortex-m1.small-multiply): Likewise.
5753 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5754 Jakub Jelinek <jakub@redhat.com>
5756 PR tree-optimization/85692
5757 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5758 source permute as well.
5760 2018-05-11 Martin Liska <mliska@suse.cz>
5763 * doc/extend.texi: Document LLVM style format for no_sanitize
5766 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5768 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5769 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5770 (mode_supports_vsx_dform_quad): Likewise.
5771 (mode_supports_vmx_dform): Move these functions to be next to the
5772 other mode_supports functions.
5773 (mode_supports_dq_form): Likewise.
5774 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5775 mode_supports_dq_form.
5776 (reg_offset_addressing_ok_p): Likewise.
5777 (offsettable_ok_by_alignment): Likewise.
5778 (rs6000_legitimate_offset_address_p): Likewise.
5779 (legitimate_lo_sum_address_p): Likewise.
5780 (rs6000_legitimize_address): Likewise.
5781 (rs6000_legitimize_reload_address): Likewise.
5782 (rs6000_secondary_reload_inner): Likewise.
5783 (rs6000_preferred_reload_class): Likewise.
5784 (rs6000_output_move_128bit): Likewise.
5786 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5788 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5789 Generate SImode target register for null target.
5790 <case IX86_BUILTIN_XGETBV>: Ditto.
5791 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5792 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5794 2018-05-10 Carl Love <cel@us.ibm.com>
5796 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5797 dcbtt and dcbtstt if operands[2] is 0.
5799 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5802 * config/i386/sse.md (usadv64qi): New expander.
5804 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5806 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5807 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5808 -maltivec=be support.
5809 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5810 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5811 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5812 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5813 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5814 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5815 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5816 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5817 altivec_vsumsws): Adjust.
5818 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5819 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5820 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5822 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5823 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5824 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5825 (altivec_lve<VI_char>x): Delete expand.
5826 (*altivec_lve<VI_char>x_internal): Rename to...
5827 (altivec_lve<VI_char>x): ... this.
5828 (altivec_lvxl_<mode>): Delete expand.
5829 (*altivec_lvxl_<mode>_internal): Rename to ...
5830 (altivec_lvxl_<mode>): ... this.
5831 (altivec_stvxl_<mode>): Delete expand.
5832 (*altivec_stvxl_<mode>_internal): Rename to ...
5833 (altivec_stvxl_<mode>): ... this.
5834 (altivec_stve<VI_char>x): Delete expand.
5835 (*altivec_stve<VI_char>x_internal): Rename to ...
5836 (altivec_stve<VI_char>x): ... this.
5837 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5838 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5839 reduc_plus_scal_<mode>): Adjust.
5840 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5842 (rs6000_cpu_cpp_builtins): Adjust.
5843 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5844 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5845 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5846 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5847 -maltivec=be support.
5848 (rs6000_split_vec_extract_var): Adjust.
5849 (rs6000_split_v4si_init): Adjust.
5850 (swap_selector_for_mode): Delete.
5851 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5852 altivec_expand_stvex_be): Delete.
5853 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5854 -maltivec=be support.
5855 (rs6000_gimple_fold_builtin): Ditto.
5856 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5858 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5859 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5860 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5861 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5862 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5863 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5864 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5865 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5866 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5867 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5868 anonymous split): Adjust.
5869 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5870 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5872 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5874 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5875 when --with-gxx-include-dir is also specified.
5876 * configure: Regenerate.
5878 2018-05-09 Jim Wilson <jimw@sifive.com>
5881 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5882 * config/riscv/t-withmultilib: New.
5883 * config/riscv/withmultilib.h: New.
5884 * doc/install.texi: Document RISC-V --with-multilib-list support.
5886 2018-05-09 Richard Biener <rguenther@suse.de>
5888 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5890 (vect_bb_vectorization_profitable_p): Adjust. Compute
5891 actual scalar cost using the cost vector and the add_stmt_cost
5894 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5896 PR rtl-optimization/85645
5897 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5898 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5900 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5902 PR rtl-optimization/85645
5903 * shrink-wrap.c (spread_components): Return a boolean saying if
5904 anything was changed.
5905 (try_shrink_wrapping_separate): Iterate spread_components until
5906 nothing changes anymore.
5908 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5910 PR rtl-optimization/85645
5911 * regrename.c (build_def_use): Also kill the chains that include the
5912 destination of a REG_CFA_REGISTER note.
5914 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5916 PR rtl-optimization/85645
5917 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5918 insn that has a REG_CFA_REGISTER note.
5920 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5922 * cfgexpand.c (expand_clobber): New function.
5923 (expand_gimple_stmt_1): Use it.
5924 * tree-vect-stmts.c (vect_clobber_variable): New function,
5926 (vectorizable_simd_clone_call): ...here.
5927 (vectorizable_store): Emit a clobber either side of an
5928 IFN_STORE_LANES sequence.
5929 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5931 2018-05-09 Tom de Vries <tom@codesourcery.com>
5934 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5935 (define_insn "trap_if_false"): Add exit after trap.
5937 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5939 PR rtl-optimization/85638
5940 * bb-reorder.c: Include common/common-target.h.
5941 (create_forwarder_block): New function extracted from...
5942 (fix_up_crossing_landing_pad): ...here. Rename into...
5943 (dw2_fix_up_crossing_landing_pad): ...this.
5944 (sjlj_fix_up_crossing_landing_pad): New function.
5945 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5946 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5947 from both partitions and exit the loop after one iteration.
5949 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5952 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5954 (Basic PowerPC Built-in Functions): The new name of the
5955 subsection previously known as "PowerPC Built-in Functions".
5956 (Basic PowerPC Built-in Functions Available on all Configurations):
5958 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5960 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5962 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5964 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5967 2018-05-08 Jim Wilson <jimw@sifive.com>
5969 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5970 (LD_EMUL_SUFFIX): New.
5971 (LINK_SPEC): Use it.
5973 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5975 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5977 (Basic PowerPC Built-in Functions): The new name of the
5978 subsection previously known as "PowerPC Built-in Functions".
5979 (Basic PowerPC Built-in Functions Available on all Configurations):
5981 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5983 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5985 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5987 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5990 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5993 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5994 after cmpelim optimization.
5996 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5998 * config.gcc: Support "goldmont".
5999 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6000 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6002 * config/i386/i386.c (m_GOLDMONT): Define.
6003 (processor_target_table): Add "goldmont".
6004 (PTA_GOLDMONT): Define.
6005 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6006 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6007 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6008 (fold_builtin_cpu): Add "goldmont".
6009 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6010 (ix86_option_override_internal): Add "goldmont".
6011 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6012 (processor_type): Add PROCESSOR_GOLDMONT.
6013 * config/i386/i386.md: Add CPU "glm".
6014 * config/i386/glm.md: New file.
6015 * config/i386/x86-tune.def: Add m_GOLDMONT.
6016 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6018 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6021 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6023 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6024 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6025 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6028 * config/i386/i386.c (ix86_fold_builtin): Handle
6029 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6032 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6033 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6035 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6038 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6039 (check_arch): Likewise.
6040 (check_fpu): Return the result rather than printing it.
6041 (end arch): Fix operator precedence.
6042 (end cpu): Likewise.
6043 (END): Print the result from check_fpu.
6045 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6046 Alan Hayward <alan.hayward@arm.com>
6047 David Sherwood <david.sherwood@arm.com>
6049 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6050 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6051 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6052 (*fcmuo<mode>_and): New patterns.
6054 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6056 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6057 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6058 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6059 (cmp_op, sve_imm_con): New code attributes.
6060 (SVE_COND_INT_CMP, imm_con): Delete.
6061 (cmp_op): Remove above unspecs from int attribute.
6062 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6064 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6065 comparison-specific unspecs.
6066 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6067 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6068 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6069 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6070 (*vec_fcm<cmp_op><mode>): Rename to...
6071 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6072 (*vec_fcmuo<mode>): Rename to...
6073 (*fcmuo<mode>): ...this and adjust likewise.
6074 (*pred_fcm<cmp_op><mode>): New pattern.
6075 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6076 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6078 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6080 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6081 (aarch64_emit_sve_predicated_cond): New function.
6082 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6083 (aarch64_emit_unspec_cond_or): Replace with...
6084 (aarch64_emit_sve_or_conds): ...this new function. Use
6085 aarch64_emit_sve_ptrue_op for the individual comparisons and
6086 aarch64_emit_binop to OR them together.
6087 (aarch64_emit_inverted_unspec_cond): Replace with...
6088 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6089 aarch64_emit_sve_ptrue_op for the comparison and
6090 aarch64_emit_unop to invert the result.
6091 (aarch64_expand_sve_vec_cmp_float): Update after the above
6092 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6094 2018-05-07 Nathan Sidwell <nathan@acm.org>
6096 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6097 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6098 (Backwards Compatibility): Likewise.
6100 2018-05-07 Luis Machado <luis.machado@linaro.org>
6104 2018-05-07 Luis Machado <luis.machado@linaro.org>
6106 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6107 <prefetch_dynamic_strides>: New const bool field.
6108 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6109 prefetch_dynamic_strides.
6110 (exynosm1_prefetch_tune): Likewise.
6111 (thunderxt88_prefetch_tune): Likewise.
6112 (thunderx_prefetch_tune): Likewise.
6113 (thunderx2t99_prefetch_tune): Likewise.
6114 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6116 (aarch64_override_options_internal): Update to set
6117 PARAM_PREFETCH_DYNAMIC_STRIDES.
6118 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6119 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6120 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6121 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6122 prefetch-dynamic-strides setting.
6124 2018-05-07 Luis Machado <luis.machado@linaro.org>
6126 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6127 <minimum_stride>: New const int field.
6128 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6129 minimum_stride field.
6130 (exynosm1_prefetch_tune): Likewise.
6131 (thunderxt88_prefetch_tune): Likewise.
6132 (thunderx_prefetch_tune): Likewise.
6133 (thunderx2t99_prefetch_tune): Likewise.
6134 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6135 (aarch64_override_options_internal): Update to set
6136 PARAM_PREFETCH_MINIMUM_STRIDE.
6137 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6138 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6139 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6141 stride is constant and is below the minimum stride threshold.
6143 2018-05-07 Luis Machado <luis.machado@linaro.org>
6145 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6148 2018-05-07 Luis Machado <luis.machado@linaro.org>
6150 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6151 <prefetch_dynamic_strides>: New const bool field.
6152 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6153 prefetch_dynamic_strides.
6154 (exynosm1_prefetch_tune): Likewise.
6155 (thunderxt88_prefetch_tune): Likewise.
6156 (thunderx_prefetch_tune): Likewise.
6157 (thunderx2t99_prefetch_tune): Likewise.
6158 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6160 (aarch64_override_options_internal): Update to set
6161 PARAM_PREFETCH_DYNAMIC_STRIDES.
6162 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6163 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6164 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6165 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6166 prefetch-dynamic-strides setting.
6168 2018-05-07 Luis Machado <luis.machado@linaro.org>
6170 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6171 <minimum_stride>: New const int field.
6172 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6173 minimum_stride field.
6174 (exynosm1_prefetch_tune): Likewise.
6175 (thunderxt88_prefetch_tune): Likewise.
6176 (thunderx_prefetch_tune): Likewise.
6177 (thunderx2t99_prefetch_tune): Likewise.
6178 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6179 (aarch64_override_options_internal): Update to set
6180 PARAM_PREFETCH_MINIMUM_STRIDE.
6181 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6182 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6183 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6184 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6185 stride is constant and is below the minimum stride threshold.
6187 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6190 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6191 the type is addressable. Don't force op into register if it has
6194 2018-05-05 Roland McGrath <mcgrathr@google.com>
6197 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6198 any section for which we don't know a specific type it should have,
6199 regardless of name. Previously this was done only for the exact
6200 names ".init_array", ".fini_array", and ".preinit_array".
6201 (default_elf_asm_named_section): Add comment about
6202 relationship with default_section_type_flags and SECTION_NOTYPE.
6203 (get_section): Don't consider it a type conflict if one side has
6204 SECTION_NOTYPE and the other doesn't, as long as neither has the
6205 SECTION_BSS et al used in the default_section_type_flags logic.
6207 2018-05-05 Tom de Vries <tom@codesourcery.com>
6210 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6211 (workaround_barsyncs): New function.
6212 (nvptx_reorg): Use workaround_barsyncs.
6213 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6214 (define_expand "nvptx_membar_cta"): New define_expand.
6215 (define_insn "*nvptx_membar_cta"): New insn.
6217 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6219 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6220 To improve optimization opportunities.
6221 * builtin-types.def: The new needed builtin types for the above.
6223 2018-05-04 Richard Biener <rguenther@suse.de>
6225 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6226 * gimple-ssa-store-merging.c
6227 (imm_store_chain_info::output_merged_store): Remove redundant create,
6228 release split_store vector contents on failure.
6229 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6230 scalar stmt vector on cache hit.
6232 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6234 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6236 * config.gcc (powerpc-xilinx-eabi*): Remove.
6237 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6239 (fusion_addis_mem_combo_load): Ditto.
6240 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6242 (rs6000_cpu_cpp_builtins): Ditto.
6243 * config/rs6000/rs6000-linux.c
6244 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6245 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6246 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6248 (rs6000_setup_reg_addr_masks): Ditto.
6249 (rs6000_init_hard_regno_mode_ok): Ditto.
6250 (rs6000_option_override_internal): Ditto.
6251 (legitimate_lo_sum_address_p): Ditto.
6252 (rs6000_legitimize_address): Ditto.
6253 (rs6000_legitimize_reload_address): Ditto.
6254 (rs6000_legitimate_address_p): Ditto.
6255 (abi_v4_pass_in_fpr): Ditto.
6256 (setup_incoming_varargs): Ditto.
6257 (rs6000_gimplify_va_arg): Ditto.
6258 (rs6000_split_multireg_move): Ditto.
6259 (rs6000_savres_strategy): Ditto.
6260 (rs6000_emit_prologue_components): Ditto.
6261 (rs6000_emit_epilogue_components): Ditto.
6262 (rs6000_emit_prologue): Ditto.
6263 (rs6000_emit_epilogue): Ditto.
6264 (rs6000_elf_file_end): Ditto.
6265 (rs6000_function_value): Ditto.
6266 (rs6000_libcall_value): Ditto.
6267 * config/rs6000/rs6000.h: Ditto.
6268 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6269 (TARGET_MINMAX): ... this. New.
6270 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6271 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6272 (*movsi_internal1_single): Delete.
6273 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6274 mfpu=, mxilinx-fpu): Delete.
6275 * config/rs6000/singlefp.h: Delete.
6276 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6277 * config/rs6000/t-rs6000: Ditto.
6278 * config/rs6000/t-xilinx: Delete.
6279 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6280 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6281 (VStype_simple): Delete.
6282 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6283 * config/rs6000/xfpu.h: Delete.
6284 * config/rs6000/xfpu.md: Delete.
6285 * config/rs6000/xilinx.h: Delete.
6286 * config/rs6000/xilinx.opt: Delete.
6287 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6288 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6290 2018-05-04 Tom de Vries <tom@codesourcery.com>
6293 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6296 2018-05-04 Richard Biener <rguenther@suse.de>
6299 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6300 (expand_complex_div_wide): Likewise.
6301 (expand_complex_operations_1): Likewise.
6302 (expand_complex_libcall): Preserve EH info of the original stmt.
6303 (tree_lower_complex): Handle removed blocks.
6304 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6305 on complex multiplication and division libcall builtins.
6307 2018-05-04 Richard Biener <rguenther@suse.de>
6310 * fold-const.c (negate_expr_p): Restrict negation of operand
6311 zero of a division to when we know that can happen without
6313 (fold_negate_expr_1): Likewise.
6315 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6318 * real.h (real_nextafter): Declare.
6319 * real.c (real_nextafter): New function.
6320 * fold-const-call.c (fold_const_nextafter): New function.
6321 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6322 CASE_CFN_NEXTTOWARD.
6323 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6324 even when arg1_mode is different from arg0_mode.
6326 2018-05-03 Nathan Sidwell <nathan@acm.org>
6328 * doc/extend.texi (Deprecated Features): Remove
6330 (Backwards Compatibility): Likewise.
6331 * doc/invoke.texi (C++ Language Options): Likewise.
6332 (C++ Dialect Options): Likewise.
6334 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6337 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6338 _mm512_mask_mullox_epi64): New intrinsics.
6340 2018-05-03 Tom de Vries <tom@codesourcery.com>
6343 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6344 dump files): Add offload-tree.
6346 2018-05-03 Richard Biener <rguenther@suse.de>
6348 PR tree-optimization/85615
6349 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6350 to loops not nested in BBs loop father to avoid creating multi-entry
6353 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6355 PR tree-optimization/70291
6356 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6357 arguments. Change return type to tree. Emit libcall as a new
6358 statement rather than replacing existing one when inplace_p is true.
6359 (expand_complex_multiplication_components): New function.
6360 (expand_complex_multiplication): Expand floating-point complex
6361 multiplication using the above.
6362 (expand_complex_division): Rename inner_type parameter to type.
6363 Update expand_complex_libcall call-site.
6364 (expand_complex_operations_1): Update expand_complex_multiplication
6365 and expand_complex_division call-sites.
6367 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6370 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6371 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6372 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6373 the highest significant bit of the shift count mask is clear. In
6374 check whether and[sq]i3 is needed verify that all significant bits
6375 of the shift count other than the highest are set.
6377 2018-05-02 Tom de Vries <tom@codesourcery.com>
6380 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6381 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6382 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6383 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6384 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6385 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6386 __builtin_goacc_parlevel_size.
6388 2018-05-02 Richard Biener <rguenther@suse.de>
6390 PR tree-optimization/85597
6391 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6392 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6395 2018-05-02 Tom de Vries <tom@codesourcery.com>
6398 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6399 dump files): Add ltrans-tree.
6401 2018-05-02 Tom de Vries <tom@codesourcery.com>
6404 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6405 dump files): Add wpa-ipa.
6407 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6409 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6410 powerpc*-*-linux*paired* target.
6411 * config/rs6000/750cl.h: Delete.
6412 * config/rs6000/paired.h: Delete.
6413 * config/rs6000/paired.md: Delete.
6414 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6416 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6417 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6418 comment. Remove paired float support.
6419 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6420 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6422 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6423 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6425 * config/rs6000/rs6000.c: Remove paired float support.
6426 (paired_expand_vector_init, paired_expand_vector_move,
6427 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6428 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6429 paired_expand_builtin, paired_expand_predicate_builtin,
6430 paired_init_builtins): Delete.
6431 * config/rs6000/rs6000.h: Remove paired float support.
6432 * config/rs6000/rs6000.md: Remove paired float support.
6433 (move_from_CR_ov_bit): Delete.
6434 * config/rs6000/rs6000.opt (mpaired): Delete.
6435 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6436 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6438 2018-05-02 Richard Biener <rguenther@suse.de>
6441 * gimplify.c (gimplify_save_expr): When in SSA form allow
6442 SAVE_EXPRs to compute to SSA vars.
6444 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6447 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6448 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6449 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6450 clobber operands[2], instead use a new pseudo. Formatting fixes.
6452 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6454 PR tree-optimization/85586
6455 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6456 exit early for statements in the same group if the accesses are
6459 2018-05-02 Tom de Vries <tom@codesourcery.com>
6462 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6465 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6467 PR tree-optimization/85143
6468 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6470 2018-05-01 Tom de Vries <tom@codesourcery.com>
6473 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6474 not found" error message.
6476 2018-05-01 Tom de Vries <tom@codesourcery.com>
6479 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6480 * vec.c (test_ordered_remove_if): New function.
6481 (vec_c_tests): Call test_ordered_remove_if.
6482 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6483 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6484 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6485 VEC_ORDERED_REMOVE_IF.
6487 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6489 PR tree-optimization/82665
6490 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6491 pointer subtraction where arguments come from a memchr call.
6493 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6495 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6496 --push-state --as-needed and --pop-state instead of --as-needed and
6497 --no-as-needed if ld supports it.
6498 * configure: Regenerated.
6501 * doc/install.texi2html: Replace _002d with - and _002a with * in
6502 generated html files using sed.
6504 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6507 * gcc-rich-location.c (blank_line_before_p): New function.
6508 (use_new_line): New function.
6509 (gcc_rich_location::add_fixit_insert_formatted): New function.
6510 * gcc-rich-location.h
6511 (gcc_rich_location::add_fixit_insert_formatted): New function.
6513 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6515 * selftest.c (assert_streq): Rename "expected" and "actual" to
6516 "val1" and "val2". Extend NULL-handling to cover both inputs
6517 symmetrically, while still requiring both to be non-NULL for a pass.
6518 * selftest.h (assert_streq): Rename "expected" and "actual" to
6520 (ASSERT_EQ): Likewise.
6521 (ASSERT_EQ_AT): Likewise.
6522 (ASSERT_KNOWN_EQ): Likewise.
6523 (ASSERT_KNOWN_EQ_AT): Likewise.
6524 (ASSERT_NE): Likewise.
6525 (ASSERT_MAYBE_NE): Likewise.
6526 (ASSERT_MAYBE_NE_AT): Likewise.
6527 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6528 the assertion to pass.
6529 (ASSERT_STREQ_AT): Likewise.
6531 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6533 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6534 interaction with -pie.
6536 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6538 * selftest.h: Fix alphabetization of per-source-file selftest
6541 2018-04-30 Jason Merrill <jason@redhat.com>
6543 PR c++/61982 - dead stores to destroyed objects.
6544 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6547 2018-04-30 Jason Merrill <jason@redhat.com>
6549 * tree.c (build_clobber): New.
6550 * tree.h: Declare it.
6551 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6553 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6555 * diagnostic-show-locus.c (layout::layout): Update for
6556 location_get_source_line returning a char_span.
6557 (struct char_span): Move to input.h.
6558 (struct correction): Update for fields in char_span becoming
6560 (struct source_line): Update for location_get_source_line
6561 returning a char_span.
6562 (layout::print_line): Likewise.
6563 * edit-context.c (edited_file::print_content): Likewise.
6564 (edited_file::print_diff_hunk): Likewise.
6565 (edited_file::print_run_of_changed_lines): Likewise.
6566 (edited_file::get_num_lines): Likewise.
6567 (edited_line::edited_line): Likewise.
6568 * final.c (asm_show_source): Likewise.
6569 * input.c (location_get_source_line): Convert return type
6570 from const char * to char_span, losing the final "line_len"
6572 (dump_location_info): Update for the above.
6573 (get_substring_ranges_for_loc): Likewise. Use a char_span
6574 when handling the literal within the line.
6575 (test_reading_source_line): Update for location_get_source_line
6576 returning a char_span.
6577 * input.h (class char_span): Move here from
6578 diagnostic-show-locus.c, converting from a struct to a class.
6579 Make data members private.
6580 (char_span::operator bool): New.
6581 (char_span::length): New.
6582 (char_span::get_buffer): New.
6583 (char_span::operator[]): New.
6584 (char_span::subspan): Make const.
6585 (char_span::xstrdup): New.
6586 (location_get_source_line): Convert return type from const char *
6587 to char_span, losing the final "line_size" param.
6589 2018-04-30 Jan Hubicka <jh@suse.cz>
6591 * lto-wrapper.c (ltrans_priorities): New static var.
6592 (cmp_priority): New.
6593 (run_gcc): Read priorities and if doing parallel build order
6594 the Makefile by them.
6596 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6598 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6600 2018-04-30 Richard Biener <rguenther@suse.de>
6602 * tree-cfg.c (verify_address): Remove base argument, add
6603 flag whether to check TREE_ADDRESSABLE and do that.
6604 (verify_expr): Remove.
6605 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6606 (verify_gimple_assign_single): Likewise.
6607 (verify_gimple_switch): Likewise.
6608 (verify_expr_location_1): Dereference tp once. Add (disabled)
6609 piece from verify_expr.
6610 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6612 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6614 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6616 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6618 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6619 (small_data_pattern): Likewise.
6620 (arc_rewrite_small_data): Likewise.
6621 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6622 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6623 (get_symbol_alignment): New function.
6624 (legitimate_small_data_address_p): Likewise.
6625 (legitimate_scaled_address): Update, call
6626 legitimate_small_data_address_p.
6627 (output_sdata): New static variable.
6628 (arc_print_operand): Update how we handle small data operands.
6629 (arc_print_operand_address): Likewise.
6630 (arc_legitimate_address_p): Update, use
6631 legitimate_small_data_address_p.
6632 (arc_rewrite_small_data_p): Remove.
6633 (arc_rewrite_small_data_1): Likewise.
6634 (arc_rewrite_small_data): Likewise.
6635 (small_data_pattern): Likewise.
6636 (compact_sda_memory_operand): Update to use
6637 legitimate_small_data_address_p and get_symbol_alignment.
6638 (prepare_move_operands): Don't rewite sdata pattern.
6639 (prepare_extend_operands): Remove.
6640 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6642 (zero_extendqisi2): Likewise.
6643 (zero_extendhisi2): Likewise.
6644 (extendqihi2): Likewise.
6645 (extendqisi2): Likewise.
6646 (extendhisi2): Likewise.
6650 * config/arc/constraints.md (Usd): Change it to memory constraint.
6652 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6654 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6655 as source of std instructions.
6656 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6657 allow 6-bit constants as source for store instructions.
6658 (movdi_insn): Update instruction pattern to allow 6-bit constants
6659 as source for store instructions.
6661 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6663 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6665 2018-04-30 Nathan Sidwell <nathan@acm.org>
6666 Sandra Loosemore <sandra@codesourcery.com>
6668 * dumpfile.c (dump_open): Allow '-' for stdout.
6669 * doc/invoke.texi (Developer Options): Document dump filename
6670 determination early. Document stdin/stdout selection.
6672 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6674 Microblaze Target: PIC data text relative
6676 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6677 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6679 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6680 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6681 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6682 New addressing mode for data-text relative position indepenedent code.
6683 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6684 'ADDRESS_SYMBOLIC_TXT_REL'.
6685 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6686 (microblaze_legitimate_pic_operand): Exclude function calls from
6687 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6688 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6690 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6691 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6692 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6693 for 'address + offset'.
6694 (microblaze_expand_prologue): Add new function prologue call for
6696 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6697 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6698 table in case of TARGET_PIC_DATA_TEXT_REL.
6699 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6700 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6701 Add new macros 'UNSPEC_TEXT',
6702 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6703 + exclude function calls from 'UNSPEC_PLT' in case of data text
6705 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6706 new target hook for generating address diff vector tables in case of
6708 * doc/tm.texi : Regenerate.
6709 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6710 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6711 of addr diff vector generation.
6712 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6713 target hook definition.
6714 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6715 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6716 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6717 option for data text relative.
6719 2018-04-30 Richard Biener <rguenther@suse.de>
6721 * tree-chrec.h (evolution_function_is_constant_p): Remove
6723 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6725 2018-04-30 Richard Biener <rguenther@suse.de>
6728 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6730 2018-04-30 Richard Biener <rguenther@suse.de>
6732 PR tree-optimization/28364
6733 PR tree-optimization/85275
6734 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6735 copying first exit test.
6737 2018-04-28 Mark Wielaard <mark@klomp.org>
6739 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6741 (dwarf_AT): Handle DW_AT_addr_base.
6742 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6744 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6747 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6748 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6749 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6750 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6752 2018-04-28 Richard Biener <rguenther@suse.de>
6754 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6755 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6756 to reflect use. Only add interesting stmts.
6758 2018-04-27 Martin Jambor <mjambor@suse.cz>
6761 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6762 the jump function allows for passing through aggregate values.
6764 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6766 * input.h (in_system_header_at): Convert from macro to inline
6768 (from_macro_expansion_at): Likewise.
6769 (from_macro_definition_at): Likewise.
6771 2018-04-27 Jeff Law <law@redhat.com>
6773 * config.gcc: Mark tile* targets as deprecated/obsolete.
6775 2018-04-27 Richard Biener <rguenther@suse.de>
6777 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6780 2018-04-27 Richard Biener <rguenther@suse.de>
6782 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6784 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6786 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6787 with Yd constraint. Set "preferred_for_speed" attribute from
6788 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6790 (*movdi_internal): Ditto.
6791 (movti_interunit splitters): Remove
6792 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6793 (movdi_interunit splitters): Ditto.
6794 * config/i386/constraints.md (Ye): Remove.
6795 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6797 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6800 * config/aarch64/constraints.md (Usg): Limit to 31.
6803 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6805 PR tree-optimization/85529
6806 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6807 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6808 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6809 zero extension or masking of the MSB bit.
6810 (optimize_range_tests): Add FIRST_BB argument, pass it through
6811 to optimize_range_tests_var_bound.
6812 (maybe_optimize_range_tests, reassociate_bb): Adjust
6813 optimize_range_tests callers.
6815 2018-04-26 Richard Biener <rguenther@suse.de>
6816 Jakub Jelinek <jakub@redhat.com>
6818 * cgraph.h (symbol_table): Just declare debug method here.
6819 * symtab.c (symbol_table::debug): Define.
6821 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6823 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6825 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6827 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6828 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6829 (*movdi_internal): Substitute Yi and Yj constraint with x
6830 and Ym and Yn constraint with y constraint. Update "isa"
6831 attribute and set "preferred_for_speed" attribute from
6832 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6833 (*movsi_internal): Ditto.
6834 (*movdf_internal): Ditto.
6835 (*movsf_internal): Ditto.
6836 (*zero_extendsidi2): Ditto.
6837 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6838 (sse2_loadld): Ditto.
6839 (*vec_extract<ssevecmodelower>_0): Ditto.
6840 (*vec_extractv4si_0_zext_sse4): Ditto.
6841 (vec_concatv2di): Ditto.
6842 (*vec_dup<mode>): Ditto.
6843 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6844 * config/i386/constraints.md (Yi): Remove.
6849 2018-04-26 Nathan Sidwell <nathan@acm.org>
6851 * dumpfile.c (dump_open): New.
6852 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6853 (dump_finish): Detect stdio/stderr by value not name.
6855 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6857 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6859 2018-04-26 Tom de Vries <tom@codesourcery.com>
6862 * config/nvptx/nvptx.c (verify_neutering_jumps)
6863 (verify_neutering_labels): New function
6864 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6866 2018-04-26 Tom de Vries <tom@codesourcery.com>
6869 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6870 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6873 2018-04-26 Richard Biener <rguenther@suse.de>
6874 Tom de Vries <tom@codesourcery.com>
6877 * lto-streamer-out.c (output_function): Fixup loops if required to match
6878 discovery done in the reader.
6880 2018-04-26 Richard Biener <rguenther@suse.de>
6882 PR tree-optimization/85116
6883 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6884 have a loop exit from the single latch predecessor. Remove
6885 case of header with just condition.
6886 (ch_base::copy_headers): Exclude infinite loops from any
6888 (pass_ch::execute): Record exits.
6890 2018-04-26 Richard Biener <rguenther@suse.de>
6892 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6893 prologue cost vector and pass it to vect_get_load_cost.
6894 (vect_get_peeling_costs_all_drs): Likewise.
6895 (vect_peeling_hash_get_lowest_cost): Likewise.
6896 (vect_enhance_data_refs_alignment): Likewise.
6898 2018-04-26 Richard Biener <rguenther@suse.de>
6901 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6902 checking of integer<->pointer conversions.
6903 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6904 sign-/zero-extending pointer types.
6905 (expand_omp_for_static_chunk): Likewise.
6907 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6908 Jean Lee <xiaoyur347@gmail.com>
6910 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6911 (TARGET_ASAN_SHADOW_OFFSET): Define.
6912 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6913 true for -fsanitize=address.
6915 2018-04-25 Mark Wielaard <mark@klomp.org>
6917 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6920 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6922 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6923 than "alu", remove explicit "memory" and "imm_disp" attributes.
6924 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6927 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6928 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6929 gen_lowpart_no_emit.
6931 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6934 * config/i386/i386.c (ix86_expand_builtin): Change memory
6935 operand to XI, extend p0 to Pmode.
6936 * config/i386/i386.md: Change unspec volatile and operand
6937 1 mode to XI, change operand 0 mode to P.
6939 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6941 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6942 GET_MODE_MASK before any checking.
6943 (nds32_can_use_bset_p): Likewise.
6944 (nds32_can_use_btgl_p): Likewise.
6946 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6948 * config/nds32/nds32-doubleword.md: New define_split pattern for
6949 illegal register number.
6951 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6953 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6955 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6957 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6959 2018-04-25 Richard Biener <rguenther@suse.de>
6961 * lto-streamer.h (LTO_major_version): Bump to 8.
6963 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6965 * BASE-VER: Set to 9.0.0.
6967 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6969 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6970 in __abskf2 and __powikf2.
6972 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6975 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6976 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6977 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6978 Use the above on operand 2. Reindent.
6979 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6981 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6984 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6986 * config/i386/i386.opt (mcet): Removed.
6987 * doc/install.texi: Remove -mcet documentation.
6988 * doc/invoke.texi: Likewise.
6990 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6993 * doc/install.texi: Remove -mcet from bootstrap-cet.
6995 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6998 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6999 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7003 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7004 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7005 containing a CONST_VECTOR.
7007 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7009 * doc/install.texi: Update newlib dependency for nvptx.
7011 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7014 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7015 instead of INTVAL when shifting x left.
7017 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7019 PR tree-optimization/85478
7020 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7021 vect_grouped_store_supported for single element vectors.
7023 2018-04-24 Richard Biener <rguenther@suse.de>
7026 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7027 load cost increase to the case of non-constant step.
7029 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7032 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7033 destination if any_malformed_asm.
7035 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7038 * expr.c (store_field): In the bitfield case, if the value comes from
7039 a function call and is returned in registers by means of a PARALLEL,
7040 do not change the mode of the temporary unless BLKmode and VOIDmode.
7042 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7044 PR rtl-optimization/85423
7045 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7046 dependencies to debug insns when the previous insn is non-debug.
7048 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7050 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7051 enums into a single definition.
7052 (fls): Fix predicates and printing.
7055 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7057 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7058 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7059 and short u6 immediate.
7060 (check_if_valid_sleep_operand): Remove.
7061 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7063 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7065 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7066 flag_always_save_lp condition.
7067 * config/nds32/nds32.opt (malways-save-lp): New option.
7069 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7071 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7072 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7073 * config/nds32/nds32.h
7074 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7075 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7077 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7079 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7080 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7082 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7083 Chung-Ju Wu <jasonwucj@gmail.com>
7085 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7087 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7088 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7090 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7092 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7094 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7096 * config/nds32/nds32-protos.h (nds32_data_alignment,
7097 nds32_local_alignment): Declare.
7098 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7099 nds32_local_alignment): New functions.
7100 (TARGET_CONSTANT_ALIGNMENT): Define.
7101 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7103 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7105 * config/nds32/nds32.c
7106 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7107 (TARGET_MODES_TIEABLE_P): Likewise.
7109 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7111 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7114 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7115 Chung-Ju Wu <jasonwucj@gmail.com>
7117 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7118 for unaligned access.
7119 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7121 * config/nds32/nds32-intrinsic.md: Likewise.
7122 * config/nds32/nds32_intrinsic.h: Likewise.
7123 * config/nds32/nds32.h (nds32_builtins): Likewise.
7124 * config/nds32/nds32.opt (munaligned-access): New option.
7125 * config/nds32/nds32.c (nds32_asm_file_start): Display
7126 flag_unaligned_access status.
7128 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7130 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7131 -mno-relax is present.
7132 * config/riscv/linux.h (LINK_SPEC): Ditto.
7134 2018-04-20 Martin Sebor <msebor@redhat.com>
7137 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7139 (gimple_fold_builtin_stxcpy_chk): Same.
7140 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7142 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7145 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7146 __powikf2 when long double is IEEE 128-bit.
7148 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7150 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7151 step to make sure stack always aligned.
7153 2018-04-20 Carl Love <cel@us.ibm.com>
7156 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7157 size check for arg0.
7159 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7160 Tom de Vries <tom@codesourcery.com>
7163 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7164 Emit insns for calls too.
7165 (nvptx_find_par): Always look for worker-level predecessor insn.
7166 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7168 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7169 (nvptx_process_pars): Propagate frames for calls.
7171 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7174 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7176 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7177 (ix86_handle_option): Don't handle OPT_mibt.
7178 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7180 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7182 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7183 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7184 (ix86_target_macros): Define __CET__ with flag_cf_protection
7185 for -fcf-protection.
7186 * config/i386/i386.c (isa2_opts): Remove -mibt.
7187 * config/i386/i386.h (TARGET_IBT): Removed.
7188 (TARGET_IBT_P): Likewise.
7189 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7190 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7191 * config/i386/i386.opt (mcet): Update help message.
7194 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7195 -mcet as an alias for -mshstk.
7197 2018-04-20 Richard Biener <rguenther@suse.de>
7200 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7201 complexity by forcing a single use of the multiply operand.
7203 2018-04-20 Martin Jambor <mjambor@suse.cz>
7206 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7207 recursion dependency to only apply to non-clones.
7209 2018-04-20 Martin Jambor <mjambor@suse.cz>
7212 * ipa-cp.c (create_specialized_node): Check that clones of
7213 self-recursive edges exist during IPA-CP.
7215 2018-04-19 Toon Moene <toon@moene.org>
7217 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7220 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7222 PR tree-optimization/85467
7223 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7224 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7225 VECTOR_CST element to type.
7227 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7230 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7231 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7232 (builtin_longjmp): Likewise.
7233 (save_stack_nonlocal): New pattern.
7234 (restore_stack_nonlocal): Likewise.
7236 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7239 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7240 Replace ASM_OUTPUT_LABEL with fprintf.
7242 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7245 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7246 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7247 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7248 define __IBT__ and __SHSTK__ for -fcf-protection.
7249 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7251 (ix86_trampoline_init): Likewise.
7252 (x86_output_mi_thunk): Likewise.
7253 (ix86_notrack_prefixed_insn_p): Likewise.
7254 (ix86_option_override_internal): Don't disallow -fcf-protection.
7255 * config/i386/i386.md (rdssp<mode>): Also enable for
7257 (incssp<mode>): Likewise.
7258 (nop_endbr): Likewise.
7259 * config/i386/i386.opt (mcet): Change help message to built-in
7263 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7264 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7265 enable CET built-in functions.
7267 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7269 * common/config/i386/i386-common.c
7270 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7271 OPTION_MASK_ISA_MOVDIRI_UNSET,
7272 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7273 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7274 * config.gcc (movdirintrin.h): New header.
7275 * config/i386/cpuid.h (bit_MOVDIRI,
7276 bit_MOVDIR64B): New bits.
7277 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7279 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7280 (VOID, PVOID, PCVOID)): New function types.
7281 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7282 __builtin_ia32_directstoreu_u64,
7283 __builtin_ia32_movdir64b): New builtins.
7284 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7285 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7287 (ix86_valid_target_attribute_inner_p): Ditto.
7288 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7289 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7290 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7291 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7292 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7293 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7294 (movdiri<mode>, movdir64b_<mode>): New.
7295 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7296 * config/i386/immintrin.h: Include movdirintrin.h.
7297 * config/i386/movdirintrin.h: New file.
7298 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7300 2018-04-19 Richard Biener <rguenther@suse.de>
7303 * cfg.c (clear_bb_flags): When loop state says we have
7304 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7306 2018-04-19 Richard Biener <rguenther@suse.de>
7308 PR tree-optimization/84737
7309 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7310 copying restrict info.
7311 (vect_setup_realignment): Use it.
7312 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7313 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7314 the first DR to all generated stores.
7315 (vectorizable_load): Likewise for loads.
7317 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7319 PR tree-optimization/85446
7320 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7321 the integral and pointer types to have the same precision.
7323 * doc/install.texi: Document --disable-cet being the default and
7326 2018-04-18 Martin Liska <mliska@suse.cz>
7328 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7331 2018-04-18 Martin Liska <mliska@suse.cz>
7334 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7337 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7338 arguments if they are comparable.
7340 2018-04-18 Martin Liska <mliska@suse.cz>
7343 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7346 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7349 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7352 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7353 ENDBR after calling __morestack.
7355 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7358 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7359 by using gcc_base_ver to generate a gcc_driver_version, and use
7360 it when generating GCC_DRIVER_NAME.
7361 * configure: Regenerate.
7363 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7366 * config.gcc: Obsolete powerpc*-*-*spe*.
7368 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7371 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7372 (stabstr_D): Change type of unum from unsigned int to
7373 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7376 2018-04-17 Jim Wilson <jimw@sifive.com>
7379 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7380 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7381 Set arg_pointer_offset after using pretend_args_size.
7383 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7385 PR rtl-optimization/85431
7386 * dse.c (record_store): Ignore zero width stores.
7389 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7390 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7391 __builtin_stack_restore rather than after it.
7392 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7393 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7394 argument instead of virtual_dynamic_stack_rtx.
7396 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7398 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7400 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7401 Add note to error message to explain internal mapping of overloaded
7402 built-in function name to non-overloaded built-in function name.
7403 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7406 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7409 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7410 where the inputs overlap with the output.
7412 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7414 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7415 (=v, v) alternative and explicit "memory" attribute.
7416 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7417 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7419 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7420 "sselog1" type instead of "sselog".
7421 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7422 "sselog". Remove explicit "memory" attribute.
7423 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7424 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7426 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7427 "sselog1" type instead of "sselog". Remove explicit "memory"
7429 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7430 use "sselog1" type instead of "sselog". Remove explicit "memory"
7432 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7433 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7435 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7436 "sselog1" type instead of "sselog". Remove explicit "memory"
7438 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7439 use "sselog1" type instead of "sselog". Remove explicit "memory"
7443 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7446 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7449 2018-04-17 Martin Jambor <mjambor@suse.cz>
7452 * ipa-cp.c (create_specialized_node): Call
7453 expand_all_artificial_thunks if necessary.
7455 2018-04-17 Martin Liska <mliska@suse.cz>
7458 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7459 in message, remote space in between '_G' and '('.
7461 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7464 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7465 avx512f_vmcmp<mode>3<round_saeonly_name>,
7466 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7467 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7468 avx512f_rndscale<mode><round_saeonly_name>,
7469 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7470 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7471 Use %<iptr>2 instead of %2 for -masm=intel.
7472 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7473 avx512f_vcvttss2usi<round_saeonly_name>,
7474 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7476 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7477 avx512f_vcvttsd2usi<round_saeonly_name>,
7478 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7479 Use %q1 instead of %1 for -masm=intel.
7480 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7481 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7482 of %3 for -masm=intel.
7483 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7485 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7487 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7489 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7490 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7492 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7494 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7495 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7496 %g1 and one with %0 and %1.
7497 (avx512er_vmrcp28<mode><round_saeonly_name>,
7498 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7500 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7501 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7502 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7503 of %0 and %{%4%} for -masm=intel.
7504 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7505 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7506 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7507 order of %0 and %{%5%}%{z%} for -masm=intel.
7509 2018-04-17 Jan Hubicka <jh@suse.cz>
7512 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7514 2018-04-17 Martin Liska <mliska@suse.cz>
7517 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7518 for target_clone error message. Make default implementation
7519 clone to be a local declaration.
7520 (separate_attrs): Add new argument and check for an empty
7522 (expand_target_clones): Handle it.
7523 (ipa_target_clone): Make redirection just for target_clones
7526 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7527 Tom de Vries <tom@codesourcery.com>
7530 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7531 tiled basic blocks without omp continue statements.
7533 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7536 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7537 vec_extract expression as having side effects to make sure it gets
7540 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7543 * config/i386/i386.c (get_builtin_code_for_version): Check
7546 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7549 * gcc/config.gcc: Support "skylake".
7550 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7552 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7553 (processor_target_table): Add "skylake".
7554 (ix86_option_override_internal): Add "skylake".
7555 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7556 PROCESSOR_CANNONLAKE.
7557 (get_builtin_code_for_version): Fix priority for
7558 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7559 PROCESSOR_SKYLAKE-AVX512.
7560 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7561 (processor_type): Add PROCESSOR_SKYLAKE.
7563 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7564 Jason Merrill <jason@redhat.com>
7567 * convert.c (convert_to_integer_1): Use direct recursion for
7568 enumeral types and types with a precision less than the number
7569 of bits in their mode.
7571 2018-04-16 Julia Koval <julia.koval@intel.com>
7574 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7575 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7577 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7580 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7581 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7582 and -mno-direct-move.
7584 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7587 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7588 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7589 Ensure negative shifts result in {0}.
7591 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7593 PR rtl-optimization/79916
7594 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7595 regs (if any) to define how to gnerate SD moves when LRA is in
7598 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7600 PR rtl-optimization/85393
7601 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7602 * except.c (expand_dw2_landing_pad_for_region): Make static.
7603 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7604 a label and unconditional jump to old_bb, rather than
7605 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7608 PR rtl-optimization/85376
7609 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7610 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7611 instead of a specific value.
7613 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7614 Bin Cheng <bin.cheng@arm.com>
7616 PR tree-optimization/82965
7617 PR tree-optimization/83991
7618 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7619 by_profile_only parameter.
7620 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7621 information if the loop was predicted to iterate too many times.
7622 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7624 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7627 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7630 2018-04-13 Martin Liska <mliska@suse.cz>
7631 Jakub Jelinek <jakub@redhat.com>
7634 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7635 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7636 * builtins.c (expand_builtin_memory_copy_args): Use
7637 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7638 handle dest_addr == pc_rtx.
7640 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7643 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7644 asked to not generate direct moves.
7645 (fix_trunc<mode>si2_stfiwx): Similar.
7646 (fix_trunc<mode>si2_internal): Similar.
7648 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7651 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7652 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7653 lookup if dest in some wider mode is known to be const0_rtx and
7654 if so, record permanent equivalence for it to be ZERO_EXTEND of
7655 the narrower mode destination.
7657 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7659 * lto-streamer-out.c (output_function): Revert 259346.
7660 * omp-expand.c (expand_oacc_for): Likewise.
7662 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7664 PR rtl-optimization/85354
7665 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7666 * sel-sched.c (sel_global_init): ... here.
7668 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7671 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7672 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7673 mode for PE-COFF targets.
7674 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7675 (i386_pe_asm_lto_end): Likewise.
7676 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7677 (TARGET_ASM_LTO_END): Likewise.
7678 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7679 (i386_pe_asm_lto_start): New function.
7680 (i386_pe_asm_lto_end): Likewise.
7682 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7683 Richard Biener <rguenther@suse.de>
7686 * lto-streamer-out.c (output_function): Fix CFG loop state before
7688 * omp-expand.c (expand_oacc_for): Handle calls to internal
7689 functions like regular functions.
7691 2018-04-12 Richard Biener <rguenther@suse.de>
7694 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7695 for the early LTO debug to properly generate references to it
7696 during DIE emission. Do not re-use that for the skeleton for
7698 (dwarf2out_early_finish): Likewise.
7700 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7703 * config/i386/sse.md
7704 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7705 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7706 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7707 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7708 and output is a reg, avoid creating invalid lowpart subreg, but
7709 instead split into a 512-bit move. Don't split if not AVX512VL,
7710 input is xmm16+ reg and output is a mem.
7711 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7712 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7713 xmm16+ reg and output is a mem.
7715 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7717 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7718 also for flag_dwarf2_cfi_asm.
7720 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7722 PR rtl-optimization/85342
7723 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7724 a bool scalar var inside of the loop instead. Don't try to update
7725 recog_data.operand after failed apply_change_group.
7727 2018-04-12 Tom de Vries <tom@codesourcery.com>
7730 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7731 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7732 array with flexible array member as array without given dimension.
7733 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7734 argument for undefined param to true.
7736 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7739 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7740 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7741 from PowerPC section.
7742 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7743 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7744 help text that is too long.
7745 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7746 help text that is too long.
7747 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7748 help text that is too long.
7750 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7752 * config/alpha/alpha.md (stack_probe_internal): Rename
7753 from "probe_stack". Update all callers.
7755 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7757 PR rtl-optimization/84566
7758 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7759 sched_macro_fuse_insns.
7761 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7764 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7765 (compute_block_dependences): ... from here.
7767 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7769 PR tree-optimization/85331
7770 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7771 from int to HOST_WIDE_INT.
7773 2018-04-11 Martin Jambor <mjambor@suse.cz>
7776 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7777 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7778 not the same as the source val.
7779 (cgraph_edge_brings_value_p): New parameter.
7780 (gather_edges_for_value): Pass destination value to
7781 cgraph_edge_brings_value_p.
7782 (perhaps_add_new_callers): Likewise.
7783 (get_info_about_necessary_edges): Likewise and exclude values brought
7784 only by self-recursive edges.
7785 (create_specialized_node): Redirect only clones of self-calling edges.
7786 (+self_recursive_pass_through_p): New function.
7787 (find_more_scalar_values_for_callers_subset): Use it.
7788 (find_aggregate_values_for_callers_subset): Likewise.
7789 (known_aggs_to_agg_replacement_list): Removed.
7790 (decide_whether_version_node): Re-calculate known constants for all
7791 remaining context clones.
7793 2018-04-11 Richard Biener <rguenther@suse.de>
7796 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7797 from early DWARF output.
7798 (dwarf2out_early_finish): Output line info unconditionally into
7799 early DWARF and add reference to it.
7801 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7804 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7805 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7806 other than V2DFmode using iptr mode attribute.
7807 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7809 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7811 PR rtl-optimization/84659
7812 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7814 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7817 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7819 (output_loc_list): Pass address of a dummy size variable even in the
7820 locview handling loop.
7821 (index_location_lists): Add comment on why skip_loc_list_entry can't
7824 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7827 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7830 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7833 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7834 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7835 and -mstring-compare-inline-limit.
7837 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7840 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7841 for stack clash protection in a register whenever we need it to be in
7844 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7846 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7847 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7849 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7852 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7854 (mlong-double-): Ditto.
7855 * config/rs6000/sysv4.opt (msdata=): Ditto.
7856 (mtls-size=): Ditto.
7858 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7860 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7861 erroneous entries for
7862 "vector int vec_ldl (int, long int *)", and
7863 "vector unsigned int vec_ldl (int, unsigned long int *)".
7864 Add comments and entries for
7865 "vector bool char vec_ldl (int, bool char *)",
7866 "vector bool short vec_ldl (int, bool short *)",
7867 "vector bool int vec_ldl (int, bool int *)",
7868 "vector bool long long vec_ldl (int, bool long long *)",
7869 "vector pixel vec_ldl (int, pixel *)",
7870 "vector long long vec_ldl (int, long long *)",
7871 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7872 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7873 type tree bool_long_long_type_node and correct definition of
7874 bool_V2DI_type_node to make reference to this new type tree.
7875 (rs6000_mangle_type): Replace erroneous reference to
7876 bool_long_type_node with bool_long_long_type_node.
7877 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7878 comments to emphasize sign distinctions for char and int types and
7879 replace RS6000_BTI_bool_long constant with
7880 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7881 use of RS6000_BTI_pixel.
7882 (bool_long_type_node): Remove this macro definition.
7883 (bool_long_long_type_node): New macro definition
7885 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7887 PR rtl-optimization/85300
7888 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7889 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7890 simplify_unary_operation fails.
7892 2018-04-10 Martin Liska <mliska@suse.cz>
7894 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7895 cgraph_edge and ipa_ref.
7897 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7901 * config/i386/sse.md
7902 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7903 computation of the VEC_MERGE selector from mask.
7904 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7905 Fix decoding of the VEC_MERGE selector into mask.
7907 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7909 PR tree-optimization/85286
7910 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7912 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7914 * final.c (final_1): Set insn_last_address as well as
7915 insn_current_address.
7917 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7920 * explow.c (emit_stack_probe): Call validize_mem on memory location
7921 before passing it to gen_probe_stack. Create address operand and
7922 legitimize it for the probe_stack_address case.
7924 2018-04-09 Jan Hubicka <jh@suse.cz>
7927 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7928 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7929 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7931 (free_lang_data): Rebuild type inheritance graph.
7933 2018-04-09 Martin Sebor <msebor@redhat.com>
7935 * invoke.texi (-finline-small-functions): Mention other optimization
7937 (-findirect-inlining, -fpartial-inlining): Same.
7938 (-finline-functions-called-once): Same.
7939 (-freorder-blocks-and-partition): Same.
7941 2018-04-09 Jan Hubicka <jh@suse.cz>
7944 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7945 jumps; choose last target that matches the criteria (i.e.
7946 no partition changes for non-crossing jumps).
7947 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7948 support for redirecting crossing jumps to non-crossing.
7950 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7952 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7953 also for naked functions.
7955 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7957 * config/arc/arc.md (add_shift): New pattern.
7958 (add_shift2): Likewise.
7959 (sub_shift): Likewise.
7960 (sub_shift_cmp0_noout): Likewise.
7961 (compare_si_ashiftsi): Likewise.
7962 (xbfu_cmp0_noout): New combine pattern.
7963 (xbfu_cmp0"): Likewise.
7964 (movsi_set_cc_insn): Place the predicable variant first.
7965 (commutative_binary_cmp0_noout): Remove clobber.
7966 (commutative_binary_cmp0): New pattern.
7967 (noncommutative_binary_cmp0): Likewise.
7968 (noncommutative_binary_cmp0_noout): Likewise.
7969 (noncommutative_binary_comparison_result_used): Removed.
7970 (rsub_cmp0): New pattern.
7971 (rsub_cmp0_noout): Likewise.
7972 (extzvsi): Changed, keep only meaningful variants.
7973 (SQH, SEZ): New iterators.
7974 (SQH_postfix): New mode attribute.
7975 (SEZ_prefix): New code attribute.
7976 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7977 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7978 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7980 (noncommutative_operator): Check the availability of barrel
7983 2018-04-09 Richard Biener <rguenther@suse.de>
7985 PR tree-optimization/85284
7986 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7987 Only use the niter constraining form of simple_iv when the exit
7990 2018-04-09 Tom de Vries <tom@codesourcery.com>
7993 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7994 (define_expand "*memory_barrier"): New define_expand.
7995 (define_insn "memory_barrier"): New insn.
7997 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7999 PR rtl-optimization/80463
8000 PR rtl-optimization/83972
8001 PR rtl-optimization/83480
8003 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8004 correct producer for the insn.
8005 (tidy_control_flow): Fixup seqnos in case of debug insns.
8007 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8009 PR rtl-optimization/83913
8011 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8012 different sched-times when merging exprs.
8014 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8016 PR rtl-optimization/83962
8018 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8019 tidy_fallthru_edge and tidy_control_flow.
8021 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8023 PR rtl-optimization/83530
8025 * sel-sched.c (force_next_insn): New global variable.
8026 (remove_insn_for_debug): When force_next_insn is true, also leave only
8027 next insn in the ready list.
8028 (sel_sched_region): When the region wasn't scheduled, make another pass
8029 over it with force_next_insn set to 1.
8031 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8033 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8035 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8036 for interrupt control.
8037 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8038 functions for interrupt control.
8039 * config/nds32/nds32-intrinsic.md: Likewise.
8040 * config/nds32/nds32_intrinsic.h: Likewise.
8041 * config/nds32/nds32.h (nds32_builtins): Likewise.
8043 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8045 * config/nds32/nds32.c (nds32_init_machine_status,
8046 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8047 strict_aligned_p field.
8048 (nds32_expand_to_rtl_hook): New function.
8049 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8050 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8052 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8053 Chung-Ju Wu <jasonwucj@gmail.com>
8055 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8056 * config/nds32/nds32-n7.md: New file.
8057 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8058 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8060 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8061 * config/nds32/nds32.md (pipeline_model): Add n7.
8062 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8063 * config/nds32/pipelines.md: Include n7 settings.
8065 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8066 Chung-Ju Wu <jasonwucj@gmail.com>
8068 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8069 * config/nds32/nds32-e8.md: New file.
8070 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8071 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8073 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8074 * config/nds32/nds32.md (pipeline_model): Add e8.
8075 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8076 * config/nds32/pipelines.md: Include e8 settings.
8078 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8079 Chung-Ju Wu <jasonwucj@gmail.com>
8081 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8082 * config/nds32/nds32-n8.md: New file.
8083 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8084 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8086 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8087 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8088 * config/nds32/nds32.md (pipeline_model): Add n8.
8089 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8090 * config/nds32/pipelines.md: Include n8 settings.
8092 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8093 Chung-Ju Wu <jasonwucj@gmail.com>
8095 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8096 * config/nds32/nds32-n9-2r1w.md: New file.
8097 * config/nds32/nds32-n9-3r2w.md: New file.
8098 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8099 nds32_register_ports): New or modify for cpu n9.
8100 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8102 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8103 * config/nds32/nds32-utils.c: New file.
8104 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8105 TARGET_MUL_SLOW): Define.
8106 * config/nds32/nds32.md (pipeline_model): New attribute.
8107 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8108 New options that support cpu n9.
8109 * config/nds32/pipelines.md: Include n9 settings.
8110 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8112 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8114 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8115 information if necessary.
8116 (output_cond_branch_compare_zero): Likewise.
8117 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8118 (nds32_target_alignment): Refine for alignment.
8119 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8120 (FUNCTION_BOUNDARY): Modify.
8121 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8123 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8125 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8127 * config/nds32/constants.md (unspec_volatile_element): Add values for
8128 TLB operation and data prefetch.
8129 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8130 functions for TLB operation and data prefetch.
8131 * config/nds32/nds32-intrinsic.md: Likewise.
8132 * config/nds32/nds32_intrinsic.h: Likewise.
8133 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8134 (nds32_print_operand): Likewise.
8135 * config/nds32/nds32.h (nds32_builtins): Likewise.
8137 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8138 Andrew Pinski <pinsika@gcc.gnu.org>
8141 * match.pd: Use constant_boolean_node of correct type instead of
8142 boolean_true_node or boolean_false_node for simplifying
8143 pointer comparisons to zero.
8145 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8147 PR tree-optimization/80021
8148 * tree.c (verify_type_variant): Make error call in verify_variant_match
8149 translatable and remove final full stop.
8151 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8153 * config/nds32/constants.md (unspec_volatile_element): Add
8154 UNSPEC_VOLATILE_EH_RETURN.
8155 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8156 nds32_output_stack_pop): Support dwarf exception handling process.
8157 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8158 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8159 exception handling process.
8160 (nds32_compute_stack_frame): Likewise.
8161 (nds32_return_addr_rtx): Likewise.
8162 (nds32_initial_elimination_offset): Likewise.
8163 (nds32_expand_prologue): Likewise.
8164 (nds32_expand_epilogue): Likewise.
8165 (nds32_dynamic_chain_address): New function.
8166 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8168 (DYNAMIC_CHAIN_ADDRESS): Define.
8169 (EH_RETURN_DATA_REGNO): Define.
8170 (EH_RETURN_STACKADJ_RTX): Define.
8171 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8172 patterns for dwarf exception handling.
8174 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8176 * config/nds32/nds32.h: Clean up obsolete macros.
8178 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8180 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8181 Add enum values for particular instructions.
8182 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8183 particular intrinsic functions.
8184 * config/nds32/nds32-intrinsic.md: Likewise.
8185 * config/nds32/nds32_intrinsic.h: Likewise.
8186 * config/nds32/nds32.h (nds32_builtins): Likewise.
8187 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8188 (btst, ave): New patterns for particular instructions.
8190 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8192 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8193 Add enum values for atomic load/store and memory sync.
8194 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8196 * config/nds32/nds32-intrinsic.md: Likewise.
8197 * config/nds32/nds32_intrinsic.h: Likewise.
8198 * config/nds32/nds32.h (nds32_builtins): Likewise.
8200 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8202 PR tree-optimization/85257
8203 * fold-const.c (native_encode_vector): If not all elts could fit
8204 and off is -1, return 0 rather than offset.
8205 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8206 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8207 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8208 adjust buffer in native_interpret_expr call.
8210 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8212 * config/nds32/constants.md (unspec_volatile_element): Add cache
8213 control enum values.
8214 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8215 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8216 * config/nds32/nds32.c (nds32_cctl_names): New.
8217 (nds32_print_operand): Handle cache control register names.
8218 * config/nds32/nds32.h (nds32_builtins): New enum values.
8219 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8221 * config/nds32/nds32.md (type): Add mmu.
8222 * config/nds32/pipelines.md (simple_insn): Add mmu.
8224 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8226 * config/nds32/nds32.md (type): Remove call.
8227 * config/nds32/pipelines.md (simple_insn): Likewise.
8229 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8231 * config/nds32/constants.md (unspec_volatile_element): Add
8232 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8233 UNSPEC_VOLATILE_FMFCFG.
8234 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8235 description for fmfcfg and fmfcsr.
8236 (bdesc_1arg): Add fmtcsr.
8237 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8238 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8239 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8240 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8241 unspec_fmfcfg): New patterns.
8242 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8243 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8244 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8245 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8246 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8247 __nds32__fmfcfg): Define.
8249 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8251 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8252 intrinsic register names.
8253 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8254 intrinsic register enum values and macros.
8256 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8258 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8259 for load/store addressing form.
8260 (nds32_print_operand_address): Likewise.
8262 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8265 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8266 based on LABEL_REF. Remove useless assertion.
8267 (pic_address_needs_scratch): Fix formatting.
8268 (sparc_legitimize_pic_address): Minor tweaks.
8269 (sparc_delegitimize_address): Adjust assertion accordingly.
8270 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8271 into symbolic_operand.
8272 (movsi_high_pic_label_ref): Likewise.
8273 (movsi_lo_sum_pic_label_ref): Likewise.
8274 (movdi_pic_label_ref): Likewise.
8275 (movdi_high_pic_label_ref): Likewise.
8276 (movdi_lo_sum_pic_label_ref): Likewise.
8278 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8280 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8281 custom LIB_SPEC setup.
8283 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8284 Kito Cheng <kito.cheng@gmail.com>
8286 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8287 * config/riscv/freebsd.h: New.
8289 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8291 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8292 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8295 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8296 Kito Cheng <kito.cheng@gmail.com>
8298 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8299 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8300 * config/nds32/nds32-protos.h (nds32_output_call,
8301 nds32_output_return): Declare.
8302 * config/nds32/nds32.md: Refine all the call and return patterns.
8304 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8307 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8308 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8310 PR rtl-optimization/84872
8311 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8312 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8315 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8317 * expr.c (copy_blkmode_to_reg): Revert 254862.
8318 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8320 2018-04-06 Richard Biener <rguenther@suse.de>
8323 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8324 after seeing a component reference with an adjacent field. Treat
8325 refs to arrays at struct end of external decls similar to
8326 refs to unconstrained commons.
8328 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8331 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8332 look through SAVE_EXPRs with non-side-effects argument. Adjust
8334 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8337 2018-04-06 Richard Biener <rguenther@suse.de>
8340 * alias.c (find_base_term): New wrapper around find_base_term
8341 unwinding CSELIB_VAL_PTR changes.
8342 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8345 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8347 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8349 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8350 constant definitions.
8351 ("nop"): lr 0,0 -> nopr r0
8352 ("nop_lr0", "nop_lr1"): New insn definitions.
8354 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8356 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8357 NDS32_V3PUSH_AVAILABLE_P macro.
8359 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8360 Chung-Ju Wu <jasonwucj@gmail.com>
8362 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8363 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8364 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8365 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8366 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8367 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8368 * config/nds32/constraints.md: New constraints and checking for hard
8369 float configuration.
8370 * config/nds32/iterators.md: New mode iterator and attribute for hard
8371 float configuration.
8372 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8374 * config/nds32/nds32-fpu.md: New file.
8375 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8376 deal with hard float code generation.
8377 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8379 (abi_type, float_reg_number): New enum type.
8380 * config/nds32/nds32-predicates.c: New predicates for hard float.
8381 * config/nds32/nds32-protos.h: Declare functions for hard float.
8382 * config/nds32/nds32.c: Implementation for hard float configuration.
8383 * config/nds32/nds32.h: Definitions for hard float configuration.
8384 * config/nds32/nds32.md: Include hard float machine description and
8385 modify patterns for hard float configuration.
8386 * config/nds32/nds32.opt: New options for hard float configuration.
8387 * config/nds32/predicates.md: New predicates for hard float
8390 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8392 * common/config/nds32/nds32-common.c
8393 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8395 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8398 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8399 CONSTRUCTOR_ELT (ctor, ...)->value.
8401 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8404 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8406 2018-04-05 Tom de Vries <tom@codesourcery.com>
8409 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8412 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8413 Kito Cheng <kito.cheng@gmail.com>
8415 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8416 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8417 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8418 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8420 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8421 Kito Cheng <kito.cheng@gmail.com>
8423 * config/nds32/constraints.md (Ufe): New memory constraint.
8424 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8425 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8426 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8428 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8429 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8431 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8433 * config/nds32/nds32.md: Use optimize_size in the condition for
8434 alu-shift instructions.
8436 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8438 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8440 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8442 * config/nds32/nds32.md (negsi2): Refine pattern.
8444 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8445 Chung-Ju Wu <jasonwucj@gmail.com>
8447 * config/nds32/iterators.md (shift_rotate): New code iterator.
8448 (shift): New code attribute.
8449 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8450 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8451 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8452 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8453 bit-wise operations.
8454 (andsi3, *andsi3): Ditto.
8455 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8456 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8457 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8458 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8459 nds32_ior_operand, nds32_xor_operand): New predicates.
8461 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8463 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8464 (addsi3, subsi3): ... this.
8466 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8468 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8470 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8472 * config/nds32/nds32.md: Adjust indention.
8474 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8476 * config/nds32/nds32.md (feature): New attribute.
8478 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8480 * config/nds32/nds32.md (subtype): New attribute.
8482 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8485 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8486 expansion to perform a bitwise AND of the argument followed by a
8487 boolean negation of the result.
8489 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8491 PR rtl-optimization/84878
8492 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8493 the basic block. Assert the use reference is not artificial and that
8494 it has an associated insn.
8496 2018-04-04 Michael Matz <matz@suse.de>
8498 * builtins.c (compute_objsize): Pass correct operand
8499 to array_at_struct_end_p.
8501 2018-04-04 Richard Biener <rguenther@suse.de>
8504 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8505 from contexts for DINFO_LEVEL_TERSE and below.
8507 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8509 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8510 resiter_operand condition.
8511 * config/nds32/nds32.md (*move<mode>): Ditto.
8513 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8514 Monk Chiang <sh.chiang04@gmail.com>
8516 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8518 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8520 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8522 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8523 Kito Cheng <kito.cheng@gmail.com>
8525 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8526 nds32_cond_code_str, output_cond_branch,
8527 output_cond_branch_compare_zero, nds32_expand_cbranch,
8528 nds32_expand_cstore, nds32_expand_movcc,
8529 nds32_output_cbranchsi4_equality_zero,
8530 nds32_output_cbranchsi4_equality_reg,
8531 nds32_output_cbranchsi4_equality_reg_or_const_int,
8532 nds32_output_cbranchsi4_greater_less_zero: New functions.
8533 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8534 nds32_expand_cstore, nds32_expand_movcc,
8535 nds32_output_cbranchsi4_equality_zero,
8536 nds32_output_cbranchsi4_equality_reg,
8537 nds32_output_cbranchsi4_equality_reg_or_const_int,
8538 nds32_output_cbranchsi4_greater_less_zero): Declare.
8539 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8540 nds32_rimm11s_operand): New predicates.
8541 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8542 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8545 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8547 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8548 * config/nds32/nds32.md: Ditto.
8549 * config/nds32/pipelines.md: Ditto.
8551 2018-04-04 Richard Biener <rguenther@suse.de>
8553 PR tree-optimization/85168
8554 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8555 propagating abnormals.
8557 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8559 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8561 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8562 Kito Cheng <kito.cheng@gmail.com>
8564 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8565 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8566 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8567 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8568 * config/nds32/nds32.md (sibcall_internal): New.
8569 (sibcall_register): Remove.
8570 (sibcall_immediate): Remove.
8571 (sibcall_value_internal): New.
8572 (sibcall_value_register): Remove.
8573 (sibcall_value_immediate): Remove.
8574 * config/nds32/predicates.md (nds32_general_register_operand): New.
8575 (nds32_call_address_operand): New.
8577 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8579 PR rtl-optimization/85167
8580 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8581 bb_defs if *split_p, instead preinitialize it to NULL.
8583 PR tree-optimization/85156
8584 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8585 evaluating the argument multiple times.
8587 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8589 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8591 (_mm_cvtpd_ps): Likewise.
8592 (_mm_cvttpd_epi32): Likewise.
8593 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8594 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8595 vector, pixel, and bool following altivec.h include.
8597 2018-04-03 Martin Sebor <msebor@redhat.com>
8599 * doc/extend.texi (Common Function Attributes): Clarify.
8600 (const attribute): Likewise.
8601 (pure attribute): Likewise.
8603 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8606 * config/i386/i386.c (ix86_expand_vector_set): Use
8607 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8609 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8611 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8612 instructions when changing rounding bits to preserve precision bits
8613 in the x87 control word.
8615 2018-04-03 Martin Liska <mliska@suse.cz>
8617 PR tree-optimization/82491
8618 * rtl.h (strip_offset_and_add): Replace += suboffset with
8619 poly_uint64 () + suboffset.
8621 2018-03-29 Martin Liska <mliska@suse.cz>
8622 Martin Jambor <mjambor@suse.cz>
8625 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8626 param_type is not an integral or pointer type.
8628 2018-04-03 Richard Biener <rguenther@suse.de>
8630 * sese.h (recompute_all_dominators): Remove.
8632 2018-04-02 Martin Sebor <msebor@redhat.com>
8634 * doc/invoke.texi (-Wrestrict): Fix typos.
8636 2018-04-02 Jim Wilson <jimw@sifive.com>
8638 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8639 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8640 (<optab>di3, <optab>si3_extend): Likewise.
8641 (<optab>si3_mask, <optab>si3_mask_1): New.
8642 (<optab>di3_mask, <optab>di3_mask_1): New.
8643 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8644 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8645 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8647 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8649 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8652 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8654 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8655 (nds32_canonicalize_comparison): New function.
8657 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8658 Kito Cheng <kito.cheng@gmail.com>
8659 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8661 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8662 * config/nds32/constants.md (unspec_volatile_element): Add
8663 UNSPEC_VOLATILE_RELAX_GROUP.
8664 * config/nds32/nds32-relax-opt.c: New file.
8665 * config/nds32/nds32-predicates.c
8666 (nds32_symbol_load_store_p): New function.
8667 * config/nds32/nds32-protos.h
8668 (nds32_symbol_load_store_p): Declare function.
8669 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8670 * config/nds32/nds32.c
8671 (nds32_register_pass): New function to register pass.
8672 (nds32_register_passes): New function to register passes.
8673 * config/nds32/nds32.md (relax_group): New pattern.
8674 * config/nds32/nds32.opt (mrelax-hint): New option.
8675 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8677 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8679 * config/nds32/t-nds32: Modify files dependency.
8681 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8683 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8684 (PROFILE_HOOK): Define its implementation.
8686 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8688 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8689 type and 32-bit size.
8691 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8694 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8695 (V_128_256): New mode iterator.
8696 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8697 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8698 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8700 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8701 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8703 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8706 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8707 NaN inputs correctly.
8709 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8712 * config/rs6000/vsx.md (??r): New mode attribute.
8713 (*vsx_mov<mode>_64bit): Use it.
8714 (*vsx_mov<mode>_32bit): Likewise.
8716 2018-03-30 Martin Sebor <msebor@redhat.com>
8718 PR tree-optimization/84818
8719 * builtins.c (check_access): Use warning_n.
8721 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8724 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8726 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8729 2018-03-30 Julia Koval <julia.koval@intel.com>
8732 * x86-tune.def (movx, partial_reg_dependency): Enable for
8735 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8738 * lra-constraints.c (process_alt_operands): Move setting
8739 this_alternative_matches below.
8741 2018-03-29 Martin Liska <mliska@suse.cz>
8744 * doc/invoke.texi: Document how LTO works with debug info.
8745 Describe auto-load support of binutils. Mention 'x86-64'
8746 as valid option value of -march option.
8748 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8750 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8753 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8754 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8755 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8758 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8761 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8762 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8763 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8764 for RS6000_BTM_POWERPC64.
8765 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8766 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8767 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8772 2018-03-28 Carl Love <cel@us.ibm.com>
8775 2017-09-27 Carl Love <cel@us.ibm.com>
8777 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8778 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8779 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8782 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8784 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8785 instead of __vector bool.
8786 (_mm_max_pu8): Likewise.
8787 (_mm_min_pi16): Likewise.
8789 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8792 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8793 (DIVWEUO): Likewise.
8795 (DIVDEUO): Likewise.
8796 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8797 DIVWEUO and DIVDEUO.
8798 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8799 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8800 (div_extend): Likewise.
8801 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8803 (__builtin_divweuo): Likewise.
8804 (__builtin_divdeo): Likewise.
8805 (__builtin_divdeuo): Likewise.
8807 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8810 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8811 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8813 PR tree-optimization/82004
8814 * gimple-match-head.c (optimize_pow_to_exp): New function.
8815 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8816 Don't fold to exp if optimize_pow_to_exp is false.
8818 2018-03-28 Martin Liska <mliska@suse.cz>
8821 * calls.c (initialize_argument_information): Fix trailing space.
8822 * common.opt: Fix typo and provide better explanation for
8823 -fsanitize-coverage option.
8824 * config/i386/i386.opt: Fix typo.
8826 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8827 Martin Liska <mliska@suse.cz>
8830 * gimplify.c (asan_poison_variable): Don't do the check for
8831 gimplify_omp_ctxp here.
8832 (gimplify_decl_expr): Do it here.
8833 (gimplify_target_expr): Likewise.
8835 2018-03-28 Martin Liska <mliska@suse.cz>
8838 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8839 chkp_type_bounds_count if MPX is not enabled.
8841 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8843 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8845 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8848 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8849 function to create the function decl for complex long double
8850 multiply and divide for -mabi=ieeelongdouble.
8851 (init_float128_ieee): Call it.
8853 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8856 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8857 -fcf-protection=branch -mibt.
8858 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8860 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8863 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8864 arm_word_relocations.
8866 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8869 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8870 extern array declarations.
8872 2018-03-27 Richard Biener <rguenther@suse.de>
8875 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8876 explicit single_use checks.
8878 2018-03-27 Richard Biener <rguenther@suse.de>
8880 PR tree-optimization/85082
8881 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8884 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8886 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8887 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8888 Turn on fasynchronous-unwind-tables and funwind-tables.
8890 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8893 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8894 (*bmi_blsr_<mode>_ccz): Ditto.
8896 2018-03-26 Tom de Vries <tom@codesourcery.com>
8898 PR tree-optimization/85063
8899 * omp-general.c (offloading_function_p): New function. Factor out
8901 * omp-offload.c (pass_omp_target_link::gate): ... here.
8902 * omp-general.h (offloading_function_p): Declare.
8903 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8904 with attribute omp declare target for offloading functions.
8906 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8908 PR tree-optimization/84005
8909 * tree-data-ref.h (get_base_for_alignment): Declare.
8910 * tree-data-ref.c (get_base_for_alignment_1): New function.
8911 (get_base_for_alignment): Likewise.
8912 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8913 get_base_for_alignment to find a suitable base object, instead
8914 of always using drb->base_address.
8916 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8919 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8920 known size by default.
8922 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8925 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8926 and non BLKmode operands.
8928 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8931 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8932 Clean up attributes.
8934 2018-03-23 Richard Biener <rguenther@suse.de>
8937 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8938 we are going to emit early debug for LTO.
8940 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8943 * function.c (match_asm_constraints_1): Don't optimize if input
8944 doesn't satisfy general_operand predicate for output's mode.
8947 * alias.c (write_dependence_p): Don't require for x_canonicalized
8948 non-VOIDmode if x has VOIDmode.
8951 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8952 just don't try to optimize it rather than assert it never happens.
8954 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8956 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8957 macro expansions for definition of ST_INTERNAL_<mode> and
8958 LD_INTERNAL_<mode> builtins.
8959 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8961 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8963 (altivec_expand_st_builtin): Likewise.
8964 (altivec_expand_builtin): Remove calls to deleted functions.
8965 (rs6000_address_for_altivec): Delete this function.
8966 * config/rs6000/vector.md: Remove expands for
8967 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8969 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8972 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8973 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8974 re-computing once computed.
8975 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8976 (arm_init_machine_status): Initialize
8977 machine->static_chain_stack_bytes.
8979 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8982 * doc/extend.texi: Add four new prototypes for vec_ld.
8983 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8984 definitions for more logical presentation.
8985 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8986 entries for V1TI variants of __builtin_altivec_ld builtin.
8987 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8988 handling of V1TI variant of LVX icode pattern.
8989 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8990 (rs6000_gimple_fold_builtin): Likewise.
8991 (altivec_init_builtins): Add code to define
8992 __builtin_altivec_lvx_v1ti function.
8994 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8997 * function.c (match_asm_constraints_1): Don't do the optimization
8998 if input isn't a REG, SUBREG, MEM or constant.
9000 2018-03-22 Tom de Vries <tom@codesourcery.com>
9002 PR tree-optimization/84956
9003 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9004 bb_has_abnormal_pred.
9006 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9009 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9010 DECL_INITIAL (decl) to decl at the end.
9011 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9014 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9016 * doc/extend.texi (__builtin_tgmath): Document when complex
9017 integer types are treated as _Complex _Float64.
9019 2018-03-21 Tom de Vries <tom@codesourcery.com>
9021 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9023 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9025 PR tree-optimization/84960
9026 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9027 if it is ENTRY block, move them into single succ of ENTRY in that case.
9029 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9031 PR tree-optimization/84811
9032 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9033 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9034 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9035 (known_subrange_p): Update accordingly. Cast each value involved
9036 in the size comparison, rather than casting the result of the
9039 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9041 PR tree-optimization/84982
9042 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9043 by flipping the least significant bit rather than all bits from
9044 bitpos to bitpos + bitsize - 1.
9046 2018-03-21 Nathan Sidwell <nathan@acm.org>
9048 * doc/extend.texi (Deprecated Features): Remove mention of
9049 long-deleted deprecations.
9051 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9054 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9055 * configure: Regenerate.
9057 2018-03-21 Tom de Vries <tom@codesourcery.com>
9059 PR tree-optimization/83126
9060 * tree-parloops.c (num_phis): New function.
9061 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9063 2018-03-21 Nathan Sidwell <nathan@acm.org>
9065 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9066 mention anon-struct/union members and trailing attributes.
9068 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9070 PR tree-optimization/84969
9071 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9072 builtin memset partitions if they set different rhs values.
9074 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9076 PR rtl-optimization/84989
9077 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9078 VEC_DUPLICATE with scalar result mode.
9080 2018-03-21 Martin Liska <mliska@suse.cz>
9083 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9084 not intended return statement.
9086 2018-03-21 Martin Liska <mliska@suse.cz>
9089 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9090 (chkp_find_bound_slots_1): Limit number of iterations.
9092 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9095 * Minor grammar fixes for x86 options.
9097 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9100 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9101 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9103 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9106 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9107 (altivec_lvx_<mode>_1op): Likewise.
9108 (altivec_stvx_<mode>_2op): Likewise.
9109 (altivec_stvx_<mode>_1op): Likewise.
9110 (altivec_lvx_<VM2:mode>): New define_expand.
9111 (altivec_stvx_<VM2:mode>): Likewise.
9112 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9113 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9114 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9115 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9116 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9117 (rs6000_gen_lvx): Likewise.
9118 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9119 (altivec_expand_stv_builtin): Likewise.
9120 (altivec_expand_builtin): Likewise.
9121 * config/rs6000/vector.md: Likewise.
9123 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9126 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9129 2018-03-20 Richard Biener <rguenther@suse.de>
9132 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9133 sign-conversions as zero, fall back to standard scalar_stmt
9136 2018-03-20 Martin Liska <mliska@suse.cz>
9139 * predict.c (rebuild_frequencies): Handle case when we have
9140 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9142 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9145 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9146 flag_section_anchors.
9147 * varasm.c (use_blocks_for_decl_p): Remove hack for
9148 dw2_force_const_mem.
9151 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9153 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9154 be created, use lowpart_subreg of operands[0] rather than operands[0]
9156 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9157 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9158 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9159 and n constraint instead of aarch64_shift_imm_di and Usd.
9160 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9161 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9163 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9166 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9167 to favor GPR over NEON registers.
9168 (<shift>di3_neon): Likewise.
9170 2018-03-20 Tom de Vries <tom@codesourcery.com>
9173 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9174 (nvptx_process_pars): Emit bar.sync asap and alap.
9176 2018-03-20 Tom de Vries <tom@codesourcery.com>
9179 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9180 seen_label if seen_label is already set.
9182 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9185 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9186 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9187 Use 1U instead of 1. Formatting fixes.
9190 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9191 instead of TREE_TYPE (s1) for the return value.
9193 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9195 PR tree-optimization/84946
9196 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9197 bitsize + bitsize in poly_uint64 rather than poly_int64.
9200 * dwarf2asm.c: Include fold-const.c.
9201 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9202 of decl rather than decl itself.
9204 PR rtl-optimization/84643
9205 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9207 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9210 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9211 calling assemble_variable.
9213 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9216 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9217 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9219 2018-03-19 Jim Wilson <jimw@sifive.com>
9222 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9223 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9224 (riscv_first_stack_step): Likewise.
9225 (riscv_option_override): Use STACK_BOUNDARY instead of
9227 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9229 (BIGGEST_ALIGNMENT): Set to 128.
9230 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9231 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9234 2018-03-19 Richard Biener <rguenther@suse.de>
9236 PR tree-optimization/84933
9237 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9238 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9240 2018-03-19 Richard Biener <rguenther@suse.de>
9242 PR tree-optimization/84859
9243 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9244 (cond_if_else_store_replacement): Perform sinking operation on
9245 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9246 Generalize what a BB with a single eligible store is.
9248 2018-03-19 Richard Biener <rguenther@suse.de>
9250 PR tree-optimization/84929
9251 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9252 chrec_is_positive against non-chrec arg.
9254 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9257 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9259 2018-03-18 Martin Liska <mliska@suse.cz>
9261 PR rtl-optimization/84635
9262 * regrename.c (build_def_use): Use matches_mode only when
9265 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9267 PR tree-optimization/84913
9268 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9269 vectorize chains of COND_EXPRs.
9271 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9273 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9275 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9277 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9279 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9281 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9283 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9284 Kito Cheng <kito.cheng@gmail.com>
9286 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9287 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9288 (nds32_adjust_reg_alloc_order): New function.
9289 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9291 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9293 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9294 nds32_print_operand, nds32_print_operand_address): Use
9295 HOST_WIDE_INT_PRINT_DEC instead.
9297 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9299 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9301 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9304 * config/i386/i386.c (initial_ix86_tune_features,
9305 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9307 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9308 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9309 rather than 1u << ix86_tune. Formatting fix.
9310 (ix86_option_override_internal): Change ix86_arch_mask from
9311 unsigned int to unsigned HOST_WIDE_INT, initialize to
9312 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9313 (ix86_function_specific_restore): Likewise.
9315 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9318 * postreload.c (reload_combine_recognize_pattern): Perform
9319 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9320 truncate_int_for_mode the result for the destination's mode.
9323 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9324 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9326 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9329 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9332 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9333 regno_allocno_class_array and sorted_pseudos.
9334 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9335 insns where regno is used.
9337 2018-03-16 Martin Liska <mliska@suse.cz>
9340 * multiple_target.c (create_dispatcher_calls): Redirect
9341 reference in the symbol table.
9343 2018-03-16 Martin Liska <mliska@suse.cz>
9346 * multiple_target.c (create_dispatcher_calls): Redirect also
9349 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9353 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9354 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9355 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9357 2018-03-16 Julia Koval <julia.koval@intel.com>
9359 * doc/invoke.texi (Skylake Server): Add CLWB.
9360 Cannonlake): Remove CLWB.
9362 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9364 PR tree-optimization/84841
9365 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9367 (FLOAT_ONE_CONST_TYPE): Define.
9368 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9369 (sort_by_operand_rank): Put entries with higher constant_type last
9370 rather than first to match comments.
9372 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9374 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9377 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9380 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9381 check and use address of target always.
9383 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9386 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9387 (indirect_thunk_bnd_needed): Likewise.
9388 (indirect_thunks_used): Likewise.
9389 (indirect_thunks_bnd_used): Likewise.
9390 (indirect_return_needed): New.
9391 (indirect_return_bnd_needed): Likewise.
9392 (output_indirect_thunk_function): Add a bool argument for
9394 (output_indirect_thunk_function): Don't generate alias for
9395 function return thunk.
9396 (ix86_code_end): Call output_indirect_thunk_function to generate
9397 function return thunks.
9398 (ix86_output_function_return): Set indirect_return_bnd_needed
9399 and indirect_return_needed instead of indirect_thunk_bnd_needed
9400 and indirect_thunk_needed.
9402 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9404 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9405 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9406 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9408 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9409 Paul Hua <paul.hua.gm@gmail.com>
9412 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9414 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9416 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9417 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9420 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9423 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9424 instead of GET_MODE_SIZE when comparing Units.
9426 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9429 * varasm.c (hash_section): Return an unchangeble hash value
9430 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9431 Return !aarch64_can_use_per_function_literal_pools_p ().
9433 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9436 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9437 rather than address of cmode as last argument to prepare_cmp_insn.
9439 2018-03-15 Julia Koval <julia.koval@intel.com>
9441 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9442 F_AVX512VNNI, F_AVX512BITALG): New.
9444 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9447 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9448 insn for floating-point loads and stores.
9450 2018-03-14 Carl Love <cel@us.ibm.com>
9452 * config/rs6000/rs6000-c.c: Add macro definitions for
9453 ALTIVEC_BUILTIN_VEC_PERMXOR.
9454 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9455 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9456 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9457 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9459 * config/doc/extend.texi: Add prototypes for vec_permxor.
9461 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9464 * diagnostic-show-locus.c (class layout_point): Convert m_line
9465 from int to linenum_type.
9466 (line_span::comparator): Use linenum "compare" function when
9467 comparing line numbers.
9468 (test_line_span): New function.
9469 (layout_range::contains_point): Convert param "row" from int to
9471 (layout_range::intersects_line_p): Likewise.
9472 (layout::will_show_line_p): Likewise.
9473 (layout::print_source_line): Likewise.
9474 (layout::should_print_annotation_line_p): Likewise.
9475 (layout::print_annotation_line): Likewise.
9476 (layout::print_leading_fixits): Likewise.
9477 (layout::annotation_line_showed_range_p): Likewise.
9478 (struct line_corrections): Likewise for field m_row.
9479 (line_corrections::line_corrections): Likewise for param "row".
9480 (layout::print_trailing_fixits): Likewise.
9481 (layout::get_state_at_point): Likewise.
9482 (layout::get_x_bound_for_row): Likewise.
9483 (layout::print_line): Likewise.
9484 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9485 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9486 * input.c (selftest::test_linenum_comparisons): New function.
9487 (selftest::input_c_tests): Call it.
9488 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9489 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9490 * selftest.h (ASSERT_GT): New macro.
9491 (ASSERT_GT_AT): New macro.
9492 (ASSERT_LT): New macro.
9493 (ASSERT_LT_AT): New macro.
9495 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9497 PR rtl-optimization/84780
9498 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9500 2018-03-14 Martin Liska <mliska@suse.cz>
9502 * tree.c (record_node_allocation_statistics): Use
9503 get_stats_node_kind.
9504 (get_stats_node_kind): New function extracted from
9505 record_node_allocation_statistics.
9506 (free_node): Use get_stats_node_kind.
9508 2018-03-14 Richard Biener <rguenther@suse.de>
9510 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9511 that the value-set of ANTIC_IN doesn't grow.
9514 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9516 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9517 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9519 2018-03-14 Julia Koval <julia.koval@intel.com>
9521 * config.gcc (icelake-client, icelake-server): New.
9523 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9524 (initial_ix86_arch_features): Ditto.
9525 (PTA_SKYLAKE): Add SGX.
9526 (PTA_ICELAKE): Remove.
9527 (PTA_ICELAKE_CLIENT): New.
9528 (PTA_ICELAKE_SERVER): New.
9529 (ix86_option_override_internal): Split up icelake on icelake client and
9531 (get_builtin_code_for_version): Ditto.
9532 (fold_builtin_cpu): Ditto.
9533 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9534 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9535 * config/i386/i386.h (processor_type): Ditto.
9536 * doc/invoke.texi: Ditto.
9538 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9541 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9542 INTEGER_CST offset, add it together with bitpos / 8 and
9543 sign extend based on POINTER_SIZE.
9547 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9550 * config/i386/constraints.md (Yc): New register constraint.
9551 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9552 Use Yc constraint for alternative 2 of operand 0. Remove
9553 preferred_for_speed attribute.
9555 2018-03-14 Richard Biener <rguenther@suse.de>
9557 PR tree-optimization/84830
9558 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9559 with the old one to avoid oscillations.
9561 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9564 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9566 (assign_by_spills): Return a flag of reload assignment failure.
9567 Do not process the reload assignment failures. Do not spill other
9568 reload pseudos if they has the same reg class. Update n if
9570 (lra_assign): Add a return arg. Set up from the result of
9571 assign_by_spills call.
9572 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9573 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9574 usage_insns if it is not NULL.
9575 (spill_hard_reg_in_range): New function.
9576 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9577 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9578 function prototypes.
9579 (lra_assign): Change prototype.
9580 * lra.c (lra): Add code to deal with fails by splitting hard reg
9583 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9585 * config/riscv/riscv.opt (mrelax): New option.
9586 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9587 "norelax" when riscv_mrelax is disabled.
9588 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9590 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9593 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9594 reassociation for int modes.
9596 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9598 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9599 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9601 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9602 * config/aarch64/aarch64-sve.md
9603 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9604 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9605 (*extend<mode><Vwide>2): Rename to...
9606 (aarch64_sve_extend<mode><Vwide>2): ...this.
9607 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9608 renaming the old pattern to...
9609 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9611 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9612 define_expand, renaming the old pattern to...
9613 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9614 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9615 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9616 account when deciding which SVE instruction the optab should use.
9617 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9619 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9621 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9622 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9623 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9624 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9625 (tlsdesc_small_<mode>): Turn a define_expand and use
9626 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9627 (tlsdesc_small_advsimd_<mode>): ...this.
9628 (tlsdesc_small_sve_<mode>): New pattern.
9630 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9632 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9633 (UNSPEC_UMUL_HIGHPART): New constants.
9634 (MUL_HIGHPART): New int iteraor.
9635 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9636 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9638 (*<su>mul<mode>3_highpart): New define_insn.
9640 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9643 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9646 2018-03-13 Martin Liska <mliska@suse.cz>
9649 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9651 (sem_item_optimizer::~sem_item_optimizer): Release it.
9652 (sem_item_optimizer::merge_classes): Register variable aliases.
9653 (sem_item_optimizer::fixup_pt_set): New function.
9654 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9655 * ipa-icf.h: Declare new variables and functions.
9657 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9660 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9661 integer_pow2p@2 and test integer_pow2p in condition.
9662 (A < 0 ? C : 0): Similarly for @1.
9665 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9666 characters starting at p contain '\0' character, don't look beyond
9670 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9671 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9674 * reg-stack.c (change_stack): Change update_end var from int to
9675 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9676 also call set_block_for_insn on the newly added insns and rescan.
9679 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9680 on the last operand.
9683 * tree.c (stabilize_reference_1): Return save_expr (e) for
9684 STATEMENT_LIST even if it doesn't have side-effects.
9686 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9688 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9690 2018-03-12 Renlin Li <renlin.li@arm.com>
9692 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9693 aarch64_output_scalar_simd_mov_immediate.
9695 2018-03-12 Martin Sebor <msebor@redhat.com>
9697 PR tree-optimization/83456
9698 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9699 for perfectly overlapping calls to memcpy.
9700 (gimple_fold_builtin_memory_chk): Same.
9701 (gimple_fold_builtin_strcpy): Handle no-warning.
9702 (gimple_fold_builtin_stxcpy_chk): Same.
9703 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9705 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9707 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9708 parameter. Use it for SFmode.
9709 (rs6000_function_arg_advance_1): Adjust.
9710 (rs6000_function_arg): Adjust.
9711 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9713 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9715 PR rtl-optimization/84169
9716 PR rtl-optimization/84780
9717 * combine.c (can_combine_p): Check for a 2-insn combination whether
9718 the destination register is used between the two insns, too.
9720 2018-03-12 Richard Biener <rguenther@suse.de>
9722 PR tree-optimization/84803
9723 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9724 for refs DR analysis didn't process.
9726 2018-03-12 Richard Biener <rguenther@suse.de>
9728 PR tree-optimization/84777
9729 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9730 force-vectorize loops ignore whether we are optimizing for size.
9732 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9734 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9735 (TARGET_MD_ASM_ADJUST): Define.
9737 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9738 Kito Cheng <kito.cheng@gmail.com>
9739 Chung-Ju Wu <jasonwucj@gmail.com>
9741 * config/nds32/nds32.c (nds32_compute_stack_frame,
9742 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9743 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9744 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9745 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9746 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9747 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9748 * config/nds32/nds32.md (prologue, epilogue): Use macro
9749 NDS32_V3PUSH_AVAILABLE_P to do checking.
9751 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9754 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9755 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9756 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9757 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9758 addition of most attributes on !orig_type_die or the attribute not
9759 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9761 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9762 Chung-Ju Wu <jasonwucj@gmail.com>
9764 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9766 * config/nds32/nds32.opt (mvh): New option.
9768 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9769 Chung-Ju Wu <jasonwucj@gmail.com>
9771 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9773 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9774 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9777 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9778 Chung-Ju Wu <jasonwucj@gmail.com>
9780 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9782 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9783 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9785 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9786 Kito Cheng <kito.cheng@gmail.com>
9787 Chung-Ju Wu <jasonwucj@gmail.com>
9789 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9790 UNSPEC_FFMISM and UNSPEC_FLMISM.
9791 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9792 for ffb, ffmism and flmism.
9793 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9794 (unspec_ffmism): Ditto.
9795 (unspec_flmism): Ditto.
9796 (nds32_expand_builtin_impl): Check if string extension is available.
9797 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9798 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9800 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9803 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9806 * lra-assigns.c (assign_by_spills): Return a flag of reload
9807 assignment failure. Do not process the reload assignment
9808 failures. Do not spill other reload pseudos if they has the same
9810 (lra_assign): Add a return arg. Set up from the result of
9811 assign_by_spills call.
9812 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9813 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9814 usage_insns if it is not NULL.
9815 (spill_hard_reg_in_range): New function.
9816 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9817 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9818 function prototypes.
9819 (lra_assign): Change prototype.
9820 * lra.c (lra): Add code to deal with fails by splitting hard reg
9823 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9826 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9828 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9831 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9832 (dw_val_node): Add val_symbolic_view.
9833 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9834 (symview_upper_bound): New.
9835 (new_line_info_table): Initialize symviews_since_reset.
9836 (dwarf2out_source_line): Count symviews_since_reset and set
9837 symview_upper_bound.
9838 (dw_val_equal_p): Handle symview.
9839 (add_AT_symview): New.
9840 (print_dw_val): Handle symview.
9841 (attr_checksum, attr_checksum_ordered): Likewise.
9842 (same_dw_val_p, size_of_die): Likewise.
9843 (value_format, output_die): Likewise.
9844 (add_high_low_attributes): Use add_AT_symview for entry_view.
9845 (dwarf2out_finish): Reset symview_upper_bound, clear
9848 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9851 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9852 Add strict argument and use it.
9853 (rs6000_split_multireg_move): Update for new strict argument.
9854 (mem_operand_gpr): Disallow all non-offsettable addresses.
9855 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9857 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9860 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9861 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9862 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9865 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9866 decl, use remap_type if we want to use the type.
9868 2018-03-09 Martin Sebor <msebor@redhat.com>
9870 PR tree-optimization/84526
9871 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9873 (builtin_access::generic_overlap): Be prepared to handle non-array
9876 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9878 PR rtl-optimization/84682
9879 * lra-constraints.c (process_address_1): Check is_address flag
9880 for address constraints.
9881 (process_alt_operands): Likewise.
9882 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9883 preprocess_constraints.
9884 * recog.h (preprocess_constraints): Add oploc parameter.
9886 * recog.c (preprocess_constraints): Test address_operand for
9887 CT_ADDRESS constraints.
9889 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9892 * lra-assigns.c (assign_by_spills): Return a flag of reload
9893 assignment failure. Do not process the reload assignment
9894 failures. Do not spill other reload pseudos if they has the same
9896 (lra_assign): Add a return arg. Set up from the result of
9897 assign_by_spills call.
9898 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9899 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9900 usage_insns if it is not NULL.
9901 (spill_hard_reg_in_range): New function.
9902 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9903 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9904 function prototypes.
9905 (lra_assign): Change prototype.
9906 * lra.c (lra): Add code to deal with fails by splitting hard reg
9909 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9912 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9913 Accept complain bool parameter. Only emit errors if it is true.
9914 (arm_parse_cpu_option_name): Likewise.
9915 (arm_target_thumb_only): Adjust callers of the above.
9916 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9917 prototype to take a default true bool parameter.
9918 (arm_parse_arch_option_name): Likewise.
9920 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9921 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9925 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9926 * configure: Regenerate.
9927 * configure.ac ("linker --version-script option"): New.
9928 ("linker soname option"): New.
9930 2018-03-09 Richard Biener <rguenther@suse.de>
9932 PR tree-optimization/84775
9933 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9934 immediate uses of predicate stmts and mark them modified.
9937 PR tree-optimization/84178
9938 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9940 (version_loop_for_if_conversion): Delay update_ssa call.
9941 (tree_if_conversion): Delay update_ssa until after predicate
9944 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9947 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9948 when the function accesses prior frames.
9950 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9953 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9954 gen_llsym, otherwise call maybe_gen_llsym.
9957 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9958 has ',' character inside of it.
9960 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9963 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9964 as clobbering CC_REGNUM.
9966 2018-03-08 Richard Biener <rguenther@suse.de>
9969 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9970 (follow_copies_to_constant): Do not follow SSA names registered
9973 2018-03-08 Richard Biener <rguenther@suse.de>
9975 PR tree-optimization/84178
9976 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9978 (version_loop_for_if_conversion): Delay update_ssa call.
9979 (tree_if_conversion): Delay update_ssa until after predicate
9982 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9984 PR tree-optimization/84178
9985 * tree-if-conv.c (release_bb_predicate): Remove the
9986 the assertion that the stmts have NULL use_ops.
9987 Discard the statements, asserting that they haven't
9988 yet been added to a BB.
9990 2018-03-08 Richard Biener <rguenther@suse.de>
9992 PR tree-optimization/84746
9993 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9994 (phi_translate): Pass in destination ANTIC_OUT set.
9995 (phi_translate_1): Likewise. For a simplified result lookup
9996 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9997 (phi_translate_set): Adjust.
9998 (do_pre_regular_insertion): Likewise.
9999 (do_pre_partial_partial_insertion): Likewise.
10001 2018-03-08 Martin Liska <mliska@suse.cz>
10003 PR gcov-profile/84735
10004 * doc/gcov.texi: Document usage of profile files.
10005 * gcov-io.h: Document changes in the format.
10007 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10011 * dwarf2out.c (struct dw_line_info_table): Update comments for
10013 (FORCE_RESET_NEXT_VIEW): New.
10014 (FORCE_RESETTING_VIEW_P): New.
10015 (RESETTING_VIEW_P): Check for -1 too.
10016 (ZERO_VIEW_P): Likewise.
10017 (new_line_info_table): Force-reset next view.
10018 (dwarf2out_begin_function): Likewise.
10019 (dwarf2out_source_line): Simplify zero_view_p initialization.
10020 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10021 view directly. Omit view when omitting .loc at line 0.
10023 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10025 PR tree-optimization/84740
10026 * tree-switch-conversion.c (process_switch): Call build_constructors
10027 only if info.phi_count is non-zero.
10029 PR tree-optimization/84739
10030 * tree-tailcall.c (find_tail_calls): Check call arguments against
10031 DECL_ARGUMENTS (current_function_decl) rather than
10032 DECL_ARGUMENTS (func) when checking for tail recursion.
10034 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10036 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10037 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10038 Volker Reichelt's entry and add entries for people that perform
10039 GCC fuzzy testing and report numerous bugs.
10041 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10044 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10045 readonly data in sdata, if that is disabled.
10046 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10047 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10048 -mreadonly-in-sdata option.
10050 2018-03-07 Martin Sebor <msebor@redhat.com>
10052 PR tree-optimization/84468
10053 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10054 basic block when looking for nul assignment.
10056 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10059 * except.h (output_function_exception_table): Adjust prototype.
10060 * except.c (output_function_exception_table): Remove FNNAME parameter
10061 and add SECTION parameter. Ouput one part of the table at a time.
10062 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10063 the first part of the exception table and emit unwind directives.
10064 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10065 (i386_pe_seh_cold_init): Likewise.
10066 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10067 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10068 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10069 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10070 * config/i386/winnt.c: Include except.h.
10071 (struct seh_frame_state): Add reg_offset, after_prologue and
10072 in_cold_section fields.
10073 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10074 (i386_pe_seh_cold_init): New function.
10075 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10076 to seh->in_cold_section.
10077 (seh_emit_push): Record the offset of the push.
10078 (seh_emit_save): Record the offet of the save.
10079 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10080 Test seh->after_prologue to disregard the epilogue.
10081 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10082 (i386_pe_end_cold_function): New function.
10084 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10087 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10088 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10091 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10093 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10094 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10096 PR middle-end/84723
10097 * multiple_target.c: Include tree-inline.h and intl.h.
10098 (expand_target_clones): Diagnose and fail if node->definition and
10099 !tree_versionable_function_p (node->decl).
10101 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10103 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10105 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10106 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10107 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10109 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10112 * combine.c (try_combine): Use reg_or_subregno instead of handling
10113 just paradoxical SUBREGs and REGs.
10115 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10117 * config/arc/arc.c (arc_finalize_pic): Remove function.
10118 (arc_must_save_register): We use single base PIC register, remove
10119 checks to save/restore the PIC register.
10120 (arc_expand_prologue): Likewise.
10121 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10123 (arc_verify_short): Likewise.
10124 (arc_attr_type): Likewise.
10125 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10126 (walk_stores): Likewise.
10127 (arc_address_cost): Make it static.
10128 (arc_verify_short): Likewise.
10129 (branch_dest): Likewise.
10130 (arc_attr_type): Likewise.
10131 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10132 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10133 (arc_final_prescan_insn): Remove inserting the nops due to
10134 hardware hazards. It is done in reorg step.
10135 (insn_length_variant_t): Remove.
10136 (insn_length_parameters_t): Likewise.
10137 (arc_insn_length_parameters): Likewise.
10138 (arc_get_insn_variants): Likewise.
10139 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10141 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10143 PR inline-asm/84683
10144 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10147 PR tree-optimization/84687
10148 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10150 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10152 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10154 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10155 Rename to ppc_speculation_barrier.
10156 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10157 __builtin_ppc_speculation_barrier.
10159 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10162 * combine.c (combine_simplify_rtx): Don't try to simplify if
10163 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10166 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10168 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10169 to 32 bytes when compiling for POWER9.
10171 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10174 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10175 regparm >= 3 with no arg reg available also for calls with
10176 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10179 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10181 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10183 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10186 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10188 2018-03-05 Richard Biener <rguenther@suse.de>
10190 PR tree-optimization/84486
10191 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10192 When inserting a __builtin_assume_aligned call set the LHS
10193 SSA name alignment info accordingly.
10195 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10197 PR tree-optimization/84114
10198 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10199 Avoid reassociation of FLOAT_MODE addition.
10201 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10203 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10204 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10205 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10206 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10207 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10208 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10209 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10211 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10212 __builtin_ia32_wbinvd): New builtins.
10213 (SPECIAL_ARGS2): New.
10214 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10215 (SPECIAL_ARGS2): New.
10216 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10217 (ix86_valid_target_attribute_inner_p): Ditto.
10218 (ix86_init_mmx_sse_builtins): Add special_args2.
10219 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10220 TARGET_WBNOINVD_P): New.
10221 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10222 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10223 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10224 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10225 * config/i386/pconfigintrin.h: New file.
10226 * config/i386/wbnoinvdintrin.h: Ditto.
10227 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10229 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10231 2018-03-05 Richard Biener <rguenther@suse.de>
10233 PR tree-optimization/84670
10234 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10236 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10237 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10238 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10239 doesn't grow if all successors (recursively) were visited at least
10242 2018-03-05 Richard Biener <rguenther@suse.de>
10244 PR tree-optimization/84650
10245 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10246 if executed in the loop pipeline.
10248 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10250 * doc/configfiles.texi (Configuration Files): Move info about
10251 conditionalizing $target-protos.h to...
10252 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10253 differs from $target-protos.h.
10255 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10256 Chung-Ju Wu <jasonwucj@gmail.com>
10258 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10259 * config/nds32/nds32-multiple.md (setmemsi): Define.
10260 * config/nds32/nds32-memory-manipulation.c
10261 (nds32_gen_dup_4_byte_to_word_value): New.
10262 (emit_setmem_word_loop): New.
10263 (emit_setmem_byte_loop): New.
10264 (nds32_expand_setmem_loop): New.
10265 (nds32_expand_setmem_loop_v3m): New.
10266 (nds32_expand_setmem_unroll): New.
10267 (nds32_expand_setmem): New.
10269 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10270 Chung-Ju Wu <jasonwucj@gmail.com>
10272 * config/nds32/nds32-memory-manipulation.c
10273 (nds32_emit_load_store): New.
10274 (nds32_emit_post_inc_load_store): New.
10275 (nds32_emit_mem_move): New.
10276 (nds32_emit_mem_move_block): New.
10277 (nds32_expand_movmemsi_loop_unknown_size): New.
10278 (nds32_expand_movmemsi_loop_known_size): New.
10279 (nds32_expand_movmemsi_loop): New.
10280 (nds32_expand_movmemsi_unroll): New.
10281 (nds32_expand_movmemqi): Rename ...
10282 (nds32_expand_movmemsi): ... to this.
10283 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10284 (movmemsi): ... to this.
10285 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10286 (nds32_expand_movmemsi): ... to this.
10288 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10289 Monk Chiang <sh.chiang04@gmail.com>
10290 Chung-Ju Wu <jasonwucj@gmail.com>
10292 * config/nds32/nds32-protos.h
10293 (nds32_expand_load_multiple): New arguments.
10294 (nds32_expand_store_multiple): Ditto.
10295 (nds32_valid_multiple_load_store): Rename ...
10296 (nds32_valid_multiple_load_store_p): ... to this.
10297 * config/nds32/nds32-memory-manipulation.c
10298 (nds32_expand_load_multiple): Refine implementation.
10299 (nds32_expand_store_multiple): Ditto.
10300 * config/nds32/nds32-multiple.md
10301 (load_multiple): Update nds32_expand_load_multiple interface.
10302 (store_multiple): Update nds32_expand_store_multiple interface.
10303 * config/nds32/nds32-predicates.c
10304 (nds32_valid_multiple_load_store): Rename ...
10305 (nds32_valid_multiple_load_store_p): ... to this and refine
10307 * config/nds32/predicates.md
10308 (nds32_load_multiple_and_update_address_operation): New predicate.
10309 (nds32_store_multiple_and_update_address_operation): New predicate.
10311 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10312 Chung-Ju Wu <jasonwucj@gmail.com>
10314 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10315 (combo): New attribute.
10316 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10318 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10320 * config/nds32/nds32.opt: Change -mcmodel= default value.
10322 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10323 Monk Chiang <sh.chiang04@gmail.com>
10324 Chung-Ju Wu <jasonwucj@gmail.com>
10326 * config/nds32/constants.md (unspec_element): New enum.
10327 * config/nds32/constraints.md (Umw): New constraint.
10328 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10329 * config/nds32/nds32-intrinsic.md: Likewise.
10330 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10331 (nds32_valid_smw_lwm_base_p): New.
10332 (nds32_output_smw_single_word): New.
10333 (nds32_output_lmw_single_word): New.
10334 (nds32_expand_unaligned_load): New.
10335 (nds32_expand_unaligned_store): New.
10336 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10337 (nds32_output_smw_single_word): Declare.
10338 (nds32_output_lmw_single_word): Declare.
10339 (nds32_expand_unaligned_load): Declare.
10340 (nds32_expand_unaligned_store): Declare.
10341 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10342 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10343 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10344 NDS32_BUILTIN_UASTORE_DW.
10345 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10348 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10349 Kito Cheng <kito.cheng@gmail.com>
10350 Chung-Ju Wu <jasonwucj@gmail.com>
10352 * config/nds32/nds32-intrinsic.c
10353 (nds32_expand_builtin_null_ftype_reg): Delete.
10354 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10355 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10356 (nds32_read_argument): New.
10357 (nds32_legitimize_target): Ditto.
10358 (nds32_legitimize_argument): Ditto.
10359 (nds32_check_constant_argument): Ditto.
10360 (nds32_expand_unop_builtin): Ditto.
10361 (nds32_expand_unopimm_builtin): Ditto.
10362 (nds32_expand_binop_builtin): Ditto.
10363 (nds32_builtin_decl_impl): Ditto.
10364 (builtin_description): Ditto.
10365 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10366 (nds32_init_builtins_impl): Ditto.
10367 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10368 (nds32_builtin_decl): New.
10369 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10370 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10372 2018-03-02 Jeff Law <law@redhat.com>
10374 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10375 (redundant_insn, fill_simple_delay_slots): Likewise.
10376 (fill_slots_from_thread): Likewise.
10377 * resource.c (mark_referenced_resources): Likewise.
10378 (mark_set_resources, find_dead_or_set_registers): Likewise.
10380 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10382 * substring-locations.h (format_warning_va): Formatting fix for
10383 ATTRIBUTE_GCC_DIAG.
10384 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10386 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10387 * substring-locations.c: Include intl.h.
10388 (format_warning_va): Turned into small wrapper around
10389 format_warning_n_va, renamed to ...
10390 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10391 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10393 (format_warning_at_substring_n): New function.
10394 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10395 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10396 format_warning_at_substring with just a shorter name instead of
10397 const function pointer.
10398 (fmtwarn_n): New function.
10399 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10400 appropriate, get rid of all the fmtstr temporaries, move conditionals
10401 with G_() wrapped string literals directly into fmtwarn arguments,
10402 cast dir.len to (int), formatting fixes.
10404 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10406 * doc/invoke.texi: Remove "Cilk Plus" references.
10408 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10409 Richard Biener <rguenther@suse.de>
10412 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10413 for error or warning attributes if CALL_FROM_THUNK_P is set.
10416 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10419 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10420 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10423 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10424 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10426 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10427 instead of -1U in last predictors element's probability member.
10429 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10432 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10433 arguments if they are comparable.
10435 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10437 PR tree-optimization/84634
10438 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10439 masks and masked_loop_p with a single loop_masks, making sure it's
10440 null for bb vectorization.
10442 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10444 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10445 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10446 loop->force_vectorize to check whether there is no alias.
10448 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10451 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10453 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10455 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10456 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10457 instead of a loop around prev_real_insn.
10458 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10461 PR inline-asm/84625
10462 * config/i386/i386.c (ix86_print_operand): Use conditional
10463 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10466 2018-03-02 Richard Biener <rguenther@suse.de>
10468 PR tree-optimization/84427
10469 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10470 (bitmap_set_subtract_values): Rewrite to handle multiple
10473 (prune_clobbered_mems): Likewise.
10474 (phi_translate): Take edge instead of pred/phiblock.
10475 (phi_translate_1): Likewise.
10476 (phi_translate_set): Likewise. Insert all translated
10477 exprs for a value into the set, keeping possibly multiple
10478 expressions per value.
10479 (compute_antic_aux): Adjust for phi_translate changes.
10480 When intersecting union the expressions and prune those
10481 not in the final value set, keeping possibly multiple
10482 expressions per value. Do not use value-insertion
10483 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10484 all expressions. Add verification that the value-sets
10485 only shrink during iteration.
10486 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10487 (do_pre_regular_insertion): Likewise.
10488 (do_pre_partial_partial_insertion): Likewise.
10490 2018-03-02 Richard Biener <rguenther@suse.de>
10493 * config/darwin.c (saved_debug_info_level): New static global.
10494 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10495 (darwin_asm_lto_end): Restore debug info generation settings.
10497 2018-03-01 Martin Liska <mliska@suse.cz>
10500 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10501 volatile arguments.
10503 2018-03-01 Richard Biener <rguenther@suse.de>
10506 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10507 type annotation with LTO when debug was disabled at compile-time.
10509 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10511 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10513 (mips_final_postscan_insn): Likewise.
10515 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10517 PR rtl-optimization/84528
10518 * alias.c (init_alias_target): Add commentary.
10519 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10520 a unique base value if the frame pointer is not eliminated
10521 to the stack pointer.
10523 2018-03-01 Tom de Vries <tom@codesourcery.com>
10525 PR rtl-optimization/83327
10526 * lra-int.h (hard_regs_spilled_into): Declare.
10527 * lra.c (hard_regs_spilled_into): Define.
10528 (init_reg_info): Init hard_regs_spilled_into.
10529 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10530 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10531 (process_bb_lives): Handle hard_regs_spilled_into.
10532 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10533 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10535 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10537 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10538 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10539 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10540 * config/rs6000/aix72.h: New file.
10542 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10544 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10545 instead of warning_at with conditional singular and plural messages
10549 * stor-layout.c (update_alignment_for_field): For
10550 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10551 && !DECL_PACKED (field), do the alignment update, just use
10552 only desired_align instead of MAX (type_align, desired_align)
10554 (place_field): Don't do known_align < desired_align handling
10555 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10556 is non-NULL, instead do it after rli->prev_field handling and
10557 only if not within a bitfield word. For DECL_PACKED (field)
10558 use type_align of BITS_PER_UNIT.
10560 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10562 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10563 superfluous parentheses and trailing spaces.
10565 2018-02-28 Richard Biener <rguenther@suse.de>
10567 PR tree-optimization/84584
10568 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10569 SCoPs with fake exit edge.
10571 2018-02-28 Martin Liska <mliska@suse.cz>
10574 * timevar.c (timer::print): Fix format to properly print 100%
10577 2018-02-28 Richard Biener <rguenther@suse.de>
10579 PR middle-end/84607
10580 * genmatch.c (capture_info::walk_match): Do not mark
10581 captured expressions without operands as expr_p given
10582 they act more like predicates and should be subject to
10583 "lost tail" side-effect preserving.
10585 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10587 PR rtl-optimization/81611
10588 * auto-inc-dec.c (attempt_change): Move dead note from
10589 mem_insn if it's the next use of regno
10590 (find_address): Take address use of reg holding
10591 non-incremented value. Add parm to limit search to the named
10593 (merge_in_block): Attempt to use a mem insn that is the next
10594 use of the original regno.
10596 2018-02-27 Martin Sebor <msebor@redhat.com>
10599 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10600 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10602 2018-02-27 Martin Sebor <msebor@redhat.com>
10604 PR translation/84207
10605 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10606 n argument to unsigned HOST_WIDE_INT.
10607 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10608 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10609 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10610 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10612 2018-02-27 Richard Biener <rguenther@suse.de>
10614 PR tree-optimization/84512
10615 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10616 Do not use the estimate returned from record_stmt_cost for
10617 the scalar iteration cost but sum properly using add_stmt_cost.
10619 2018-02-27 Richard Biener <rguenther@suse.de>
10621 PR tree-optimization/84466
10622 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10623 Adjust last change to less strictly validate use operands.
10625 2018-02-27 Martin Liska <mliska@suse.cz>
10627 PR gcov-profile/84548
10628 * gcov.c (process_file): Allow partial overlap and consider it
10629 also as group functions.
10630 (output_lines): Properly calculate range of lines for a group.
10632 2018-02-27 Martin Liska <mliska@suse.cz>
10634 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10635 'ggc' suffixes. Change first column width.
10636 (timer::print): Fix formatting of the column.
10638 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10640 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10641 preserve inline entry blocks for the sake of debug inline
10642 entry point markers alone.
10643 (remove_unused_locals): Suggest in comments a better place to
10644 force the preservation of inline entry blocks that are
10645 otherwise unused, but do not preserve them.
10647 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10649 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10651 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10654 * config/i386/constraints.md (Bs): Replace
10655 ix86_indirect_branch_register with
10656 TARGET_INDIRECT_BRANCH_REGISTER.
10658 * config/i386/i386.md (indirect_jump): Likewise.
10659 (tablejump): Likewise.
10660 (*sibcall_memory): Likewise.
10661 (*sibcall_value_memory): Likewise.
10662 Peepholes of indirect call and jump via memory: Likewise.
10663 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10664 (*sibcall_value_GOT_32): Likewise.
10665 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10666 (GOT_memory_operand): Likewise.
10667 (call_insn_operand): Likewise.
10668 (sibcall_insn_operand): Likewise.
10669 (GOT32_symbol_operand): Likewise.
10670 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10672 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10674 PR rtl-optimization/83496
10675 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10676 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10677 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10678 redundant insn, if any.
10679 (relax_delay_slots): Likewise.
10680 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10682 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10684 PR tree-optimization/83965
10685 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10686 that grouped statements are part of a reduction chain. Return
10687 true if the statement is not marked as a reduction itself but
10688 is part of a group.
10689 (vect_recog_dot_prod_pattern): Don't check whether the statement
10690 is part of a group here.
10691 (vect_recog_sad_pattern): Likewise.
10692 (vect_recog_widen_sum_pattern): Likewise.
10694 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10697 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10699 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10702 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10704 (ix86_output_indirect_function_return): New prototype.
10705 (ix86_split_simple_return_pop_internal): Likewise.
10706 * config/i386/i386.c (indirect_return_via_cx): New.
10707 (indirect_return_via_cx_bnd): Likewise.
10708 (indirect_thunk_name): Handle return va CX_REG.
10709 (output_indirect_thunk_function): Create alias for
10710 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10711 (ix86_output_indirect_jmp): Remove the bool argument.
10712 (ix86_output_indirect_function_return): New function.
10713 (ix86_split_simple_return_pop_internal): Likewise.
10714 * config/i386/i386.md (*indirect_jump): Don't pass false
10715 to ix86_output_indirect_jmp.
10716 (*tablejump_1): Likewise.
10717 (simple_return_pop_internal): Change it to define_insn_and_split.
10718 Call ix86_split_simple_return_pop_internal to split it for
10719 -mfunction-return=.
10720 (simple_return_indirect_internal): Call
10721 ix86_output_indirect_function_return instead of
10722 ix86_output_indirect_jmp.
10724 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10727 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10728 memset and value initialization afterwards.
10730 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10732 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10734 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10737 * common/config/aarch64/aarch64-common.c
10738 (aarch_option_optimization_table[]): Switch
10739 off fomit-frame-pointer
10741 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10742 Chung-Ju Wu <jasonwucj@gmail.com>
10744 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10746 (store_multiple): Ditto.
10748 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10750 * config.gcc: Add --with-cpu support for nds32 target.
10751 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10752 * config/nds32/nds32.opt: Add -mcpu= option.
10754 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10756 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10757 isel=yes): Warn for these deprecated options.
10759 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10761 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10762 ISA_2_5_MASKS_EMBEDDED.
10764 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10766 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10767 p->max as pointers rather than using iterative_hash_expr.
10769 2018-02-23 Carl Love <cel@us.ibm.com>
10771 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10772 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10774 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10775 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10776 P8V_BUILTIN_VEC_VUNSIGNED2.
10778 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10781 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10782 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10784 * lra-constraints.c (curr_insn_transform): Set up
10785 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10787 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10788 * lra-eliminations.c (spill_pseudos): Ditto.
10789 (process_insn_for_elimination): Ditto.
10790 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10791 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10792 LRA_NON_CLOBBERED_ALT.
10794 2018-02-22 Martin Sebor <msebor@redhat.com>
10796 PR tree-optimization/84480
10797 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10798 to maybe_diag_stxncpy_trunc. Call it.
10799 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10800 from gimple_fold_builtin_strcpy. Print inlining stack.
10801 (handle_builtin_stxncpy): Print inlining stack.
10802 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10804 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10807 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10808 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10809 and -fcheck-pointer-bounds are used together.
10810 (indirect_thunk_prefix): New enum.
10811 (indirect_thunk_need_prefix): New function.
10812 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10813 "_nt" instead of "_bnd" for NOTRACK prefix.
10814 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10815 (output_indirect_thunk_function): Likewise.
10817 (ix86_code_end): Update output_indirect_thunk_function calls.
10818 (ix86_output_indirect_branch_via_reg): Replace
10819 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10820 (ix86_output_indirect_branch_via_push): Likewise.
10821 (ix86_output_function_return): Likewise.
10822 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10823 incompatible with -fcf-protection=branch and
10824 -fcheck-pointer-bounds.
10826 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10829 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10830 Change gcc_assert call to output_operand_lossage.
10832 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10834 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10836 2018-02-22 DJ Delorie <dj@redhat.com>
10837 Sebastian Perta <sebastian.perta@renesas.com>
10838 Oleg Endo <olegendo@gcc.gnu.org>
10840 * config/rx/rx.c (rx_rtx_costs): New function.
10841 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10843 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10845 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10847 2018-02-22 Martin Liska <mliska@suse.cz>
10850 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10851 Add "native" as a possible value.
10853 2018-02-22 Martin Liska <mliska@suse.cz>
10856 * config/i386/i386.c (ix86_option_override_internal):
10857 Add "native" as a possible value for -march and -mtune.
10859 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10862 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10863 to all type variants.
10865 PR tree-optimization/84503
10866 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10867 width as info->bitpos + info->bitsize - start.
10868 (merged_store_group::merge_overlapping): Simplify width computation.
10869 (check_no_overlap): New function.
10870 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10871 start + width and last_order of the group, fail if check_no_overlap
10873 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10874 to group if check_no_overlap fails.
10876 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10878 * config/rs6000/altivec.md: Delete contraint arguments to
10879 define_expand, define_split, and define_peephole2, and in
10880 define_insn_and_split if always unused.
10881 * config/rs6000/darwin.md: Ditto.
10882 * config/rs6000/dfp.md: Ditto.
10883 * config/rs6000/rs6000.md: Ditto.
10884 * config/rs6000/sync.md: Ditto.
10885 * config/rs6000/vector.md: Ditto.
10886 * config/rs6000/vsx.md: Ditto.
10888 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10890 * config/rs6000/altivec.md: Write output control strings as braced
10891 blocks instead of double-quoted strings.
10892 * config/rs6000/darwin.md: Ditto.
10893 * config/rs6000/rs6000.md: Ditto.
10894 * config/rs6000/vector.md: Ditto.
10895 * config/rs6000/vsx.md: Ditto.
10897 2018-02-21 Jason Merrill <jason@redhat.com>
10899 PR c++/84314 - ICE with templates and fastcall attribute.
10900 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10902 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10904 * ipa-cp.c (determine_versionability): Fix comment typos.
10906 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10909 * ipa-cp.c (determine_versionability): Do not version functions caling
10912 2018-02-21 Martin Liska <mliska@suse.cz>
10915 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10916 Add "native" as a possible value.
10917 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10918 the macro when native cpu detection is available.
10920 2018-02-21 Martin Liska <mliska@suse.cz>
10923 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10924 Add "native" as a possible value.
10925 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10926 when native cpu detection is available.
10928 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10929 Martin Sebor <msebor@redhat.com>
10931 PR tree-optimization/84478
10932 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10934 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10935 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10936 support which is conservatively correct, for 2 only stay conservative
10937 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10938 argument to the 2 argument get_range_strlen, adjust 6 arg
10939 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10941 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10942 (gimple_fold_builtin_strlen): Pass true as last argument to
10945 2018-02-20 Martin Sebor <msebor@redhat.com>
10947 PR middle-end/84095
10948 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10949 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10950 (builtin_memref::builtin_memref): Factor out parts into
10951 set_base_and_offset and call it.
10953 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10955 PR middle-end/84406
10956 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10957 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10958 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10959 search at the associated MODE_INT.
10961 2018-02-20 Jeff Law <law@redhat.com>
10963 PR middle-end/82123
10964 PR tree-optimization/81592
10965 PR middle-end/79257
10966 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10967 for range data rather than using global data.
10968 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10969 range data rather than using global data.
10970 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10971 pass it to children as needed.
10972 (struct directive::fmtresult): Similarly.
10973 (struct directive::set_width): Similarly.
10974 (struct directive::set_precision): Similarly.
10975 (format_integer, format_directive, parse_directive): Similarly.
10976 (format_none): Accept unnamed vr_values parameter.
10977 (format_percent, format_floating, format_character): Similarly.
10978 (format_string, format_plain): Similarly.
10979 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10980 the EVRP range analyzer for range data rather than using global data.
10981 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10982 gimple-ssa-evrp-analyze.h
10983 (class sprintf_dom_walker): Add after_dom_children member function.
10984 Add evrp_range_analyzer member.
10985 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10986 range analyzer as needed.
10987 (sprintf_dom_walker::after_dom_children): New member function.
10988 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10990 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10991 (evrp_range_analyzer::pop_to_marker): Likewise.
10993 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10995 PR tree-optimization/84419
10996 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10997 with the required type if its current type is compatible but
11000 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11002 PR middle-end/82004
11003 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11004 after vectorization.
11006 2018-02-20 Martin Liska <mliska@suse.cz>
11009 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11010 possible values if we don't have a hint.
11012 2018-02-20 Martin Liska <mliska@suse.cz>
11016 * final.c (shorten_branches): Build align_tab array with one
11018 * opts.c (finish_options): Add alignment option limit check.
11019 (MAX_CODE_ALIGN): Likewise.
11020 (MAX_CODE_ALIGN_VALUE): Likewise.
11021 * doc/invoke.texi: Document maximum allowed option value for
11022 all -falign-* options.
11024 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11027 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11028 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11029 * var-tracking.c (emit_note_insn_var_location): Remove all references
11030 to NOTE_INSN_CALL_ARG_LOCATION.
11031 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11032 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11033 Use copy_rtx_if_shared.
11034 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11035 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11036 (dwarf2out_var_location): Remove handling of
11037 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11039 * final.c (final_scan_insn): Remove all references to
11040 NOTE_INSN_CALL_ARG_LOCATION.
11041 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11042 before dumping final insns.
11043 * except.c (emit_note_eh_region_end): Remove all references to
11044 NOTE_INSN_CALL_ARG_LOCATION.
11045 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11046 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11047 * config/arc/arc.c (hwloop_optimize): Likewise.
11048 * config/arm/arm.c (create_fix_barrier): Likewise.
11049 * config/s390/s390.c (s390_chunkify_start): Likewise.
11050 * config/sh/sh.c (find_barrier): Likewise.
11051 * config/i386/i386.c (rest_of_insert_endbranch,
11052 ix86_seh_fixup_eh_fallthru): Likewise.
11053 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11054 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11055 * config/frv/frv.c (frv_function_prologue): Likewise.
11056 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11058 (note_outside_basic_block_p): Remove all references to
11059 NOTE_INSN_CALL_ARG_LOCATION.
11060 * gengtype.c (adjust_field_rtx_def): Likewise.
11061 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11063 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11064 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11067 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11070 PR tree-optimization/84452
11071 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11072 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11075 2018-02-19 Martin Liska <mliska@suse.cz>
11078 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11080 2018-02-19 Martin Liska <mliska@suse.cz>
11081 Richard Sandiford <richard.sandiford@linaro.org>
11083 PR tree-optimization/82491
11084 * gimple-fold.c (get_base_constructor): Make earlier bail out
11087 2018-02-19 Carl Love <cel@us.ibm.com>
11089 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11090 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11092 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11093 P8V_BUILTIN_VEC_NEG.
11095 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11097 * config/rl78/rl78.md (movdf): New define expand.
11099 2018-02-19 Martin Liska <mliska@suse.cz>
11102 * doc/invoke.texi: Fix typo.
11103 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11105 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11107 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11108 handle rs6000_single_float and rs6000_double_float specially for
11111 2018-02-16 Jeff Law <law@redhat.com>
11113 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11114 (pop_regs): Use it.
11116 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11119 * ipa-inline.c (inline_small_functions): Fix a typo.
11121 2018-02-16 Nathan Sidwell <nathan@acm.org>
11123 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11125 2018-02-16 Carl Love <cel@us.ibm.com>
11127 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11128 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11129 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11130 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11131 expansion to P8V_BUILTIN_VEC_FLOAT2.
11133 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11135 PR rtl-optimization/70023
11136 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11137 src_regno into account.
11139 2018-02-16 Carl Love <cel@us.ibm.com>
11141 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11142 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11143 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11144 * config/rs6000/rs6000.c: Remove case statements for
11145 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11146 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11147 and P9V_BUILTIN_VEC_VINSERT4B.
11148 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11149 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11150 * config/rs6000/vsx.md:
11151 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11154 2018-02-16 Carl Love <cel@us.ibm.com>
11156 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11158 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11160 * config/rs6000/rs6000-c.c: Add the definitions for
11161 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11162 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11163 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11164 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11165 definition for insert4b and define insn *insert3b_internal.
11166 * doc/extend.texi: Add documentation for vec_extract4b.
11168 2018-02-16 Nathan Sidwell <nathan@acm.org>
11170 * doc/extend.texi (Backwards Compatibility): Mention friend
11171 injection. Note for-scope is deprecated.
11172 * doc/invoke.texi (-ffriend-injection): Deprecate.
11174 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11176 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11177 that moved to I2, also allow destinations that are a paradoxical
11178 subreg (instead of a normal reg).
11180 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11183 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11186 2018-02-16 Richard Biener <rguenther@suse.de>
11188 PR tree-optimization/84037
11189 PR tree-optimization/84016
11191 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11192 Adjust vec_construct for the fact we need additional higher latency
11193 128bit inserts for AVX256 and AVX512 vector builds.
11194 (ix86_add_stmt_cost): Scale vector construction cost for
11197 2018-02-16 Richard Biener <rguenther@suse.de>
11199 PR tree-optimization/84417
11200 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11201 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11202 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11204 2018-02-16 Martin Liska <mliska@suse.cz>
11207 * internal-fn.def (ASAN_CHECK): Set proper flags.
11208 (ASAN_MARK): Likewise.
11210 2018-02-16 Julia Koval <julia.koval@intel.com>
11212 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11213 from PTA_CANNONLAKE.
11215 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11218 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11219 Use ++iter rather than iter++ for std::list iterators.
11220 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11221 defer deleting them until all nodes in the forest are processed. Do
11222 free even leaf nodes. Change to_process into auto_vec.
11225 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11226 * vec.h (vec_default_construct): Use memset instead of placement new
11227 if BROKEN_VALUE_INITIALIZATION is defined.
11228 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11229 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11232 PR rtl-optimization/83723
11233 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11234 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11235 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11237 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11239 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11241 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11243 PR rtl-optimization/81443
11244 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11245 from inner REGs to paradoxical SUBREGs.
11247 2018-02-16 Richard Biener <rguenther@suse.de>
11249 PR tree-optimization/84399
11250 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11251 For operands we can analyze at their definition make sure we can
11252 analyze them at each use as well.
11254 2018-02-16 Richard Biener <rguenther@suse.de>
11256 PR tree-optimization/84190
11257 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11258 volatile accesses if the decl isn't volatile.
11260 2018-02-15 Jason Merrill <jason@redhat.com>
11262 PR c++/84314 - ICE with templates and fastcall attribute.
11263 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11264 TYPE_CANONICAL on an existing type.
11266 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11268 PR tree-optimization/84383
11269 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11270 dstoff nor call operand_equal_p if dstbase is NULL.
11272 PR tree-optimization/84334
11273 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11274 also a CONSTANT_CLASS_P, punt.
11276 2018-02-14 Jim Wilson <jimw@sifive.com>
11278 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11279 first SMALL_OPERAND check. New local min_second_step. Move assert
11280 to where locals are set. Add TARGET_RVC support.
11281 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11283 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11285 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11287 2018-02-14 Martin Sebor <msebor@redhat.com>
11289 PR tree-optimization/83698
11290 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11291 arrays constrain the offset range to their bounds.
11292 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11293 (builtin_access::overlap): Avoid setting the size of overlap if it's
11295 (maybe_diag_overlap): Also consider arrays when deciding what values
11296 of offsets to include in diagnostics.
11298 2018-02-14 Martin Sebor <msebor@redhat.com>
11301 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11302 that correspond to the kind of a declaration.
11304 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11307 * config/pa/pa.md: Load address of PIC label using the linkage table
11308 if the label is nonlocal.
11310 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11312 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11313 warning message if user requests -maltivec=be.
11314 * doc/invoke.texi: Document deprecation of -maltivec=be.
11316 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11319 * config/rs6000/rs6000-c.c: Update definitions for
11320 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11321 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11323 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11326 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11327 add _get_ssp intrinsics. Remove argument from
11328 __builtin_ia32_rdssp[d|q].
11329 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11330 * config/i386/i386-builtin.def: Remove argument from
11331 __builtin_ia32_rdssp[d|q].
11332 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11333 ix86_expand_special_args_builtin for _rdssp[d|q].
11334 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11335 Clear register before usage.
11336 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11337 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11339 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11341 PR tree-optimization/84357
11342 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11343 operand 1 of an ARRAY_REF too.
11345 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11348 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11349 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11351 (set_of_reg): New struct.
11352 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11353 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11354 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11356 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11357 Split into bitclr, bitset, bitinvert patterns if appropriate.
11358 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11359 use rx_fuse_in_memory_bitop.
11360 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11361 to named insn, correct maximum insn length.
11363 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11366 * machmode.def: Define a complex mode for PARTIAL_INT.
11367 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11369 * doc/rtl.texi: Document CSPImode.
11370 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11372 (msp430_hard_regno_nregs_with_padding): Likewise.
11374 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11377 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11379 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11381 PR rtl-optimization/84169
11382 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11383 we generated a parallel as new i3 and we split that to new i2 and i3
11384 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11385 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11386 those to i2, not i1. Partially rewrite this scan code.
11388 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11391 * stor-layout.c (place_field): For variable length fields, adjust
11392 offset_align afterwards not just based on the field's alignment,
11393 but also on the size.
11395 PR middle-end/84309
11396 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11397 of exps and logs in the use_exp2 case.
11399 2018-02-13 Jeff Law <law@redhat.com>
11401 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11402 entry for "vector".
11404 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11407 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11411 * common.opt (gas-loc-support, gas-locview-support): New.
11412 (ginline-points, ginternal-reset-location-views): New.
11413 * doc/invoke.texi: Document them. Use @itemx where intended.
11414 (gvariable-location-views): Adjust.
11415 * target.def (reset_location_view): New.
11416 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11417 (TARGET_RESET_LOCATION_VIEW): New.
11418 * doc/tm.texi: Rebuilt.
11419 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11420 (dwarf2out_default_as_locview_support): New.
11421 (output_asm_line_debug_info): Use option variables.
11422 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11423 (output_loc_list): Likewise.
11424 (add_high_low_attributes): Check option variables.
11425 Don't output entry view attribute in strict mode.
11426 (gen_inlined_subroutine_die): Check option variables.
11427 (dwarf2out_inline_entry): Likewise.
11428 (init_sections_and_labels): Likewise.
11429 (dwarf2out_early_finish): Likewise.
11430 (maybe_reset_location_view): New, from...
11431 (dwarf2out_var_location): ... here. Call it.
11432 * debug.h (dwarf2out_default_as_loc_support): Declare.
11433 (dwarf2out_default_as_locview_support): Declare.
11434 * hooks.c (hook_int_rtx_insn_0): New.
11435 * hooks.h (hook_int_rtx_insn_0): Declare.
11436 * toplev.c (process_options): Take -gas-loc-support and
11437 -gas-locview-support from dwarf2out. Enable
11438 -gvariable-location-views by default only with locview
11439 assembler support. Enable -ginternal-reset-location-views by
11440 default only if the target defines the corresponding hook.
11441 Enable -ginline-points by default if location views are
11442 enabled; force it disabled if statement frontiers are
11444 * tree-inline.c (expand_call_inline): Check option variables.
11445 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11447 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11449 PR tree-optimization/84321
11450 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11451 handling. Also check whether the anti-range contains any values
11452 that satisfy the mask; switch to a VR_RANGE if not.
11454 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11457 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11459 2018-02-13 Martin Jambor <mjambor@suse.cz>
11462 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11463 of call statements, also set location of a load to a temporary.
11465 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11467 * config/rl78/rl78.c (add_vector_labels): New function.
11468 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11469 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11470 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11471 which checks that no arguments are passed.
11472 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11473 * doc/extend.texi: Documentation for the new attribute.
11475 2018-02-13 Andreas Schwab <schwab@suse.de>
11477 * config/riscv/linux.h (CPP_SPEC): Define.
11479 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11482 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11483 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11484 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11485 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11486 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11487 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11488 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11489 temporarily for AES and PCLMUL builtins.
11491 PR tree-optimization/84339
11492 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11493 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11496 PR middle-end/84309
11497 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11498 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11499 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11501 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11503 * omp-simd-clone.h: New file.
11504 * omp-simd-clone.c: Include omp-simd-clone.h.
11505 (expand_simd_clones): No longer static.
11506 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11507 cgraph.h and omp-simd-clone.h.
11508 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11509 (vect_recog_widen_shift_pattern): Formatting fix.
11510 (vect_pattern_recog_1): Don't check optab for calls.
11513 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11514 operands[2] into a REG before using gen_lowpart on it.
11516 2018-02-12 Jeff Law <law@redhat.com>
11519 * config/sh/sh.c (find_barrier): Consider a sibling call
11522 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11523 successfully back substituting a reg.
11525 2018-02-12 Richard Biener <rguenther@suse.de>
11527 PR tree-optimization/84037
11528 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11529 parameter, move visited init to caller.
11530 (vect_slp_analyze_operations): Separate cost from validity
11531 check, initialize visited once for all instances.
11532 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11534 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11536 (vect_model_promotion_demotion_cost): Likewise.
11537 (vectorizable_bswap): Guard cost modeling with !slp_node
11538 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11540 (vectorizable_call): Likewise.
11541 (vectorizable_conversion): Likewise.
11542 (vectorizable_assignment): Likewise.
11543 (vectorizable_shift): Likewise.
11544 (vectorizable_operation): Likewise.
11545 (vectorizable_store): Likewise.
11546 (vectorizable_load): Likewise.
11547 (vectorizable_condition): Likewise.
11548 (vectorizable_comparison): Likewise.
11550 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11553 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11554 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11555 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11557 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11559 PR middle-end/83665
11560 * params.def (inline-min-speedup): Increase from 8 to 15.
11561 (max-inline-insns-auto): Decrease from 40 to 30.
11562 * ipa-split.c (consider_split): Add some buffer for function to
11563 be considered inlining candidate.
11564 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11567 2018-02-12 Richard Biener <rguenther@suse.de>
11569 PR tree-optimization/84037
11570 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11571 matched stmts if we cannot swap the non-matched ones.
11573 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11575 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11576 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11577 _mm_maskz_scalef_round_ss): New intrinsics.
11578 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11579 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11580 __builtin_ia32_scalefss_round): Remove.
11581 (__builtin_ia32_scalefsd_mask_round,
11582 __builtin_ia32_scalefss_mask_round): New intrinsics.
11583 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11584 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11585 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11586 "<round_constraint>")): Changed to ...
11587 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11588 "<round_scalar_constraint>")): ... this.
11589 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11590 %0, %1, %2<round_op3>}"): Changed to ...
11591 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11592 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11593 %2<round_scalar_mask_op3>}"): ... this.
11594 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11596 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11598 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11599 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11600 (_mm_maskz_sqrt_round_ss): New intrinsics.
11601 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11602 (__builtin_ia32_sqrtsd_mask_round)
11603 (__builtin_ia32_sqrtss_mask_round): New builtins.
11604 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11605 (__builtin_ia32_sqrtss_round): Remove.
11606 (__builtin_ia32_sqrtsd_mask_round)
11607 (__builtin_ia32_sqrtss_mask_round): New builtins.
11608 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11609 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11610 ((match_operand:VF_128 1 "vector_operand"
11611 "xBm,<round_constraint>")): Changed to ...
11612 ((match_operand:VF_128 1 "vector_operand"
11613 "xBm,<round_scalar_constraint>")): ... this.
11614 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11615 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11616 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11617 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11618 %<iptr>1<round_scalar_mask_op3>}): ... this.
11619 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11620 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11622 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11625 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11626 Cast vec_cmpeq result to correct type.
11627 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11628 Cast vec_cmpgt result to correct type.
11630 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11632 * final.c (final_scan_insn_1): Renamed from...
11633 (final_scan_insn): ... this. New wrapper, to recover
11634 seen from the outermost call in recursive ones.
11635 * config/sparc/sparc.c (output_return): Drop seen from call.
11636 (output_sibcall): Likewise.
11637 * config/visium/visium.c (output_branch): Likewise.
11639 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11641 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11644 2018-02-10 Alan Modra <amodra@gmail.com>
11647 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11648 Specify LR as an input.
11650 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11653 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11654 remove_member_access_dummy_vars): New functions.
11655 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11656 lower_omp_1, execute_lower_omp): Use them.
11658 PR rtl-optimization/84308
11659 * shrink-wrap.c (spread_components): Release todo vector.
11661 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11663 PR rtl-optimization/57193
11664 * ira-color.c (struct allocno_color_data): Add member
11665 conflict_allocno_hard_prefs.
11666 (update_conflict_allocno_hard_prefs): New.
11667 (bucket_allocno_compare_func): Add a preference based on
11668 conflict_allocno_hard_prefs.
11669 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11670 (color_allocnos): Remove a dead code. Initiate
11671 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11673 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11676 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11677 constraint from =wa to wa. Avoid a subreg on the output operand,
11678 instead use a pseudo and subreg it in a move.
11679 (p9_xxbrd_<mode>): Changed to ...
11680 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11681 (p9_xxbrd_v2df): New expander.
11682 (p9_xxbrw_<mode>): Changed to ...
11683 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11684 (p9_xxbrw_v4sf): New expander.
11686 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11688 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11690 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11693 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11694 multiply in 32-bit mode.
11695 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11696 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11699 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11701 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11702 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11703 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11704 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11706 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11709 * dwarf2out.c (is_trivial_indirect_ref): New function.
11710 (dwarf2out_late_global_decl): Do not generate a location
11711 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11712 and that are not defined in the current unit.
11714 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11716 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11717 instead of a libcall for UNORDERED.
11719 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11722 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11723 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11725 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11728 * config/s390/s390.c (s390_set_current_function): Invoke
11729 s390_indirect_branch_settings also if fndecl didn't change.
11731 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11733 * config/rs6000/rs6000.md (blockage): Set length to zero.
11735 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11737 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11739 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11742 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11743 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11747 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11748 PARALLEL incoming that failed vt_get_decl_and_offset check.
11750 PR middle-end/84237
11751 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11752 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11754 (get_variable_section): For decls in named .bss* sections pass true as
11755 second argument to bss_initializer_p.
11757 2018-02-09 Marek Polacek <polacek@redhat.com>
11758 Jakub Jelinek <jakub@redhat.com>
11761 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11762 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11763 Sync some changes from cxx_fold_indirect_ref.
11765 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11767 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11769 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11770 (BLOCK_INLINE_ENTRY_LABEL): New.
11771 (dwarf2out_var_location): Disregard inline entry markers.
11772 (inline_entry_data): New struct.
11773 (inline_entry_data_hasher): New hashtable type.
11774 (inline_entry_data_hasher::hash): New.
11775 (inline_entry_data_hasher::equal): New.
11776 (inline_entry_data_table): New variable.
11777 (add_high_low_attributes): Add DW_AT_entry_pc and
11778 DW_AT_GNU_entry_view attributes if a pending entry is found
11779 in inline_entry_data_table. Add old entry_pc attribute only
11780 if debug nonbinding markers are disabled.
11781 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11782 markers are enabled.
11783 (block_within_block_p, dwarf2out_inline_entry): New.
11784 (dwarf2out_finish): Check that no entries remained in
11785 inline_entry_data_table.
11786 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11787 (final_scan_insn, notice_source_line): Likewise.
11788 (rest_of_clean_state): Skip inline entry markers.
11789 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11791 * gimple.c (gimple_build_debug_inline_entry): New.
11792 * gimple.h (enum gimple_debug_subcode): Add
11793 GIMPLE_DEBUG_INLINE_ENTRY.
11794 (gimple_build_debug_inline_entry): Declare.
11795 (gimple_debug_inline_entry_p): New.
11796 (gimple_debug_nonbind_marker_p): Adjust.
11797 * insn-notes.def (INLINE_ENTRY): New.
11798 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11799 inline entry marker notes.
11800 (print_insn): Likewise.
11801 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11802 (INSN_DEBUG_MARKER_KIND): Likewise.
11803 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11804 * tree-inline.c (expand_call_inline): Build and insert
11805 debug_inline_entry stmt.
11806 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11807 inline entry blocks early, if nonbind markers are enabled.
11808 (dump_scope_block): Dump fragment info.
11809 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11810 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11811 (gimple_build_debug_inline_entry): New.
11812 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11813 Enable/disable inline entry points too.
11814 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11815 (DEBUG_INSN): Describe inline entry markers.
11817 * common.opt (gvariable-location-views): New.
11818 (gvariable-location-views=incompat5): New.
11819 * config.in: Rebuilt.
11820 * configure: Rebuilt.
11821 * configure.ac: Test assembler for view support.
11822 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11823 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11824 * dwarf2out.c (var_loc_view): New typedef.
11825 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11826 (dwarf2out_locviews_in_attribute): New.
11827 (dwarf2out_locviews_in_loclist): New.
11828 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11829 (enum dw_line_info_opcode): Add LI_adv_address.
11830 (struct dw_line_info_table): Add view.
11831 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11832 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11833 (zero_view_p): New variable.
11834 (ZERO_VIEW_P): New macro.
11835 (output_asm_line_debug_info): New.
11836 (struct var_loc_node): Add view.
11837 (add_AT_view_list, AT_loc_list): New.
11838 (add_var_loc_to_decl): Add view param. Test it against last.
11839 (new_loc_list): Add view params. Record them.
11840 (AT_loc_list_ptr): Handle loc and view lists.
11841 (view_list_to_loc_list_val_node): New.
11842 (print_dw_val): Handle dw_val_class_view_list.
11843 (size_of_die): Likewise.
11844 (value_format): Likewise.
11845 (loc_list_has_views): New.
11846 (gen_llsym): Set vl_symbol too.
11847 (maybe_gen_llsym, skip_loc_list_entry): New.
11848 (dwarf2out_maybe_output_loclist_view_pair): New.
11849 (output_loc_list): Output view list or entries too.
11850 (output_view_list_offset): New.
11851 (output_die): Handle dw_val_class_view_list.
11852 (output_dwarf_version): New.
11853 (output_compilation_unit_header): Use it.
11854 (output_skeleton_debug_sections): Likewise.
11855 (output_rnglists, output_line_info): Likewise.
11856 (output_pubnames, output_aranges): Update version comments.
11857 (output_one_line_info_table): Output view numbers in asm comments.
11858 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11859 Call maybe_gen_llsym.
11860 (loc_list_from_tree_1): Adjust.
11861 (add_AT_location_description): Create view list attribute if
11862 needed, check it's absent otherwise.
11863 (convert_cfa_to_fb_loc_list): Adjust.
11864 (maybe_emit_file): Call output_asm_line_debug_info for test.
11865 (dwarf2out_var_location): Reset views as needed. Precompute
11866 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11867 attribute. Set view.
11868 (new_line_info_table): Reset next view.
11869 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11870 (dwarf2out_source_line): Likewise. Output view resets and labels to
11871 the assembler, or select appropriate line info opcodes.
11872 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11873 (optimize_string_length): Catch it. Adjust.
11874 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11875 dw_val_class_view_list, and remove it if no longer needed.
11876 (hash_loc_list): Hash view numbers.
11877 (loc_list_hasher::equal): Compare them.
11878 (optimize_location_lists): Check whether a view list symbol is
11879 needed, and whether the locview attribute is present, and
11880 whether they match. Remove the locview attribute if no longer
11882 (index_location_lists): Call skip_loc_list_entry for test.
11883 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11884 Use output_dwarf_version.
11885 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11886 (struct dw_val_node): Add val_view_list.
11887 * final.c (SEEN_NEXT_VIEW): New.
11888 (set_next_view_needed): New.
11889 (clear_next_view_needed): New.
11890 (maybe_output_next_view): New.
11891 (final_start_function): Rename to...
11892 (final_start_function_1): ... this. Take pointer to FIRST,
11893 add SEEN parameter. Emit param bindings in the initial view.
11894 (final_start_function): Reintroduce SEEN-less interface.
11895 (final): Rename to...
11896 (final_1): ... this. Take SEEN parameter. Output final pending
11897 next view at the end.
11898 (final): Reintroduce seen-less interface.
11899 (final_scan_insn): Output pending next view before switching
11900 sections or ending a block. Mark the next view as needed when
11901 outputting variable locations. Notify debug backend of section
11902 changes, and of location view changes.
11903 (rest_of_handle_final): Adjust.
11904 * toplev.c (process_options): Autodetect value for debug variable
11905 location views option. Warn on incompat5 without -gdwarf-5.
11906 * doc/invoke.texi (gvariable-location-views): New.
11907 (gvariable-location-views=incompat5): New.
11908 (gno-variable-location-views): New.
11910 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11912 PR tree-optimization/84136
11913 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11914 that the result of find_edge is non-NULL.
11916 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11919 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11920 storing integer register in SImode. Fix cost of 256 and 512
11921 byte aligned SSE register store.
11923 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11925 * config/i386/i386.c (ix86_multiplication_cost): Fix
11926 multiplication cost for TARGET_AVX512DQ.
11928 2018-02-08 Marek Polacek <polacek@redhat.com>
11930 PR tree-optimization/84238
11931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11934 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11936 PR tree-optimization/84265
11937 * tree-vect-stmts.c (vectorizable_store): Don't treat
11938 VMAT_CONTIGUOUS accesses as grouped.
11939 (vectorizable_load): Likewise.
11941 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11943 PR tree-optimization/81635
11944 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11945 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11946 (test_round_for_mask): New functions.
11947 (wide_int_cc_tests): Call test_round_for_mask.
11948 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11949 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11950 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11951 range returned by get_range_info.
11953 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11956 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11957 * symtab.c: Include builtins.h
11958 (symtab_node::output_to_lto_symbol_table_p): Move here
11959 from lto-streamer-out.c:output_symbol_p.
11960 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11961 (output_symbol_p): Move all logic to symtab.c
11962 (produce_symtab): Update.
11964 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11966 * config/s390/s390-opts.h (enum indirect_branch): Define.
11967 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11968 (s390_indirect_branch_via_thunk)
11969 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11970 (enum s390_indirect_branch_type): Define.
11971 * config/s390/s390.c (struct s390_frame_layout, struct
11972 machine_function): Remove.
11973 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11974 (indirect_branch_table_label_no, indirect_branch_table_name):
11976 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11977 (enum s390_indirect_branch_option): Define.
11978 (s390_return_addr_from_memory): New function.
11979 (s390_handle_string_attribute): New function.
11980 (s390_attribute_table): Add new attribute handler.
11981 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11982 (s390_indirect_branch_via_thunk): New function.
11983 (s390_indirect_branch_via_inline_thunk): New function.
11984 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11985 sibling call optimization for non z10 compiles.
11986 (s390_emit_call): Force indirect branch target to be a single
11987 register. Add r1 clobber for non-z10 compiles.
11988 (s390_emit_epilogue): Emit return jump via return_use expander.
11989 (s390_reorg): Handle JUMP_INSNs as execute targets.
11990 (s390_option_override_internal): Perform validity checks for the
11991 new command line options.
11992 (s390_indirect_branch_attrvalue): New function.
11993 (s390_indirect_branch_settings): New function.
11994 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11995 (s390_output_indirect_thunk_function): New function.
11996 (s390_code_end): Implement target hook.
11997 (s390_case_values_threshold): Implement target hook.
11998 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12000 * config/s390/s390.h (struct s390_frame_layout)
12001 (struct machine_function): Move here from s390.c.
12002 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12003 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12004 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12005 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12006 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12007 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12008 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12009 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12010 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12011 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12012 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12013 (mnemonic attribute): Add values which aren't recognized
12015 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12016 pattern for branch conversion. Fix mnemonic attribute.
12017 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12018 indirect branch via thunk if requested.
12019 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12020 ("*indirect_jump"): Disable for branch conversion using out of
12022 ("indirect_jump_via_thunk<mode>_z10")
12023 ("indirect_jump_via_thunk<mode>")
12024 ("indirect_jump_via_inlinethunk<mode>_z10")
12025 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12026 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12027 ("casesi_jump_via_inlinethunk<mode>_z10")
12028 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12029 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12030 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12031 ("*indirect2_jump"): Disable for branch conversion.
12032 ("casesi_jump"): Turn into expander and expand patterns for branch
12034 ("return_use"): New expander.
12035 ("*return"): Emit return via thunk and rename it to ...
12036 ("*return<mode>"): ... this one.
12037 * config/s390/s390.opt: Add new options and and enum for the
12040 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12042 * lra-constraints.c (match_reload): Unconditionally use
12043 gen_lowpart_SUBREG, rather than selecting between that
12044 and equivalent gen_rtx_SUBREG code.
12046 2018-02-08 Richard Biener <rguenther@suse.de>
12048 PR tree-optimization/84233
12049 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12050 changed flag instead of boguously re-using phi_inserted.
12052 2018-02-08 Martin Jambor <mjambor@suse.cz>
12054 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12055 static local variables.
12057 2018-02-08 Richard Biener <rguenther@suse.de>
12059 PR tree-optimization/84278
12060 * tree-vect-stmts.c (vectorizable_store): When looking for
12061 smaller vector types to perform grouped strided loads/stores
12062 make sure the mode is supported by the target.
12063 (vectorizable_load): Likewise.
12065 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12067 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12068 Increase LDP/STP opportunities by adding adjacent callee-saves.
12070 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12072 PR rtl-optimization/84068
12073 PR rtl-optimization/83459
12074 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12076 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12078 PR tree-optimization/84224
12079 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12080 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12081 non-zero arguments.
12083 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12086 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12087 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12088 count, remove one USE.
12090 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12092 * doc/install.texi (Configuration): Document the
12093 --with-long-double-format={ibm,ieee} PowerPC configuration
12097 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12098 Convert from define_expand to be define_insn_and_split. Rework
12099 float/double/_Float128 conversions to QI/HI/SImode to work with
12100 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12101 conversions to QI/HImode types did a store and then a load to
12102 truncate the value. For conversions to VSX registers, don't split
12103 the insn, instead emit the code directly. Use the code iterator
12104 any_fix to combine signed and unsigned conversions.
12105 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12106 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12107 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12108 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12109 (fix_<mode>di2_hw): Likewise.
12110 (fixuns_<mode>di2_hw): Likewise.
12111 (fix_<mode>si2_hw): Likewise.
12112 (fixuns_<mode>si2_hw): Likewise.
12113 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12114 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12115 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12116 fix<uns>_trunc<SFDF:mode>si2_p8.
12117 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12118 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12119 (fix<uns>_<mode>_mem): Likewise.
12120 (fctiw<u>z_<mode>_mem): Likewise.
12121 (fix<uns>_<mode>_mem): Likewise.
12122 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12123 the register allocator from doing a direct move to the GPRs to do
12124 a store, and instead use the ISA 3.0 store byte/half-word from
12125 vector register instruction. For IEEE 128-bit floating point,
12126 also optimize stores of 32-bit ints.
12127 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12129 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12131 * genextract.c (push_pathstr_operand): New function to support
12133 (walk_rtx): Call push_pathstr_operand.
12134 (print_path): Support [a-zA-Z].
12136 2018-02-07 Richard Biener <rguenther@suse.de>
12138 PR tree-optimization/84037
12139 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12140 (cse_and_gimplify_to_preheader): Declare.
12141 (vect_get_place_in_interleaving_chain): Likewise.
12142 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12144 (_loop_vec_info::~_loop_vec_info): Delete it.
12145 (cse_and_gimplify_to_preheader): New function.
12146 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12147 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12148 (vectorizable_load): Likewise. For grouped stores always base
12149 the IV on the first element.
12150 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12151 condition before gimplifying.
12153 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12155 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12156 *DIV_EXPR and *MOD_EXPR.
12158 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12161 * config/i386/i386.c (ix86_option_override_internal): Mask out
12162 the CF_SET bit when checking -fcf-protection.
12164 2018-02-07 Tom de Vries <tom@codesourcery.com>
12167 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12170 2018-02-07 Richard Biener <rguenther@suse.de>
12172 PR tree-optimization/84204
12173 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12176 PR tree-optimization/84205
12177 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12178 special-case isl_ast_op_zdiv_r.
12180 PR tree-optimization/84223
12181 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12182 Only add conditions from within the region.
12183 (gather_bbs::after_dom_children): Adjust.
12185 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12188 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12189 * config/avr/avr.md: Only post-reload split REG-REG moves if
12190 either register is GENERAL_REG_P.
12192 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12194 PR tree-optimization/84235
12195 * tree-ssa-scopedtables.c
12196 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12197 if the subtraction is performed in floating point type where NaNs are
12198 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12199 build 1. Formatting fix.
12201 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12204 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12205 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12206 and skip it regardless of bb boundaries. Use CALL_P macro,
12207 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12208 unnecessarily, formatting fix.
12210 2018-02-06 Michael Collison <michael.collison@arm.com>
12212 * config/arm/thumb2.md:
12213 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12214 (*thumb_mov_notscc): Ditto.
12216 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12219 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12220 unsigned_fix, not "s".
12222 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12224 * configure.ac (gcc_fn_eh_frame_ro): New function.
12225 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12226 correct .eh_frame permissions.
12227 * configure: Regenerate.
12229 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12231 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12232 irrelevant options.
12234 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12236 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12237 Display warning message for -mno-speculate-indirect-jumps.
12239 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12241 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12243 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12245 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12247 PR tree-optimization/84225
12248 * tree-eh.c (find_trapping_overflow): Only call
12249 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12251 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12254 * config/i386/i386.c: Reimplement the check of possible options
12255 -mibt/-mshstk conbination. Change error messages.
12256 * doc/invoke.texi: Fix a typo: remove extra '='.
12258 2018-02-06 Marek Polacek <polacek@redhat.com>
12260 PR tree-optimization/84228
12261 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12263 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12266 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12267 emitted arch directives.
12268 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12269 __ARM_FEATURE_COPROC before changing architectures.
12271 2018-02-06 Richard Biener <rguenther@suse.de>
12273 * config/i386/i386.c (print_reg): Fix typo.
12274 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12276 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12278 * configure: Regenerate.
12280 2018-02-05 Martin Sebor <msebor@redhat.com>
12282 PR tree-optimization/83369
12283 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12286 2018-02-05 Martin Liska <mliska@suse.cz>
12288 * doc/invoke.texi: Cherry-pick upstream r323995.
12290 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12292 * ira.c (ira_init_register_move_cost): Adjust comment.
12294 2018-02-05 Martin Liska <mliska@suse.cz>
12296 PR gcov-profile/84137
12297 * doc/gcov.texi: Fix typo in documentation.
12299 2018-02-05 Martin Liska <mliska@suse.cz>
12301 PR gcov-profile/83879
12302 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12303 using dlopen functionality.
12305 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12307 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12308 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12309 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12310 _mm_maskz_range_round_ss): New intrinsics.
12311 (__builtin_ia32_rangesd128_round)
12312 (__builtin_ia32_rangess128_round): Remove.
12313 (__builtin_ia32_rangesd128_mask_round,
12314 __builtin_ia32_rangess128_mask_round): New builtins.
12315 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12316 __builtin_ia32_rangess128_round): Remove.
12317 (__builtin_ia32_rangesd128_mask_round,
12318 __builtin_ia32_rangess128_mask_round): New builtins.
12319 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12320 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12321 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12322 "<round_saeonly_constraint>")): Changed to ...
12323 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12324 "<round_saeonly_scalar_constraint>")): ... this.
12325 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12326 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12327 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12328 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12329 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12331 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12333 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12335 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12336 Remove all values except native, 8540 and 8548.
12338 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12340 * config/i386/i386.c (ix86_output_function_return): Pass
12341 INVALID_REGNUM, instead of -1, as invalid register number to
12342 indirect_thunk_name and output_indirect_thunk.
12344 2018-02-02 Julia Koval <julia.koval@intel.com>
12346 * config.gcc: Add -march=icelake.
12347 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12348 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12349 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12350 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12351 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12352 (processor_target_table): Add icelake.
12353 (ix86_option_override_internal): Handle new PTAs.
12354 (get_builtin_code_for_version): Handle icelake.
12355 (M_INTEL_COREI7_ICELAKE): New.
12356 (fold_builtin_cpu): Handle icelake.
12357 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12358 * doc/invoke.texi: Add -march=icelake.
12360 2018-02-02 Julia Koval <julia.koval@intel.com>
12362 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12363 to wide_int_bitmask.
12364 * wide-int-bitmask.h: New.
12366 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12369 * config/i386/i386.md: Replace Pmode with word_mode in
12370 builtin_setjmp_setup and builtin_longjmp to support x32.
12372 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12376 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12378 (rs6000_supported_cpu_names): New static variable.
12379 (linux_cpu_translation_table): Likewise.
12380 (elf_platform) <cpu>: Define new static variable and use it.
12381 Translate kernel AT_PLATFORM name to canonical name if needed.
12382 Error if platform name is unknown.
12384 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12387 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12389 2018-02-01 Jeff Law <law@redhat.com>
12392 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12393 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12394 the scratch if RELEASE_VIA_POP is false.
12395 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12396 If we have to save a temporary register, decrement SIZE appropriately.
12397 Pass new arguments to release_scratch_register_on_entry.
12398 (ix86_adjust_stack_and_probe): Likewise.
12399 (ix86_emit_probe_stack_range): Pass new arguments to
12400 release_scratch_register_on_entry.
12402 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12404 PR rtl-optimization/84157
12405 * combine.c (change_zero_ext): Use REG_P predicate in
12406 front of HARD_REGISTER_P predicate.
12408 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12410 * config/avr/avr.c (avr_option_override): Move disabling of
12411 -fdelete-null-pointer-checks to...
12412 * common/config/avr/avr-common.c (avr_option_optimization_table):
12415 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12417 PR tree-optimization/81635
12418 * tree-data-ref.c (split_constant_offset_1): For types that
12419 wrap on overflow, try to use range info to prove that wrapping
12422 2018-02-01 Renlin Li <renlin.li@arm.com>
12425 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12426 TAILCALL_ADDR_REGS.
12427 (aarch64_register_move_cost): Likewise.
12428 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12429 TAILCALL_ADDR_REGS.
12430 (REG_CLASS_NAMES): Likewise.
12431 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12432 TAILCALL_ADDR_REGS. Remove IP registers.
12433 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12435 2018-02-01 Richard Biener <rguenther@suse.de>
12437 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12438 for specifying RPO order and allow NULL for that.
12439 * domwalk.c (dom_walker::dom_walker): Likewise.
12440 (dom_walker::walk): Handle NULL RPO order.
12441 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12443 (rewrite_update_dom_walker): Likewise.
12444 (mark_def_dom_walker): Likewise.
12446 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12448 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12449 (aarch64_maybe_expand_sve_subreg_move): Declare.
12450 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12451 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12453 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12454 that are semantically a reverse operation.
12455 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12456 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12457 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12459 (aarch64_can_change_mode_class): For big-endian, forbid changes
12460 between two SVE modes if they have different element sizes.
12462 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12464 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12465 the TImode handling for big-endian targets.
12467 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12469 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12470 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12472 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12473 Remove BSWAP handing for big-endian targets and use the form of
12474 LD1RQ appropariate for the mode.
12476 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12478 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12479 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12480 duplicated element.
12482 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12485 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12486 check for operands that need to go through aarch64_sve_reload_be.
12488 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12490 PR tree-optimization/81661
12491 PR tree-optimization/84117
12492 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12493 * tree-eh.c: Include gimplify.h.
12494 (find_trapping_overflow, replace_trapping_overflow,
12495 rewrite_to_non_trapping_overflow): New functions.
12496 * tree-vect-loop.c: Include tree-eh.h.
12497 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12498 * tree-data-ref.c: Include tree-eh.h.
12499 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12501 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12503 PR rtl-optimization/84123
12504 * combine.c (change_zero_ext): Check if hard register satisfies
12505 can_change_dest_mode before calling gen_lowpart_SUBREG.
12507 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12510 * ira.c (ira_init_register_move_cost): Remove assert.
12512 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12514 PR rtl-optimization/84071
12515 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12516 * doc/tm.texi: Regenerate.
12518 2018-01-31 Richard Biener <rguenther@suse.de>
12520 PR tree-optimization/84132
12521 * tree-data-ref.c (analyze_miv_subscript): Properly
12522 check whether evolution_function_is_affine_multivariate_p
12523 before calling gcd_of_steps_may_divide_p.
12525 2018-01-31 Julia Koval <julia.koval@intel.com>
12528 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12529 * config/i386/i386.md (rdpid_rex64) New.
12530 (rdpid): Make 32bit only.
12532 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12535 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12536 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12538 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12541 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12543 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12545 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12547 PR rtl-optimization/84071
12548 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12549 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12551 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12553 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12554 (arc_attribute_table): Add 'aux' attribute.
12555 (arc_in_small_data_p): Consider aux like variables.
12556 (arc_is_aux_reg_p): New function.
12557 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12558 (arc_get_aux_arg): New function.
12559 (prepare_move_operands): Handle aux-register access.
12560 (arc_handle_aux_attribute): New function.
12561 * doc/extend.texi (ARC Variable attributes): Add subsection.
12563 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12565 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12566 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12567 (arc_attribute_table): Add 'uncached' attribute.
12568 (arc_print_operand): Print '.di' flag for uncached memory
12570 (arc_in_small_data_p): Do not consider for small data the uncached
12572 (arc_is_uncached_mem_p): New function.
12573 * config/arc/predicates.md (compact_store_memory_operand): Check
12574 for uncached memory accesses.
12575 (nonvol_nonimm_operand): Likewise.
12576 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12578 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12581 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12582 falign-loops=): Add Optimization flag.
12584 2018-01-30 Jeff Law <law@redhat.com>
12587 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12588 INT_REGISTERS_SAVED. Check it prior to calling
12589 get_scratch_register_on_entry.
12590 (ix86_adjust_stack_and_probe): Similarly.
12591 (ix86_emit_probe_stack_range): Similarly.
12592 (ix86_expand_prologue): Corresponding changes.
12594 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12597 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12598 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12600 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12603 * lra-constraints.c (curr_insn_transform): Process AND in the
12606 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12608 PR rtl-optimization/83986
12609 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12610 dependence against last_pending_memory_flush in addition to
12611 pending_jump_insns.
12613 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12615 PR tree-optimization/81611
12616 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12619 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12622 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12625 2018-01-30 Richard Biener <rguenther@suse.de>
12626 Jakub Jelinek <jakub@redhat.com>
12628 PR tree-optimization/84111
12629 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12630 inner loops added during recursion, as they don't have up-to-date
12633 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12636 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12637 (can_inline_edge_by_limits_p): ... here.
12638 (can_early_inline_edge_p, check_callers,
12639 update_caller_keys, update_callee_keys, recursive_inlining,
12640 add_new_edges_to_heap, speculation_useful_p,
12641 inline_small_functions,
12642 inline_small_functions, flatten_function,
12643 inline_to_all_callers_1): Update.
12645 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12647 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12650 2018-01-30 Richard Biener <rguenther@suse.de>
12652 PR tree-optimization/83008
12653 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12654 invariant and constant vector uses in stmts when they need
12655 more than one stmt.
12657 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12660 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12661 * configure: Regenerate.
12663 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12665 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12667 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12668 Use gen_rtx_REG rather than gen_lowpart.
12670 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12672 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12673 rather than 0 when creating partial subregs.
12675 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12677 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12680 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12683 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12684 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12685 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12686 flags. This restores the settings used before the 2017-07-24.
12687 Turning off pre increment/decrement/modify allows IVOPTS to
12688 optimize DF/SF loops where the index is an int.
12690 2018-01-29 Richard Biener <rguenther@suse.de>
12691 Kelvin Nilsen <kelvin@gcc.gnu.org>
12694 * tree-vect-stmts.c (vectorizable_call): Don't call
12695 targetm.vectorize_builtin_md_vectorized_function if callee is
12698 2018-01-22 Carl Love <cel@us.ibm.com>
12700 * doc/extend.tex: Fix typo in second arg in
12701 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12703 2018-01-29 Richard Biener <rguenther@suse.de>
12705 PR tree-optimization/84086
12706 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12707 (flush_ssaname_freelist): When SSA names were released reset
12708 the SCEV hash table.
12710 2018-01-29 Richard Biener <rguenther@suse.de>
12712 PR tree-optimization/84057
12713 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12714 removed paths when removing edges.
12716 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12718 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12719 -mfunction-return=@var{choice}.
12721 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12723 PR diagnostic/84034
12724 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12725 Handle CR like TAB.
12726 (layout::print_source_line): Likewise.
12727 (test_get_line_width_without_trailing_whitespace): Add test cases.
12729 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12731 PR middle-end/84040
12732 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12735 2018-01-26 Jim Wilson <jimw@sifive.com>
12737 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12739 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12742 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12744 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12745 and CMP + SUB-immediate -> SUBS.
12747 2018-01-26 Martin Sebor <msebor@redhat.com>
12749 PR tree-optimization/83896
12750 * tree-ssa-strlen.c (get_string_len): Rename...
12751 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12752 Avoid assuming length is constant.
12753 (handle_char_store): Use HOST_WIDE_INT for string length.
12755 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12758 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12759 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12761 2018-01-26 Richard Biener <rguenther@suse.de>
12763 PR rtl-optimization/84003
12764 * dse.c (record_store): Only record redundant stores when
12765 the earlier store aliases at least all accesses the later one does.
12767 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12769 PR rtl-optimization/83985
12770 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12771 REG_CFA_RESTORE insns.
12772 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12773 REG_CFA_RESTORE insns here.
12776 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12777 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12779 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12781 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12782 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12783 (arc_init): Likewise.
12784 (arc_override_options): Likewise.
12785 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12787 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12789 * config/arc/arc.h (TARGET_DBNZ): Define.
12790 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12791 properly set the tune attribute.
12792 (dbnz): Use TARGET_DBNZ guard.
12793 * config/arc/arc.opt (mtune): Add core3 option.
12795 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12797 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12798 recognize new pic like addresses.
12799 (arc_delegitimize_address): Clean up.
12801 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12803 * config/arc/arc-arches.def: Option mrf16 valid for all
12805 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12806 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12807 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12808 * config/arc/arc-tables.opt: Regenerate.
12809 * config/arc/arc.c (arc_conditional_register_usage): Handle
12810 reduced register file case.
12811 (arc_file_start): Set must have build attributes.
12812 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12813 mrf16 option value.
12814 * config/arc/arc.opt (mrf16): Add new option.
12815 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12816 * config/arc/genmultilib.awk: Handle new mrf16 option.
12817 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12818 * config/arc/t-multilib: Regenerate.
12819 * doc/invoke.texi (ARC Options): Document mrf16 option.
12821 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12823 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12824 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12825 (arc_attribute_table): Add 'secure_call' attribute.
12826 (arc_print_operand): Print secure call operand.
12827 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12829 (arc_is_secure_call_p): New function. * config/arc/arc.md
12830 (call_i): Add support for sjli instruction.
12831 (call_value_i): Likewise.
12832 * config/arc/constraints.md (Csc): New constraint.
12834 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12835 John Eric Martin <John.Martin@emmicro-us.com>
12837 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12838 * config/arc/arc.c (_arc_jli_section): New struct.
12839 (arc_jli_section): New type.
12840 (rc_jli_sections): New static variable.
12841 (arc_handle_jli_attribute): New function.
12842 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12843 (arc_file_end): New function.
12844 (TARGET_ASM_FILE_END): Define.
12845 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12846 (arc_add_jli_section): New function.
12847 (jli_call_scan): Likewise.
12848 (arc_reorg): Call jli_call_scan.
12849 (arc_output_addsi): Remove 'S' from printing asm operand.
12850 (arc_is_jli_call_p): New function.
12851 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12853 (movhi_insn): Likewise.
12854 (movsi_insn): Likewise.
12855 (movsi_set_cc_insn): Likewise.
12856 (loadqi_update): Likewise.
12857 (load_zeroextendqisi_update): Likewise.
12858 (load_signextendqisi_update): Likewise.
12859 (loadhi_update): Likewise.
12860 (load_zeroextendhisi_update): Likewise.
12861 (load_signextendhisi_update): Likewise.
12862 (loadsi_update): Likewise.
12863 (loadsf_update): Likewise.
12864 (movsicc_insn): Likewise.
12865 (bset_insn): Likewise.
12866 (bxor_insn): Likewise.
12867 (bclr_insn): Likewise.
12868 (bmsk_insn): Likewise.
12869 (bicsi3_insn): Likewise.
12870 (cmpsi_cc_c_insn): Likewise.
12871 (movsi_ne): Likewise.
12872 (movsi_cond_exec): Likewise.
12873 (clrsbsi2): Likewise.
12874 (norm_f): Likewise.
12884 (call_i): Remove 'S' asm letter, add jli instruction.
12885 (call_value_i): Likewise.
12886 * config/arc/arc.op (mjli-always): New option.
12887 * config/arc/constraints.md (Cji): New constraint.
12888 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12890 (subsf3_fpx): Likewise.
12891 (mulsf3_fpx): Likewise.
12892 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12894 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12895 function attrbutes.
12896 * doc/invoke.texi (ARC): Document mjli-always option.
12898 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12900 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12901 avoid addition with 0 and use incw and decw where possible.
12903 2018-01-26 Richard Biener <rguenther@suse.de>
12905 PR tree-optimization/81082
12906 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12907 association if it requires casting to unsigned.
12908 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12909 from fold_plusminus_mult_expr to catch important cases late when
12910 range info is available.
12912 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12914 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12915 * configure.ac (hidden_linkonce): New test.
12916 * configure: Regenerate.
12917 * config.in: Regenerate.
12919 2018-01-26 Julia Koval <julia.koval@intel.com>
12921 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12922 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12923 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12924 _mm_mask_bitshuffle_epi64_mask): Fix type.
12925 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12926 USI_FTYPE_V4DI_V4DI_USI): Remove.
12927 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12928 __builtin_ia32_vpshufbitqmb256_mask,
12929 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12930 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12931 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12933 2018-01-26 Alan Modra <amodra@gmail.com>
12936 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12937 UNSPEC_VBPERMQ. Sort other unspecs.
12939 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12941 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12943 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12945 PR middle-end/83055
12946 * predict.c (drop_profile): Do not push/pop cfun; update also
12948 (handle_missing_profiles): Fix logic looking for zero profiles.
12950 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12952 PR middle-end/83977
12953 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12954 on functions with #pragma omp declare simd or functions with simd
12956 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12957 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12958 Remove trailing \n from warning_at calls.
12960 2018-01-25 Tom de Vries <tom@codesourcery.com>
12963 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12964 for neutered workers.
12966 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12969 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12970 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12972 2018-01-24 Jeff Law <law@redhat.com>
12975 * i386.c (get_probe_interval): Move to earlier point.
12976 (ix86_compute_frame_layout): If -fstack-clash-protection and
12977 the frame is larger than the probe interval, then use pushes
12978 to save registers rather than reg->mem moves.
12979 (ix86_expand_prologue): Remove conditional for int_registers_saved
12982 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12985 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12986 min/max for never referenced object.
12988 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12990 PR middle-end/83977
12991 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12993 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12994 attributes from DECL_ATTRIBUTES (decl) without affecting
12995 DECL_ATTRIBUTES (current_function_decl).
12996 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12997 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12999 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13001 PR tree-optimization/83979
13002 * fold-const.c (fold_comparison): Use constant_boolean_node
13003 instead of boolean_{true,false}_node.
13005 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13007 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13010 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13012 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13013 Simplify the clause that sets the length attribute.
13014 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13015 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13016 clause that sets the length attribute.
13017 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13019 2018-01-24 Tom de Vries <tom@codesourcery.com>
13022 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13023 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13024 Add strict parameter.
13025 (prevent_branch_around_nothing): Insert dummy insn between branch to
13026 label and label with no ptx insn inbetween.
13027 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13029 2018-01-24 Tom de Vries <tom@codesourcery.com>
13032 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13033 for neutered threads in warp.
13034 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13036 2018-01-24 Richard Biener <rguenther@suse.de>
13038 PR tree-optimization/83176
13039 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13042 2018-01-24 Richard Biener <rguenther@suse.de>
13044 PR tree-optimization/82819
13045 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13046 code generating pluses that are no-ops in the target precision.
13048 2018-01-24 Richard Biener <rguenther@suse.de>
13050 PR middle-end/84000
13051 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13053 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13055 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13056 to merge probabilities.
13057 * predict.c (probably_never_executed): Also mark as cold functions
13058 with global 0 profile and guessed local profile.
13059 * profile-count.c (profile_probability::combine_with_count): New
13061 * profile-count.h (profile_probability::operator*,
13062 profile_probability::operator*=, profile_probability::operator/,
13063 profile_probability::operator/=): Reduce precision to adjusted
13064 and set value to guessed on contradictory divisions.
13065 (profile_probability::combine_with_freq): Remove.
13066 (profile_probability::combine_wiht_count): Declare.
13067 (profile_count::force_nonzero):: Set to adjusted.
13068 (profile_count::probability_in):: Set quality to adjusted.
13069 * tree-ssa-tail-merge.c (replace_block_by): Use
13070 combine_with_count.
13072 2018-01-23 Andrew Waterman <andrew@sifive.com>
13073 Jim Wilson <jimw@sifive.com>
13075 * config/riscv/riscv.c (riscv_stack_boundary): New.
13076 (riscv_option_override): Set riscv_stack_boundary. Handle
13077 riscv_preferred_stack_boundary_arg.
13078 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13079 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13080 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13081 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13082 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13083 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13085 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13088 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13089 of struct ix86_frame.
13090 (ix86_expand_epilogue): Likewise. Add a local variable for
13091 the reg_save_offset field in struct ix86_frame.
13093 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13095 PR tree-optimization/82604
13096 * tree-loop-distribution.c (enum partition_kind): New enum item
13097 PKIND_PARTIAL_MEMSET.
13098 (partition_builtin_p): Support above new enum item.
13099 (generate_code_for_partition): Ditto.
13100 (compute_access_range): Differentiate cases that equality can be
13101 proven at all loops, the innermost loops or no loops.
13102 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13103 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13104 (finalize_partitions, distribute_loop): Don't fuse partition of
13105 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13106 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13107 parloop is enabled.
13109 2018-01-23 Martin Liska <mliska@suse.cz>
13111 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13112 order to ignore the predictor.
13113 (PRED_POLYMORPHIC_CALL): Likewise.
13114 (PRED_RECURSIVE_CALL): Likewise.
13116 2018-01-23 Martin Liska <mliska@suse.cz>
13118 * tree-profile.c (tree_profiling): Print function header to
13119 aware reader which function we are working on.
13120 * value-prof.c (gimple_find_values_to_profile): Do not print
13121 not interesting value histograms.
13123 2018-01-23 Martin Liska <mliska@suse.cz>
13125 * profile-count.h (enum profile_quality): Add
13126 profile_uninitialized as the first value. Do not number values
13127 as they are zero based.
13128 (profile_count::verify): Update sanity check.
13129 (profile_probability::verify): Likewise.
13131 2018-01-23 Nathan Sidwell <nathan@acm.org>
13133 * doc/invoke.texi (ffor-scope): Deprecate.
13135 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13137 PR tree-optimization/83510
13138 * domwalk.c (set_all_edges_as_executable): New function.
13139 (dom_walker::dom_walker): Convert bool param
13140 "skip_unreachable_blocks" to enum reachability. Move setup of
13141 edge flags to set_all_edges_as_executable and only do it when
13142 reachability is REACHABLE_BLOCKS.
13143 * domwalk.h (enum dom_walker::reachability): New enum.
13144 (dom_walker::dom_walker): Convert bool param
13145 "skip_unreachable_blocks" to enum reachability.
13146 (set_all_edges_as_executable): New decl.
13147 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13148 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13150 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13151 but converting true to REACHABLE_BLOCKS.
13152 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13154 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13155 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13156 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13158 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13159 if check_all_array_refs will be called.
13161 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13163 * tree.c (selftest::test_location_wrappers): Add more test
13166 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13168 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13169 (selftest::test_bit_in_range): Likewise.
13171 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13174 * doc/sourcebuild.texi (vect_float): Say that the selector
13175 only describes the situation when -funsafe-math-optimizations is on.
13176 (vect_float_strict): Document.
13178 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13180 PR tree-optimization/83965
13181 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13182 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13183 instead of checking only for a reduction.
13184 (vect_recog_widen_sum_pattern): Likewise.
13186 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13188 * predict.c (probably_never_executed): Only use precise profile info.
13189 (compute_function_frequency): Skip after inlining hack since we now
13190 have quality checking.
13192 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13194 * profile-count.h (profile_probability::very_unlikely,
13195 profile_probability::unlikely, profile_probability::even): Set
13196 precision to guessed.
13198 2018-01-23 Richard Biener <rguenther@suse.de>
13200 PR tree-optimization/83963
13201 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13202 Properly terminate dominator walk when crossing the exit edge not
13203 when visiting its source block.
13205 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13208 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13209 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13211 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13213 PR tree-optimization/83957
13214 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13215 semicolon after for body surrounded by braces.
13217 PR tree-optimization/83081
13218 * profile-count.h (profile_probability::split): New method.
13219 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13220 Use profile_probability::split.
13221 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13222 when splitting a single conditional jump into 2.
13224 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13226 PR tree-optimization/69452
13227 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13232 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13233 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13234 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13236 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13238 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13240 * config/rl78/rl78.md (movdi): New define_expand.
13241 * config/rl78/rl78.c (rl78_split_movdi): New function.
13243 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13246 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13248 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13249 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13250 128-bit to produce an UNSPEC move to get the double word with the
13251 signbit and then a shift directly to do signbit.
13252 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13253 implementation with a new version that just does either a direct
13254 move or a regular move. Move memory interface to separate insns.
13255 Move insns so they are next to the expander.
13256 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13257 with signbit move. Split big and little endian case.
13258 (signbit<mode>2_dm_mem_le): Likewise.
13259 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13260 (signbit<mode>2_dm2): Likewise.
13262 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13264 * config/rl78/rl78.md (anddi3): New define_expand.
13266 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13268 * config/rl78/rl78.md (umindi3): New define_expand.
13270 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13272 * config/rl78/rl78.md (smindi3): New define_expand.
13274 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13276 * config/rl78/rl78.md (smaxdi3): New define_expand.
13278 2018-01-22 Carl Love <cel@us.ibm.com>
13280 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13281 LVX_V1TI): Add macro expansion.
13282 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13283 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13284 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13285 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13286 Change check to determine if the instruction is a byte reversing
13287 entry. Fix typo in comment.
13288 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13289 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13290 Add def_builtin calls for new builtins.
13291 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13292 Add define_insn expansion.
13294 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13296 * config/rl78/rl78.md (umaxdi3): New define_expand.
13298 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13300 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13301 for non-QImode registers.
13303 2018-01-22 Richard Biener <rguenther@suse.de>
13305 PR tree-optimization/83963
13306 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13307 including the loop exit block.
13308 (scop_detection::merge_sese): Likewise.
13309 (scop_detection::add_scop): Do it here instead.
13311 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13313 * doc/sourcebuild.texi (arm_softfloat): Document.
13315 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13318 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13319 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13320 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13322 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13323 David Edelsohn <dje.gcc@gmail.com>
13326 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13327 Change "crset eq" to "crset 2".
13328 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13329 (*call_indirect_aix<mode>_nospec): Likewise.
13330 (*call_value_indirect_aix<mode>_nospec): Likewise.
13331 (*call_indirect_elfv2<mode>_nospec): Likewise.
13332 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13333 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13334 change assembly output from . to $.
13335 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13336 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13337 (*tablejump<mode>_internal1_nospec): Likewise.
13339 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13342 * config/sh/sh_optimize_sett_clrt.cc:
13343 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13345 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13347 PR tree-optimization/83940
13348 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13349 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13350 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13351 use it to pass back the definition type.
13352 (vect_check_store_rhs): Likewise rhs_dt_out.
13353 (vect_build_gather_load_calls): Add a mask_dt argument and use
13354 it instead of a call to vect_is_simple_use.
13355 (vectorizable_store): Update calls to vect_check_load_store_mask
13356 and vect_check_store_rhs. Use the dt returned by the latter instead
13357 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13358 instead of calls to vect_is_simple_use. Pass the scalar rather
13359 than the vector operand to vect_is_simple_use when handling
13360 second and subsequent copies of an rhs value.
13361 (vectorizable_load): Update calls to vect_check_load_store_mask
13362 and vect_build_gather_load_calls. Use the cached mask_dt and
13363 gs_info.offset_dt instead of calls to vect_is_simple_use.
13365 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13367 PR middle-end/83945
13368 * tree-emutls.c: Include gimplify.h.
13369 (lower_emutls_2): New function.
13370 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13371 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13372 it before further processing.
13375 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13376 UINTVAL (trueop1) instead of INTVAL (op1).
13378 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13382 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13383 INCOMING_FRAME_SP_OFFSET if not defined.
13384 (scan_trace): Add ENTRY argument. If true and
13385 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13386 emit a note to adjust the CFA offset.
13387 (create_cfi_notes): Adjust scan_trace callers.
13388 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13389 INCOMING_FRAME_SP_OFFSET in the CIE.
13390 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13391 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13393 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13394 * doc/tm.texi: Regenerated.
13396 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13398 PR rtl-optimization/83147
13399 * lra-constraints.c (remove_inheritance_pseudos): Use
13400 lra_substitute_pseudo_within_insn.
13402 2018-01-19 Tom de Vries <tom@codesourcery.com>
13403 Cesar Philippidis <cesar@codesourcery.com>
13406 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13408 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13411 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13412 spaces for function labels.
13414 2018-01-19 Martin Liska <mliska@suse.cz>
13416 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13417 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13418 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13419 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13420 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13421 (PRED_CONST_RETURN): Change from 69 to 65.
13422 (PRED_NULL_RETURN): Change from 91 to 71.
13423 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13424 (PRED_LOOP_GUARD): Change from 66 to 73.
13426 2018-01-19 Martin Liska <mliska@suse.cz>
13428 * predict.c (predict_insn_def): Add new assert.
13429 (struct branch_predictor): Change type to signed integer.
13430 (test_prediction_value_range): Amend test to cover
13431 PROB_UNINITIALIZED.
13432 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13433 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13434 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13435 (PRED_LOOP_IV_COMPARE): Likewise.
13436 * predict.h (PROB_UNINITIALIZED): Define new constant.
13438 2018-01-19 Martin Liska <mliska@suse.cz>
13440 * predict.c (dump_prediction): Add new format for
13441 analyze_brprob.py script which is enabled with -details
13443 * profile-count.h (precise_p): New function.
13445 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13447 PR tree-optimization/83922
13448 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13449 there are no statements that need masking.
13450 (vect_active_double_reduction_p): New function.
13451 (vect_analyze_loop_operations): Use it when handling phis that
13452 are not in the loop header.
13454 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13456 PR tree-optimization/83914
13457 * tree-vect-loop.c (vectorizable_induction): Don't convert
13458 init_expr or apply the peeling adjustment for inductions
13459 that are nested within the vectorized loop.
13461 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13463 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13466 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13470 * function.h (gimplify_parameters): Add gimple_seq * argument.
13471 * function.c: Include gimple.h and options.h.
13472 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13473 for the added local temporaries if needed.
13474 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13475 if there are any parameter cleanups, wrap whole body into a
13476 try/finally with the cleanups.
13478 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13481 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13482 Use GET_MODE_CLASS for scalar floating point.
13484 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13488 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13489 Fix call of call_cgraph_insertion_hooks.
13491 2018-01-18 Martin Sebor <msebor@redhat.com>
13493 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13495 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13498 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13501 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13504 * common.opt: (-ffile-prefix-map): New option.
13505 * opts.c (common_handle_option): Defer it.
13506 * opts-global.c (handle_common_deferred_options): Handle it.
13507 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13508 * file-prefix-map.h: New file.
13509 (remap_debug_filename, add_debug_prefix_map): ...here.
13510 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13511 * final.c (debug_prefix_map, add_debug_prefix_map
13512 remap_debug_filename): Move to...
13513 * file-prefix-map.c: New file.
13514 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13515 generalize, get rid of alloca(), use strrchr() instead of strchr().
13516 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13517 Implement in terms of add_prefix_map().
13518 (remap_macro_filename, remap_debug_filename): Implement in term of
13520 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13521 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13522 * dbxout.c: Include file-prefix-map.h.
13523 * varasm.c: Likewise.
13524 * vmsdbgout.c: Likewise.
13525 * xcoffout.c: Likewise.
13526 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13527 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13528 * doc/invoke.texi (-ffile-prefix-map): Document.
13529 (-fdebug-prefix-map): Update description.
13531 2018-01-18 Martin Liska <mliska@suse.cz>
13533 * config/i386/i386.c (indirect_thunk_name): Document that also
13535 (output_indirect_thunk): Document why both instructions
13536 (pause and lfence) are generated.
13538 2018-01-18 Richard Biener <rguenther@suse.de>
13540 PR tree-optimization/83887
13541 * graphite-scop-detection.c
13542 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13543 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13544 (scop_detection::merge_sese): Re-implement with a flood-fill
13545 algorithm that properly finds a SESE region if it exists.
13547 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13550 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13551 pointer_diff optimizations use view_convert instead of convert.
13553 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13555 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13556 Generate different code for -mno-speculate-indirect-jumps.
13557 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13558 (*call_indirect_aix<mode>): Disable for
13559 -mno-speculate-indirect-jumps.
13560 (*call_indirect_aix<mode>_nospec): New define_insn.
13561 (*call_value_indirect_aix<mode>): Disable for
13562 -mno-speculate-indirect-jumps.
13563 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13564 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13565 -mno-speculate-indirect-jumps.
13566 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13568 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13570 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13571 long double type, set the flags for noting the default long double
13572 type, even if we don't pass or return a long double type.
13574 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13577 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13580 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13582 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13583 support for merge[hl].
13584 (fold_mergehl_helper): New helper function.
13585 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13586 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13587 (altivec_vmrglw_direct): Add xxmrglw insn.
13589 2018-01-17 Andrew Waterman <andrew@sifive.com>
13591 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13592 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13594 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13597 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13598 call the lto_location_cache before reading the
13599 DECL_SOURCE_LOCATION of the types.
13601 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13602 Richard Sandiford <richard.sandiford@linaro.org>
13604 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13605 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13606 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13607 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13608 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13610 * config/aarch64/constraints.md (aarch64_movti_operand):
13612 * config/aarch64/predicates.md (Uti): Add new constraint.
13614 2018-01-17 Carl Love <cel@us.ibm.com>
13616 * config/rs6000/vsx.md (define_expand xl_len_r,
13617 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13618 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13620 (define_expand, define_insn): Move the shift left from the
13621 define_insn to the define_expand for lxvl and stxvl instructions.
13622 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13623 and XL_LEN_R definitions to PURE.
13625 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13627 * config/i386/i386.c (indirect_thunk_name): Declare regno
13628 as unsigned int. Compare regno with INVALID_REGNUM.
13629 (output_indirect_thunk): Ditto.
13630 (output_indirect_thunk_function): Ditto.
13631 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13632 in the call to output_indirect_thunk_function.
13634 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13636 PR middle-end/83884
13637 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13638 rather than the size of inner_type to determine the stack slot size
13639 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13641 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13644 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13647 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13649 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13650 endian Linux systems to optionally enable multilibs for selecting
13651 the long double type if the user configured an explicit type.
13652 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13653 have no long double multilibs if not defined.
13654 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13655 warn if the user used -mabi={ieee,ibm}longdouble and we built
13656 multilibs for long double.
13657 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13658 appropriate multilib option.
13659 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13661 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13662 for building long double multilibs.
13663 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13665 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13667 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13670 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13672 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13675 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13678 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13681 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13683 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13684 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13686 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13688 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13689 different rtl trees depending on TARGET_64BIT.
13690 (rs6000_gen_lvx): Likewise.
13692 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13694 * config/visium/visium.md (nop): Tweak comment.
13695 (hazard_nop): Likewise.
13697 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13699 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13700 -mspeculate-indirect-jumps.
13701 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13702 for -mno-speculate-indirect-jumps.
13703 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13704 (*call_value_indirect_elfv2<mode>): Disable for
13705 -mno-speculate-indirect-jumps.
13706 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13707 (indirect_jump): Emit different RTL for
13708 -mno-speculate-indirect-jumps.
13709 (*indirect_jump<mode>): Disable for
13710 -mno-speculate-indirect-jumps.
13711 (*indirect_jump<mode>_nospec): New define_insn.
13712 (tablejump): Emit different RTL for
13713 -mno-speculate-indirect-jumps.
13714 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13715 (tablejumpsi_nospec): New define_expand.
13716 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13717 (tablejumpdi_nospec): New define_expand.
13718 (*tablejump<mode>_internal1): Disable for
13719 -mno-speculate-indirect-jumps.
13720 (*tablejump<mode>_internal1_nospec): New define_insn.
13721 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13724 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13726 * caller-save.c (insert_save): Drop unnecessary parameter. All
13729 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13730 Richard Biener <rguenth@suse.de>
13733 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13734 return early, inline manually is_gimple_sizepos. Make sure if we
13735 call gimplify_expr we don't end up with a gimple constant.
13736 * tree.c (variably_modified_type_p): Don't return true for
13737 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13738 * gimplify.h (is_gimple_sizepos): Remove.
13740 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13742 PR tree-optimization/83857
13743 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13744 vectorizable_live_operation for pure SLP statements.
13745 (vectorizable_live_operation): Handle PHIs.
13747 2018-01-16 Richard Biener <rguenther@suse.de>
13749 PR tree-optimization/83867
13750 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13751 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13753 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13756 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13757 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13758 If off is not INTEGER_CST, issue a may not be aligned warning
13759 rather than isn't aligned. Use isn%'t rather than isn't.
13760 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13762 <case MULT_EXPR>: Improve the case when bottom and one of the
13763 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13764 operand, in that case check if the other operand is multiple of
13765 bottom divided by the INTEGER_CST operand.
13767 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13770 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13771 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13772 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13773 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13774 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13775 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13776 (pa_function_arg_size): New function.
13778 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13780 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13781 in a separate statement.
13783 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13785 PR tree-optimization/83847
13786 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13787 group gathers and scatters.
13789 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13791 PR rtl-optimization/86620
13792 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13794 PR rtl-optimization/83213
13795 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13796 to last if both are JUMP_INSNs.
13798 PR tree-optimization/83843
13799 * gimple-ssa-store-merging.c
13800 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13801 store_immediate_info for bswap/nop orig_stores.
13803 2018-01-15 Andrew Waterman <andrew@sifive.com>
13805 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13807 <UDIV>: Increase cost if !TARGET_DIV.
13809 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13811 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13812 (define_attr "cr_logical_3op"): New.
13813 (cceq_ior_compare): Adjust.
13814 (cceq_ior_compare_complement): Adjust.
13815 (*cceq_rev_compare): Adjust.
13816 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13817 (is_cracked_insn): Adjust.
13818 (insn_must_be_first_in_group): Adjust.
13819 * config/rs6000/40x.md: Adjust.
13820 * config/rs6000/440.md: Adjust.
13821 * config/rs6000/476.md: Adjust.
13822 * config/rs6000/601.md: Adjust.
13823 * config/rs6000/603.md: Adjust.
13824 * config/rs6000/6xx.md: Adjust.
13825 * config/rs6000/7450.md: Adjust.
13826 * config/rs6000/7xx.md: Adjust.
13827 * config/rs6000/8540.md: Adjust.
13828 * config/rs6000/cell.md: Adjust.
13829 * config/rs6000/e300c2c3.md: Adjust.
13830 * config/rs6000/e500mc.md: Adjust.
13831 * config/rs6000/e500mc64.md: Adjust.
13832 * config/rs6000/e5500.md: Adjust.
13833 * config/rs6000/e6500.md: Adjust.
13834 * config/rs6000/mpc.md: Adjust.
13835 * config/rs6000/power4.md: Adjust.
13836 * config/rs6000/power5.md: Adjust.
13837 * config/rs6000/power6.md: Adjust.
13838 * config/rs6000/power7.md: Adjust.
13839 * config/rs6000/power8.md: Adjust.
13840 * config/rs6000/power9.md: Adjust.
13841 * config/rs6000/rs64.md: Adjust.
13842 * config/rs6000/titan.md: Adjust.
13844 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13846 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13847 ix86_indirect_branch_register logic.
13849 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13851 * config/i386/constraints.md (Bs): Update
13852 ix86_indirect_branch_register check. Don't check
13853 ix86_indirect_branch_register with GOT_memory_operand.
13855 * config/i386/predicates.md (GOT_memory_operand): Don't check
13856 ix86_indirect_branch_register here.
13857 (GOT32_symbol_operand): Likewise.
13859 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13861 * config/i386/predicates.md (constant_call_address_operand):
13862 Rewrite ix86_indirect_branch_register logic.
13863 (sibcall_insn_operand): Likewise.
13865 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13867 * config/i386/constraints.md (Bs): Replace
13868 ix86_indirect_branch_thunk_register with
13869 ix86_indirect_branch_register.
13871 * config/i386/i386.md (indirect_jump): Likewise.
13872 (tablejump): Likewise.
13873 (*sibcall_memory): Likewise.
13874 (*sibcall_value_memory): Likewise.
13875 Peepholes of indirect call and jump via memory: Likewise.
13876 * config/i386/i386.opt: Likewise.
13877 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13878 (GOT_memory_operand): Likewise.
13879 (call_insn_operand): Likewise.
13880 (sibcall_insn_operand): Likewise.
13881 (GOT32_symbol_operand): Likewise.
13883 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13885 PR middle-end/83837
13886 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13887 type rather than type addr's type points to.
13888 (expand_omp_atomic_mutex): Likewise.
13889 (expand_omp_atomic): Likewise.
13891 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13894 * config/i386/i386.c (output_indirect_thunk_function): Use
13895 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13896 for __x86_return_thunk.
13898 2018-01-15 Richard Biener <rguenther@suse.de>
13900 PR middle-end/83850
13901 * expmed.c (extract_bit_field_1): Fix typo.
13903 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13906 * config/arm/iterators.md (VF): New mode iterator.
13907 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13908 Remove integer-related logic from pattern.
13909 (neon_vabd<mode>_3): Likewise.
13911 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13913 PR middle-end/82694
13914 * common.opt (fstrict-overflow): No longer an alias.
13915 (fwrapv-pointer): New option.
13916 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13917 also for pointer types based on flag_wrapv_pointer.
13918 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13919 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13920 opts->x_flag_wrapv got set.
13921 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13922 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13923 POINTER_TYPE_OVERFLOW_UNDEFINED.
13924 * match.pd: Likewise in address comparison pattern.
13925 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13927 2018-01-15 Richard Biener <rguenther@suse.de>
13930 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13931 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13932 Reset type names to their identifier if their TYPE_DECL doesn't
13933 have linkage (and thus is used for ODR and devirt).
13934 (save_debug_info_for_decl): Remove.
13935 (save_debug_info_for_type): Likewise.
13936 (add_tree_to_fld_list): Adjust.
13937 * tree-pretty-print.c (dump_generic_node): Make dumping of
13938 type names more robust.
13940 2018-01-15 Richard Biener <rguenther@suse.de>
13942 * BASE-VER: Bump to 8.0.1.
13944 2018-01-14 Martin Sebor <msebor@redhat.com>
13947 * builtins.c (check_access): Avoid warning when the no-warning bit
13950 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13952 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13953 * ira-color (allocno_hard_regs_compare): Likewise.
13955 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13958 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13959 Use .pushsection/.popsection.
13961 2018-01-14 Martin Sebor <msebor@redhat.com>
13964 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13966 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13968 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13969 entry from extra_headers.
13970 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13971 extra_headers, make the list bitwise identical to the i?86-*-* one.
13973 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13975 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13976 -mcmodel=large with -mindirect-branch=thunk,
13977 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13978 -mfunction-return=thunk-extern.
13979 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13980 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13981 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13983 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13985 * config/i386/i386.c (print_reg): Print the name of the full
13986 integer register without '%'.
13987 (ix86_print_operand): Handle 'V'.
13988 * doc/extend.texi: Document 'V' modifier.
13990 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13992 * config/i386/constraints.md (Bs): Disallow memory operand for
13993 -mindirect-branch-register.
13995 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13996 (GOT_memory_operand): Likewise.
13997 (call_insn_operand): Likewise.
13998 (sibcall_insn_operand): Likewise.
13999 (GOT32_symbol_operand): Likewise.
14000 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14001 for -mindirect-branch-register.
14002 (tablejump): Likewise.
14003 (*sibcall_memory): Likewise.
14004 (*sibcall_value_memory): Likewise.
14005 Disallow peepholes of indirect call and jump via memory for
14006 -mindirect-branch-register.
14007 (*call_pop): Replace m with Bw.
14008 (*call_value_pop): Likewise.
14009 (*sibcall_pop_memory): Replace m with Bs.
14010 * config/i386/i386.opt (mindirect-branch-register): New option.
14011 * doc/invoke.texi: Document -mindirect-branch-register option.
14013 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14015 * config/i386/i386-protos.h (ix86_output_function_return): New.
14016 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14017 set function_return_type.
14018 (indirect_thunk_name): Add ret_p to indicate thunk for function
14020 (output_indirect_thunk_function): Pass false to
14021 indirect_thunk_name.
14022 (ix86_output_indirect_branch_via_reg): Likewise.
14023 (ix86_output_indirect_branch_via_push): Likewise.
14024 (output_indirect_thunk_function): Create alias for function
14025 return thunk if regno < 0.
14026 (ix86_output_function_return): New function.
14027 (ix86_handle_fndecl_attribute): Handle function_return.
14028 (ix86_attribute_table): Add function_return.
14029 * config/i386/i386.h (machine_function): Add
14030 function_return_type.
14031 * config/i386/i386.md (simple_return_internal): Use
14032 ix86_output_function_return.
14033 (simple_return_internal_long): Likewise.
14034 * config/i386/i386.opt (mfunction-return=): New option.
14035 (indirect_branch): Mention -mfunction-return=.
14036 * doc/extend.texi: Document function_return function attribute.
14037 * doc/invoke.texi: Document -mfunction-return= option.
14039 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14041 * config/i386/i386-opts.h (indirect_branch): New.
14042 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14043 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14044 with local indirect jump when converting indirect call and jump.
14045 (ix86_set_indirect_branch_type): New.
14046 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14047 (indirectlabelno): New.
14048 (indirect_thunk_needed): Likewise.
14049 (indirect_thunk_bnd_needed): Likewise.
14050 (indirect_thunks_used): Likewise.
14051 (indirect_thunks_bnd_used): Likewise.
14052 (INDIRECT_LABEL): Likewise.
14053 (indirect_thunk_name): Likewise.
14054 (output_indirect_thunk): Likewise.
14055 (output_indirect_thunk_function): Likewise.
14056 (ix86_output_indirect_branch_via_reg): Likewise.
14057 (ix86_output_indirect_branch_via_push): Likewise.
14058 (ix86_output_indirect_branch): Likewise.
14059 (ix86_output_indirect_jmp): Likewise.
14060 (ix86_code_end): Call output_indirect_thunk_function if needed.
14061 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14063 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14064 (ix86_attribute_table): Add indirect_branch.
14065 * config/i386/i386.h (machine_function): Add indirect_branch_type
14066 and has_local_indirect_jump.
14067 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14069 (tablejump): Likewise.
14070 (*indirect_jump): Use ix86_output_indirect_jmp.
14071 (*tablejump_1): Likewise.
14072 (simple_return_indirect_internal): Likewise.
14073 * config/i386/i386.opt (mindirect-branch=): New option.
14074 (indirect_branch): New.
14077 (thunk-inline): Likewise.
14078 (thunk-extern): Likewise.
14079 * doc/extend.texi: Document indirect_branch function attribute.
14080 * doc/invoke.texi: Document -mindirect-branch= option.
14082 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14085 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14087 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14089 * ipa-inline.c (want_inline_small_function_p): Return false if
14090 inlining has already failed with CIF_FINAL_ERROR.
14091 (update_caller_keys): Call want_inline_small_function_p before
14093 (update_callee_keys): Likewise.
14095 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14097 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14099 (rs6000_quadword_masked_address_p): Likewise.
14100 (quad_aligned_load_p): Likewise.
14101 (quad_aligned_store_p): Likewise.
14102 (const_load_sequence_p): Add comment to describe the outer-most loop.
14103 (mimic_memory_attributes_and_flags): New function.
14104 (rs6000_gen_stvx): Likewise.
14105 (replace_swapped_aligned_store): Likewise.
14106 (rs6000_gen_lvx): Likewise.
14107 (replace_swapped_aligned_load): Likewise.
14108 (replace_swapped_load_constant): Capitalize argument name in
14109 comment describing this function.
14110 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14111 and stores that access quad-word aligned addresses and replace
14112 with stvx or lvx instructions when appropriate.
14113 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14114 New function prototype.
14115 (rs6000_quadword_masked_address_p): Likewise.
14116 (rs6000_gen_lvx): Likewise.
14117 (rs6000_gen_stvx): Likewise.
14118 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14119 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14120 when memory address is aligned.
14121 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14122 this split to select lvx instruction when memory address is aligned.
14123 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14124 instruction when memory address is aligned.
14125 (*vsx_le_perm_load_v16qi): Likewise.
14126 (four unnamed splitters): Modify to select the stvx instruction
14127 when memory is aligned.
14129 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14131 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14132 which appears in the queue multiple times.
14134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14135 Alan Hayward <alan.hayward@arm.com>
14136 David Sherwood <david.sherwood@arm.com>
14138 * tree-vectorizer.h (vec_lower_bound): New structure.
14139 (_loop_vec_info): Add check_nonzero and lower_bounds.
14140 (LOOP_VINFO_CHECK_NONZERO): New macro.
14141 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14142 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14143 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14144 fields. Make seg_len the distance travelled, not including the
14146 (dr_direction_indicator): Declare.
14147 (dr_zero_step_indicator): Likewise.
14148 (dr_known_forward_stride_p): Likewise.
14149 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14151 (runtime_alias_check_p): Allow runtime alias checks with
14153 (operator ==): Compare access_size and align.
14154 (prune_runtime_alias_test_list): Rework for new distinction between
14155 the access_size and seg_len.
14156 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14158 (get_segment_min_max): New function.
14159 (create_intersect_range_checks): Use it.
14160 (dr_step_indicator): New function.
14161 (dr_direction_indicator): Likewise.
14162 (dr_zero_step_indicator): Likewise.
14163 (dr_known_forward_stride_p): Likewise.
14164 * tree-loop-distribution.c (data_ref_segment_size): Return
14165 DR_STEP * (niters - 1).
14166 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14168 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14169 (vect_preserves_scalar_order_p): New function, split out from...
14170 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14171 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14172 (vect_vfa_access_size): New function.
14173 (vect_vfa_align): Likewise.
14174 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14175 (dump_lower_bound): New function.
14176 (vect_check_lower_bound): Likewise.
14177 (vect_small_gap_p): Likewise.
14178 (vectorizable_with_step_bound_p): Likewise.
14179 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14180 depencies if the vectorization factor is 1. Convert the checks
14181 for nonzero steps into checks on the bounds of DR_STEP. Try using
14182 a bunds check for variable steps if the minimum required step is
14183 relatively small. Update calls to the dr_with_seg_len
14184 constructor and to vect_compile_time_alias.
14185 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14187 (vect_loop_versioning): Call it.
14188 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14190 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14193 Alan Hayward <alan.hayward@arm.com>
14194 David Sherwood <david.sherwood@arm.com>
14196 * doc/sourcebuild.texi (vect_scatter_store): Document.
14197 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14199 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14201 * genopinit.c (main): Add supports_vec_scatter_store and
14202 supports_vec_scatter_store_cached to target_optabs.
14203 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14204 IFN_MASK_SCATTER_STORE.
14205 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14207 * internal-fn.h (internal_store_fn_p): Declare.
14208 (internal_fn_stored_value_index): Likewise.
14209 * internal-fn.c (scatter_store_direct): New macro.
14210 (expand_scatter_store_optab_fn): New function.
14211 (direct_scatter_store_optab_supported_p): New macro.
14212 (internal_store_fn_p): New function.
14213 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14214 IFN_MASK_SCATTER_STORE.
14215 (internal_fn_mask_index): Likewise.
14216 (internal_fn_stored_value_index): New function.
14217 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14218 for scatter stores.
14219 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14220 * optabs-query.c (supports_vec_scatter_store_p): New function.
14221 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14222 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14223 true for scatter stores.
14224 (vect_gather_scatter_fn_p): Handle scatter stores too.
14225 (vect_check_gather_scatter): Consider using scatter stores if
14226 supports_vec_scatter_store_p.
14227 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14228 scatter stores too.
14229 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14230 internal_fn_stored_value_index.
14231 (check_load_store_masking): Handle scatter stores too.
14232 (vect_get_store_rhs): Make public.
14233 (vectorizable_call): Use internal_store_fn_p.
14234 (vectorizable_store): Handle scatter store internal functions.
14235 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14236 when deciding whether the end of the group has been reached.
14237 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14238 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14239 (mask_scatter_store<mode>): New insns.
14241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14242 Alan Hayward <alan.hayward@arm.com>
14243 David Sherwood <david.sherwood@arm.com>
14245 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14246 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14247 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14249 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14250 Use vect_truncate_gather_scatter_offset if we can't treat the
14251 operation as a normal gather load or scatter store.
14252 (get_group_load_store_type): Take the gather_scatter_info
14253 as argument. Try using a gather load or scatter store for
14254 single-element groups.
14255 (get_load_store_type): Update calls to get_group_load_store_type
14256 and vect_use_strided_gather_scatters_p.
14258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14259 Alan Hayward <alan.hayward@arm.com>
14260 David Sherwood <david.sherwood@arm.com>
14262 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14263 optional tree argument.
14264 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14266 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14267 but continue to use the current value as a fallback.
14268 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14269 to compare the updates.
14270 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14271 (get_load_store_type): Use it when handling a strided access.
14272 (vect_get_strided_load_store_ops): New function.
14273 (vect_get_data_ptr_increment): Likewise.
14274 (vectorizable_load): Handle strided gather loads. Always pass
14275 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14278 Alan Hayward <alan.hayward@arm.com>
14279 David Sherwood <david.sherwood@arm.com>
14281 * doc/md.texi (gather_load@var{m}): Document.
14282 (mask_gather_load@var{m}): Likewise.
14283 * genopinit.c (main): Add supports_vec_gather_load and
14284 supports_vec_gather_load_cached to target_optabs.
14285 * optabs-tree.c (init_tree_optimization_optabs): Use
14286 ggc_cleared_alloc to allocate target_optabs.
14287 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14288 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14290 * internal-fn.h (internal_load_fn_p): Declare.
14291 (internal_gather_scatter_fn_p): Likewise.
14292 (internal_fn_mask_index): Likewise.
14293 (internal_gather_scatter_fn_supported_p): Likewise.
14294 * internal-fn.c (gather_load_direct): New macro.
14295 (expand_gather_load_optab_fn): New function.
14296 (direct_gather_load_optab_supported_p): New macro.
14297 (direct_internal_fn_optab): New function.
14298 (internal_load_fn_p): Likewise.
14299 (internal_gather_scatter_fn_p): Likewise.
14300 (internal_fn_mask_index): Likewise.
14301 (internal_gather_scatter_fn_supported_p): Likewise.
14302 * optabs-query.c (supports_at_least_one_mode_p): New function.
14303 (supports_vec_gather_load_p): Likewise.
14304 * optabs-query.h (supports_vec_gather_load_p): Declare.
14305 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14306 and memory_type field.
14307 (NUM_PATTERNS): Bump to 15.
14308 * tree-vect-data-refs.c: Include internal-fn.h.
14309 (vect_gather_scatter_fn_p): New function.
14310 (vect_describe_gather_scatter_call): Likewise.
14311 (vect_check_gather_scatter): Try using internal functions for
14312 gather loads. Recognize existing calls to a gather load function.
14313 (vect_analyze_data_refs): Consider using gather loads if
14314 supports_vec_gather_load_p.
14315 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14316 (vect_get_gather_scatter_offset_type): Likewise.
14317 (vect_convert_mask_for_vectype): Likewise.
14318 (vect_add_conversion_to_patterm): Likewise.
14319 (vect_try_gather_scatter_pattern): Likewise.
14320 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14321 (vect_vect_recog_func_ptrs): Add it.
14322 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14323 internal_fn_mask_index and internal_gather_scatter_fn_p.
14324 (check_load_store_masking): Take the gather_scatter_info as an
14325 argument and handle gather loads.
14326 (vect_get_gather_scatter_ops): New function.
14327 (vectorizable_call): Check internal_load_fn_p.
14328 (vectorizable_load): Likewise. Handle gather load internal
14330 (vectorizable_store): Update call to check_load_store_masking.
14331 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14332 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14333 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14334 (aarch64_gather_scale_operand_d): New predicates.
14335 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14336 (mask_gather_load<mode>): New insns.
14338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14339 Alan Hayward <alan.hayward@arm.com>
14340 David Sherwood <david.sherwood@arm.com>
14342 * optabs.def (fold_left_plus_optab): New optab.
14343 * doc/md.texi (fold_left_plus_@var{m}): Document.
14344 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14345 * internal-fn.c (fold_left_direct): Define.
14346 (expand_fold_left_optab_fn): Likewise.
14347 (direct_fold_left_optab_supported_p): Likewise.
14348 * fold-const-call.c (fold_const_fold_left): New function.
14349 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14350 * tree-parloops.c (valid_reduction_p): New function.
14351 (gather_scalar_reductions): Use it.
14352 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14353 (vect_finish_replace_stmt): Declare.
14354 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14355 (needs_fold_left_reduction_p): New function, split out from...
14356 (vect_is_simple_reduction): ...here. Accept reductions that
14357 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14358 (vect_force_simple_reduction): Also store the reduction type in
14359 the assignment's STMT_VINFO_REDUC_TYPE.
14360 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14361 (merge_with_identity): New function.
14362 (vect_expand_fold_left): Likewise.
14363 (vectorize_fold_left_reduction): Likewise.
14364 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14365 scalar phi in place for it. Check for target support and reject
14366 cases that would reassociate the operation. Defer the transform
14367 phase to vectorize_fold_left_reduction.
14368 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14369 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14370 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14374 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14375 call to ifc_temp_var.
14377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14378 Alan Hayward <alan.hayward@arm.com>
14379 David Sherwood <david.sherwood@arm.com>
14381 * target.def (legitimize_address_displacement): Take the original
14382 offset as a poly_int.
14383 * targhooks.h (default_legitimize_address_displacement): Update
14385 * targhooks.c (default_legitimize_address_displacement): Likewise.
14386 * doc/tm.texi: Regenerate.
14387 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14388 as an argument, moving assert of ad->disp == ad->disp_term to...
14389 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14390 Try calling targetm.legitimize_address_displacement before expanding
14391 the address rather than afterwards, and adjust for the new interface.
14392 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14393 Match the new hook interface. Handle SVE addresses.
14394 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14395 new hook interface.
14397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14399 * Makefile.in (OBJS): Add early-remat.o.
14400 * target.def (select_early_remat_modes): New hook.
14401 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14402 * doc/tm.texi: Regenerate.
14403 * targhooks.h (default_select_early_remat_modes): Declare.
14404 * targhooks.c (default_select_early_remat_modes): New function.
14405 * timevar.def (TV_EARLY_REMAT): New timevar.
14406 * passes.def (pass_early_remat): New pass.
14407 * tree-pass.h (make_pass_early_remat): Declare.
14408 * early-remat.c: New file.
14409 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14411 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14414 Alan Hayward <alan.hayward@arm.com>
14415 David Sherwood <david.sherwood@arm.com>
14417 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14418 vfm1 with a bound_epilog parameter.
14419 (vect_do_peeling): Update calls accordingly, and move the prologue
14420 call earlier in the function. Treat the base bound_epilog as 0 for
14421 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14422 this base when peeling for gaps.
14423 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14424 with fully-masked loops.
14425 (vect_estimate_min_profitable_iters): Handle the single peeled
14426 iteration in that case.
14428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14429 Alan Hayward <alan.hayward@arm.com>
14430 David Sherwood <david.sherwood@arm.com>
14432 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14433 single-element interleaving even if the size is not a power of 2.
14434 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14435 accesses for single-element interleaving if the group size is
14438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14439 Alan Hayward <alan.hayward@arm.com>
14440 David Sherwood <david.sherwood@arm.com>
14442 * doc/md.texi (fold_extract_last_@var{m}): Document.
14443 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14444 * optabs.def (fold_extract_last_optab): New optab.
14445 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14446 * internal-fn.c (fold_extract_direct): New macro.
14447 (expand_fold_extract_optab_fn): Likewise.
14448 (direct_fold_extract_optab_supported_p): Likewise.
14449 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14450 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14451 EXTRACT_LAST_REDUCTION.
14452 (get_initial_def_for_reduction): Do not create an initial vector
14453 for EXTRACT_LAST_REDUCTION reductions.
14454 (vectorizable_reduction): Leave the scalar phi in place for
14455 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14456 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14457 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14458 transform phase to vectorizable_condition.
14459 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14461 (vect_finish_stmt_generation): ...here.
14462 (vect_finish_replace_stmt): New function.
14463 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14464 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14466 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14469 Alan Hayward <alan.hayward@arm.com>
14470 David Sherwood <david.sherwood@arm.com>
14472 * doc/md.texi (extract_last_@var{m}): Document.
14473 * optabs.def (extract_last_optab): New optab.
14474 * internal-fn.def (EXTRACT_LAST): New internal function.
14475 * internal-fn.c (cond_unary_direct): New macro.
14476 (expand_cond_unary_optab_fn): Likewise.
14477 (direct_cond_unary_optab_supported_p): Likewise.
14478 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14479 loops using EXTRACT_LAST.
14480 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14481 (extract_last_<mode>): ...this optab.
14482 (vec_extract<mode><Vel>): Update accordingly.
14484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14485 Alan Hayward <alan.hayward@arm.com>
14486 David Sherwood <david.sherwood@arm.com>
14488 * target.def (empty_mask_is_expensive): New hook.
14489 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14490 * doc/tm.texi: Regenerate.
14491 * targhooks.h (default_empty_mask_is_expensive): Declare.
14492 * targhooks.c (default_empty_mask_is_expensive): New function.
14493 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14494 if the target says that empty masks are expensive.
14495 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14497 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14500 Alan Hayward <alan.hayward@arm.com>
14501 David Sherwood <david.sherwood@arm.com>
14503 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14504 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14505 (vect_use_loop_mask_for_alignment_p): New function.
14506 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14507 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14508 niters_skip argument. Make sure that the first niters_skip elements
14509 of the first iteration are inactive.
14510 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14511 Update call to vect_set_loop_masks_directly.
14512 (get_misalign_in_elems): New function, split out from...
14513 (vect_gen_prolog_loop_niters): ...here.
14514 (vect_update_init_of_dr): Take a code argument that specifies whether
14515 the adjustment should be added or subtracted.
14516 (vect_update_init_of_drs): Likewise.
14517 (vect_prepare_for_masked_peels): New function.
14518 (vect_do_peeling): Skip prologue peeling if we're using a mask
14519 instead. Update call to vect_update_inits_of_drs.
14520 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14522 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14523 alignment. Do not include the number of peeled iterations in
14524 the minimum threshold in that case.
14525 (vectorizable_induction): Adjust the start value down by
14526 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14527 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14528 Take the number of skipped iterations into account when calculating
14530 * tree-vect-stmts.c (vect_gen_while_not): New function.
14532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14533 Alan Hayward <alan.hayward@arm.com>
14534 David Sherwood <david.sherwood@arm.com>
14536 * doc/sourcebuild.texi (vect_fully_masked): Document.
14537 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14538 default value to 0.
14539 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14541 (vect_analyze_loop_2): ...here. Don't check the vectorization
14542 factor against the number of loop iterations if the loop is
14545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14546 Alan Hayward <alan.hayward@arm.com>
14547 David Sherwood <david.sherwood@arm.com>
14549 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14550 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14551 (dump_groups): Update accordingly.
14552 (iv_use::mem_type): New member variable.
14553 (address_p): New function.
14554 (record_use): Add a mem_type argument and initialize the new
14556 (record_group_use): Add a mem_type argument. Use address_p.
14557 Remove obsolete null checks of base_object. Update call to record_use.
14558 (find_interesting_uses_op): Update call to record_group_use.
14559 (find_interesting_uses_cond): Likewise.
14560 (find_interesting_uses_address): Likewise.
14561 (get_mem_type_for_internal_fn): New function.
14562 (find_address_like_use): Likewise.
14563 (find_interesting_uses_stmt): Try find_address_like_use before
14564 calling find_interesting_uses_op.
14565 (addr_offset_valid_p): Use the iv mem_type field as the type
14566 of the addressed memory.
14567 (add_autoinc_candidates): Likewise.
14568 (get_address_cost): Likewise.
14569 (split_small_address_groups_p): Use address_p.
14570 (split_address_groups): Likewise.
14571 (add_iv_candidate_for_use): Likewise.
14572 (autoinc_possible_for_pair): Likewise.
14573 (rewrite_groups): Likewise.
14574 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14575 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14576 (get_alias_ptr_type_for_ptr_address): New function.
14577 (rewrite_use_address): Rewrite address uses in calls that were
14578 identified by find_address_like_use.
14580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14581 Alan Hayward <alan.hayward@arm.com>
14582 David Sherwood <david.sherwood@arm.com>
14584 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14586 * gimple-expr.h (is_gimple_addressable: Likewise.
14587 * gimple-expr.c (is_gimple_address): Likewise.
14588 * internal-fn.c (expand_call_mem_ref): New function.
14589 (expand_mask_load_optab_fn): Use it.
14590 (expand_mask_store_optab_fn): Likewise.
14592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14593 Alan Hayward <alan.hayward@arm.com>
14594 David Sherwood <david.sherwood@arm.com>
14596 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14597 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14598 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14599 (cond_umax@var{mode}): Document.
14600 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14601 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14602 (cond_umin_optab, cond_umax_optab): New optabs.
14603 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14604 (COND_IOR, COND_XOR): New internal functions.
14605 * internal-fn.h (get_conditional_internal_fn): Declare.
14606 * internal-fn.c (cond_binary_direct): New macro.
14607 (expand_cond_binary_optab_fn): Likewise.
14608 (direct_cond_binary_optab_supported_p): Likewise.
14609 (get_conditional_internal_fn): New function.
14610 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14611 Cope with reduction statements that are vectorized as calls rather
14613 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14614 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14615 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14616 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14617 (UNSPEC_COND_EOR): New unspecs.
14618 (optab): Add mappings for them.
14619 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14620 (sve_int_op, sve_fp_op): New int attributes.
14622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14623 Alan Hayward <alan.hayward@arm.com>
14624 David Sherwood <david.sherwood@arm.com>
14626 * optabs.def (while_ult_optab): New optab.
14627 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14628 * internal-fn.def (WHILE_ULT): New internal function.
14629 * internal-fn.h (direct_internal_fn_supported_p): New override
14630 that takes two types as argument.
14631 * internal-fn.c (while_direct): New macro.
14632 (expand_while_optab_fn): New function.
14633 (convert_optab_supported_p): Likewise.
14634 (direct_while_optab_supported_p): New macro.
14635 * wide-int.h (wi::udiv_ceil): New function.
14636 * tree-vectorizer.h (rgroup_masks): New structure.
14637 (vec_loop_masks): New typedef.
14638 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14639 and fully_masked_p.
14640 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14641 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14642 (vect_max_vf): New function.
14643 (slpeel_make_loop_iterate_ntimes): Delete.
14644 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14645 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14646 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14647 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14648 internal-fn.h, stor-layout.h and optabs-query.h.
14649 (vect_set_loop_mask): New function.
14650 (add_preheader_seq): Likewise.
14651 (add_header_seq): Likewise.
14652 (interleave_supported_p): Likewise.
14653 (vect_maybe_permute_loop_masks): Likewise.
14654 (vect_set_loop_masks_directly): Likewise.
14655 (vect_set_loop_condition_masked): Likewise.
14656 (vect_set_loop_condition_unmasked): New function, split out from
14657 slpeel_make_loop_iterate_ntimes.
14658 (slpeel_make_loop_iterate_ntimes): Rename to..
14659 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14660 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14661 (vect_do_peeling): Update call accordingly.
14662 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14664 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14665 mask_compare_type, can_fully_mask_p and fully_masked_p.
14666 (release_vec_loop_masks): New function.
14667 (_loop_vec_info): Use it to free the loop masks.
14668 (can_produce_all_loop_masks_p): New function.
14669 (vect_get_max_nscalars_per_iter): Likewise.
14670 (vect_verify_full_masking): Likewise.
14671 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14672 retries, and free the mask rgroups before retrying. Check loop-wide
14673 reasons for disallowing fully-masked loops. Make the final decision
14674 about whether use a fully-masked loop or not.
14675 (vect_estimate_min_profitable_iters): Do not assume that peeling
14676 for the number of iterations will be needed for fully-masked loops.
14677 (vectorizable_reduction): Disable fully-masked loops.
14678 (vectorizable_live_operation): Likewise.
14679 (vect_halve_mask_nunits): New function.
14680 (vect_double_mask_nunits): Likewise.
14681 (vect_record_loop_mask): Likewise.
14682 (vect_get_loop_mask): Likewise.
14683 (vect_transform_loop): Handle the case in which the final loop
14684 iteration might handle a partial vector. Call vect_set_loop_condition
14685 instead of slpeel_make_loop_iterate_ntimes.
14686 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14687 (check_load_store_masking): New function.
14688 (prepare_load_store_mask): Likewise.
14689 (vectorizable_store): Handle fully-masked loops.
14690 (vectorizable_load): Likewise.
14691 (supportable_widening_operation): Use vect_halve_mask_nunits for
14693 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14694 (vect_gen_while): New function.
14695 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14696 (aarch64_uqdec<mode>): New insn.
14698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14699 Alan Hayward <alan.hayward@arm.com>
14700 David Sherwood <david.sherwood@arm.com>
14702 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14703 (reduc_xor_scal_optab): New optabs.
14704 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14705 (reduc_xor_scal_@var{m}): Document.
14706 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14707 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14708 internal functions.
14709 * fold-const-call.c (fold_const_call): Handle them.
14710 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14711 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14712 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14713 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14714 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14715 (UNSPEC_XORV): New unspecs.
14716 (optab): Add entries for them.
14717 (BITWISEV): New int iterator.
14718 (bit_reduc_op): New int attributes.
14720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14721 Alan Hayward <alan.hayward@arm.com>
14722 David Sherwood <david.sherwood@arm.com>
14724 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14725 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14726 * optabs.def (vec_shl_insert_optab): New optab.
14727 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14728 (duplicate_and_interleave): Likewise.
14729 * tree-vect-loop.c: Include internal-fn.h.
14730 (neutral_op_for_slp_reduction): New function, split out from
14731 get_initial_defs_for_reduction.
14732 (get_initial_def_for_reduction): Handle option 2 for variable-length
14733 vectors by loading the neutral value into a vector and then shifting
14734 the initial value into element 0.
14735 (get_initial_defs_for_reduction): Replace the code argument with
14736 the neutral value calculated by neutral_op_for_slp_reduction.
14737 Use gimple_build_vector for constant-length vectors.
14738 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14739 but the first group_size elements have a neutral value.
14740 Use duplicate_and_interleave otherwise.
14741 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14742 Update call to get_initial_defs_for_reduction. Handle SLP
14743 reductions for variable-length vectors by creating one vector
14744 result for each scalar result, with the elements associated
14745 with other scalar results stubbed out with the neutral value.
14746 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14747 Require IFN_VEC_SHL_INSERT for double reductions on
14748 variable-length vectors, or SLP reductions that have
14749 a neutral value. Require can_duplicate_and_interleave_p
14750 support for variable-length unchained SLP reductions if there
14751 is no neutral value, such as for MIN/MAX reductions. Also require
14752 the number of vector elements to be a multiple of the number of
14753 SLP statements when doing variable-length unchained SLP reductions.
14754 Update call to vect_create_epilog_for_reduction.
14755 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14756 and remove initial values.
14757 (duplicate_and_interleave): Make public.
14758 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14759 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14762 Alan Hayward <alan.hayward@arm.com>
14763 David Sherwood <david.sherwood@arm.com>
14765 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14766 (can_duplicate_and_interleave_p): New function.
14767 (vect_get_and_check_slp_defs): Take the vector of statements
14768 rather than just the current one. Remove excess parentheses.
14769 Restriction rejectinon of vect_constant_def and vect_external_def
14770 for variable-length vectors to boolean types, or types for which
14771 can_duplicate_and_interleave_p is false.
14772 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14773 (duplicate_and_interleave): New function.
14774 (vect_get_constant_vectors): Use gimple_build_vector for
14775 constant-length vectors and suitable variable-length constant
14776 vectors. Use duplicate_and_interleave for other variable-length
14777 vectors. Don't defer the update when inserting new statements.
14779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14780 Alan Hayward <alan.hayward@arm.com>
14781 David Sherwood <david.sherwood@arm.com>
14783 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14784 min_profitable_iters doesn't go negative.
14786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14787 Alan Hayward <alan.hayward@arm.com>
14788 David Sherwood <david.sherwood@arm.com>
14790 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14791 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14792 * optabs.def (vec_mask_load_lanes_optab): New optab.
14793 (vec_mask_store_lanes_optab): Likewise.
14794 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14795 (MASK_STORE_LANES): Likewise.
14796 * internal-fn.c (mask_load_lanes_direct): New macro.
14797 (mask_store_lanes_direct): Likewise.
14798 (expand_mask_load_optab_fn): Handle masked operations.
14799 (expand_mask_load_lanes_optab_fn): New macro.
14800 (expand_mask_store_optab_fn): Handle masked operations.
14801 (expand_mask_store_lanes_optab_fn): New macro.
14802 (direct_mask_load_lanes_optab_supported_p): Likewise.
14803 (direct_mask_store_lanes_optab_supported_p): Likewise.
14804 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14806 (vect_load_lanes_supported): Likewise.
14807 * tree-vect-data-refs.c (strip_conversion): New function.
14808 (can_group_stmts_p): Likewise.
14809 (vect_analyze_data_ref_accesses): Use it instead of checking
14810 for a pair of assignments.
14811 (vect_store_lanes_supported): Take a masked_p parameter.
14812 (vect_load_lanes_supported): Likewise.
14813 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14814 vect_store_lanes_supported and vect_load_lanes_supported.
14815 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14816 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14817 parameter. Don't allow gaps for masked accesses.
14818 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14819 and vect_load_lanes_supported.
14820 (get_load_store_type): Take a masked_p parameter and update
14821 call to get_group_load_store_type.
14822 (vectorizable_store): Update call to get_load_store_type.
14823 Handle IFN_MASK_STORE_LANES.
14824 (vectorizable_load): Update call to get_load_store_type.
14825 Handle IFN_MASK_LOAD_LANES.
14827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14828 Alan Hayward <alan.hayward@arm.com>
14829 David Sherwood <david.sherwood@arm.com>
14831 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14833 * config/aarch64/aarch64-protos.h
14834 (aarch64_sve_struct_memory_operand_p): Declare.
14835 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14836 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14837 (VPRED, vpred): Handle SVE structure modes.
14838 * config/aarch64/constraints.md (Utx): New constraint.
14839 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14840 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14841 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14842 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14843 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14844 structure modes. Split into pieces after RA.
14845 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14846 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14848 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14849 SVE structure modes.
14850 (aarch64_classify_address): Likewise.
14851 (sizetochar): Move earlier in file.
14852 (aarch64_print_operand): Handle SVE register lists.
14853 (aarch64_array_mode): New function.
14854 (aarch64_sve_struct_memory_operand_p): Likewise.
14855 (TARGET_ARRAY_MODE): Redefine.
14857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14858 Alan Hayward <alan.hayward@arm.com>
14859 David Sherwood <david.sherwood@arm.com>
14861 * target.def (array_mode): New target hook.
14862 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14863 * doc/tm.texi: Regenerate.
14864 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14865 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14866 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14867 targetm.array_mode.
14868 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14872 Alan Hayward <alan.hayward@arm.com>
14873 David Sherwood <david.sherwood@arm.com>
14875 * fold-const.c (fold_binary_loc): Check the argument types
14876 rather than the result type when testing for a vector operation.
14878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14880 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14881 * doc/tm.texi: Regenerate.
14883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14884 Alan Hayward <alan.hayward@arm.com>
14885 David Sherwood <david.sherwood@arm.com>
14887 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14888 (sve): Document new AArch64 extension.
14889 * doc/md.texi (w): Extend the description of the AArch64
14890 constraint to include SVE vectors.
14891 (Upl, Upa): Document new AArch64 predicate constraints.
14892 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14894 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14895 (msve-vector-bits=): New option.
14896 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14897 SVE when these are disabled.
14898 (sve): New extension.
14899 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14900 modes. Adjust their number of units based on aarch64_sve_vg.
14901 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14902 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14903 aarch64_addr_query_type.
14904 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14905 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14906 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14907 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14908 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14909 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14910 (aarch64_simd_imm_zero_p): Delete.
14911 (aarch64_check_zero_based_sve_index_immediate): Declare.
14912 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14913 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14914 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14915 (aarch64_sve_float_mul_immediate_p): Likewise.
14916 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14917 rather than an rtx.
14918 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14919 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14920 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14921 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14922 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14923 (aarch64_regmode_natural_size): Likewise.
14924 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14925 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14927 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14928 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14929 for VG and the SVE predicate registers.
14930 (V_ALIASES): Add a "z"-prefixed alias.
14931 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14932 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14933 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14934 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14935 (REG_CLASS_NAMES): Add entries for them.
14936 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14937 and the predicate registers.
14938 (aarch64_sve_vg): Declare.
14939 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14940 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14941 (REGMODE_NATURAL_SIZE): Define.
14942 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14944 * config/aarch64/aarch64.c: Include cfgrtl.h.
14945 (simd_immediate_info): Add a constructor for series vectors,
14946 and an associated step field.
14947 (aarch64_sve_vg): New variable.
14948 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14949 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14950 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14951 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14952 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14953 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14954 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14955 (aarch64_get_mask_mode): New functions.
14956 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14957 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14958 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14959 predicate modes and predicate registers. Explicitly restrict
14960 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14961 to store a vector mode if it is recognized by
14962 aarch64_classify_vector_mode.
14963 (aarch64_regmode_natural_size): New function.
14964 (aarch64_hard_regno_caller_save_mode): Return the original mode
14966 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14967 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14968 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14969 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14971 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14972 does not overlap dest if the function is frame-related. Handle
14974 (aarch64_split_add_offset): New function.
14975 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14976 them aarch64_add_offset.
14977 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14978 and update call to aarch64_sub_sp.
14979 (aarch64_add_cfa_expression): New function.
14980 (aarch64_expand_prologue): Pass extra temporary registers to the
14981 functions above. Handle the case in which we need to emit new
14982 DW_CFA_expressions for registers that were originally saved
14983 relative to the stack pointer, but now have to be expressed
14984 relative to the frame pointer.
14985 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14987 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14988 IP0 and IP1 values for SVE frames.
14989 (aarch64_expand_vec_series): New function.
14990 (aarch64_expand_sve_widened_duplicate): Likewise.
14991 (aarch64_expand_sve_const_vector): Likewise.
14992 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14993 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14994 into the register, rather than emitting a SET directly.
14995 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14996 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14997 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14998 (offset_9bit_signed_scaled_p): New functions.
14999 (aarch64_replicate_bitmask_imm): New function.
15000 (aarch64_bitmask_imm): Use it.
15001 (aarch64_cannot_force_const_mem): Reject expressions involving
15002 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15003 (aarch64_classify_index): Handle SVE indices, by requiring
15004 a plain register index with a scale that matches the element size.
15005 (aarch64_classify_address): Handle SVE addresses. Assert that
15006 the mode of the address is VOIDmode or an integer mode.
15007 Update call to aarch64_classify_symbol.
15008 (aarch64_classify_symbolic_expression): Update call to
15009 aarch64_classify_symbol.
15010 (aarch64_const_vec_all_in_range_p): New function.
15011 (aarch64_print_vector_float_operand): Likewise.
15012 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15013 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15014 and the FP immediates 1.0 and 0.5.
15015 (aarch64_print_address_internal): Handle SVE addresses.
15016 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15017 (aarch64_regno_regclass): Handle predicate registers.
15018 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15020 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15021 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15022 (aarch64_convert_sve_vector_bits): New function.
15023 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15024 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15025 rather than an rtx.
15026 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15027 Handle SVE vector and predicate modes. Accept VL-based constants
15028 that need only one temporary register, and VL offsets that require
15029 no temporary registers.
15030 (aarch64_conditional_register_usage): Mark the predicate registers
15031 as fixed if SVE isn't available.
15032 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15033 Return true for SVE vector and predicate modes.
15034 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15035 rather than an unsigned int. Handle SVE modes.
15036 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15038 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15040 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15041 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15042 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15043 (aarch64_sve_float_mul_immediate_p): New functions.
15044 (aarch64_sve_valid_immediate): New function.
15045 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15046 Explicitly reject structure modes. Check for INDEX constants.
15047 Handle PTRUE and PFALSE constants.
15048 (aarch64_check_zero_based_sve_index_immediate): New function.
15049 (aarch64_simd_imm_zero_p): Delete.
15050 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15051 vector modes. Accept constants in the range of CNT[BHWD].
15052 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15053 ask for an Advanced SIMD mode.
15054 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15055 (aarch64_simd_vector_alignment): Handle SVE predicates.
15056 (aarch64_vectorize_preferred_vector_alignment): New function.
15057 (aarch64_simd_vector_alignment_reachable): Use it instead of
15059 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15060 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15062 (MAX_VECT_LEN): Delete.
15063 (expand_vec_perm_d): Add a vec_flags field.
15064 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15065 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15066 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15068 (aarch64_evpc_rev): Rename to...
15069 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15070 (aarch64_evpc_rev_global): New function.
15071 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15072 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15074 (aarch64_evpc_sve_tbl): New function.
15075 (aarch64_expand_vec_perm_const_1): Update after rename of
15076 aarch64_evpc_rev. Handle SVE permutes too, trying
15077 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15078 than aarch64_evpc_tbl.
15079 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15080 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15081 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15082 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15083 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15084 (aarch64_expand_sve_vcond): New functions.
15085 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15086 of aarch64_vector_mode_p.
15087 (aarch64_dwarf_poly_indeterminate_value): New function.
15088 (aarch64_compute_pressure_classes): Likewise.
15089 (aarch64_can_change_mode_class): Likewise.
15090 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15091 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15092 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15093 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15094 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15095 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15096 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15097 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15099 (Dn, Dl, Dr): Accept const as well as const_vector.
15100 (Dz): Likewise. Compare against CONST0_RTX.
15101 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15102 of "vector" where appropriate.
15103 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15104 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15105 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15106 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15107 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15108 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15109 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15110 (v_int_equiv): Extend to SVE modes.
15111 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15113 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15114 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15115 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15116 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15117 (SVE_COND_FP_CMP): New int iterators.
15118 (perm_hilo): Handle the new unpack unspecs.
15119 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15121 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15122 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15123 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15124 (aarch64_equality_operator, aarch64_constant_vector_operand)
15125 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15126 (aarch64_sve_nonimmediate_operand): Likewise.
15127 (aarch64_sve_general_operand): Likewise.
15128 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15129 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15130 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15131 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15132 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15133 (aarch64_sve_float_arith_immediate): Likewise.
15134 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15135 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15136 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15137 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15138 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15139 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15140 (aarch64_sve_float_arith_operand): Likewise.
15141 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15142 (aarch64_sve_float_mul_operand): Likewise.
15143 (aarch64_sve_vec_perm_operand): Likewise.
15144 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15145 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15146 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15147 as well as const_vector.
15148 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15149 in file. Use CONST0_RTX and CONSTM1_RTX.
15150 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15151 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15152 Use aarch64_simd_imm_zero.
15153 * config/aarch64/aarch64-sve.md: New file.
15154 * config/aarch64/aarch64.md: Include it.
15155 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15156 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15157 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15158 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15159 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15160 (sve): New attribute.
15161 (enabled): Disable instructions with the sve attribute unless
15163 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15164 aarch64_expand_mov_immediate.
15165 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15166 CNT[BHSD] immediates.
15167 (movti): Split CONST_POLY_INT moves into two halves.
15168 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15169 Split additions that need a temporary here if the destination
15170 is the stack pointer.
15171 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15172 (*add<mode>3_poly_1): New instruction.
15173 (set_clobber_cc): New expander.
15175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15177 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15178 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15179 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15180 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15181 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15182 Change innermode from fixed_mode_size to machine_mode.
15183 (simplify_subreg): Update call accordingly. Handle a constant-sized
15184 subreg of a variable-length CONST_VECTOR.
15186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15187 Alan Hayward <alan.hayward@arm.com>
15188 David Sherwood <david.sherwood@arm.com>
15190 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15191 (add_offset_to_base): New function, split out from...
15192 (create_mem_ref): ...here. When handling a scale other than 1,
15193 check first whether the address is valid without the offset.
15194 Add it into the base if so, leaving the index and scale as-is.
15196 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15199 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15200 fold_for_warn before checking if arg2 is INTEGER_CST.
15202 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15204 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15205 (store_multiple_operation): Delete.
15206 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15207 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15208 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15209 guarded by TARGET_STRING.
15210 (rs6000_output_load_multiple): Delete.
15211 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15212 OPTION_MASK_STRING / TARGET_STRING handling.
15213 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15214 (const rs6000_opt_masks) <"string">: Change mask to 0.
15215 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15216 (MASK_STRING): Delete.
15217 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15219 (load_multiple): Delete.
15226 (store_multiple): Delete.
15233 (movmemsi_8reg): Delete.
15234 (corresponding unnamed define_insn): Delete.
15235 (movmemsi_6reg): Delete.
15236 (corresponding unnamed define_insn): Delete.
15237 (movmemsi_4reg): Delete.
15238 (corresponding unnamed define_insn): Delete.
15239 (movmemsi_2reg): Delete.
15240 (corresponding unnamed define_insn): Delete.
15241 (movmemsi_1reg): Delete.
15242 (corresponding unnamed define_insn): Delete.
15243 * config/rs6000/rs6000.opt (mno-string): New.
15244 (mstring): Replace by deprecation warning stub.
15245 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15247 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15249 * regrename.c (regrename_do_replace): If replacing the same
15250 reg multiple times, try to reuse last created gen_raw_REG.
15253 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15254 main to workaround a bug in GDB.
15256 2018-01-12 Tom de Vries <tom@codesourcery.com>
15259 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15261 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15263 PR rtl-optimization/80481
15264 * ira-color.c (get_cap_member): New function.
15265 (allocnos_conflict_by_live_ranges_p): Use it.
15266 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15267 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15269 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15272 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15273 (*saddl_se_1): Ditto.
15274 (*ssubsi_1): Ditto.
15275 (*ssubl_se_1): Ditto.
15277 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15279 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15280 rather than wi::to_widest for DR_INITs.
15281 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15282 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15283 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15285 (vect_analyze_group_access_1): Note that here.
15287 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15289 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15290 polynomial type sizes.
15292 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15294 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15295 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15296 (gimple_add_tmp_var): Likewise.
15298 2018-01-12 Martin Liska <mliska@suse.cz>
15300 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15301 (gimple_alloc_sizes): Likewise.
15302 (dump_gimple_statistics): Use PRIu64 in printf format.
15303 * gimple.h: Change uint64_t to int.
15305 2018-01-12 Martin Liska <mliska@suse.cz>
15307 * tree-core.h: Use uint64_t instead of int.
15308 * tree.c (tree_node_counts): Likewise.
15309 (tree_node_sizes): Likewise.
15310 (dump_tree_statistics): Use PRIu64 in printf format.
15312 2018-01-12 Martin Liska <mliska@suse.cz>
15314 * Makefile.in: As qsort_chk is implemented in vec.c, add
15315 vec.o to linkage of gencfn-macros.
15316 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15317 passing the info to record_node_allocation_statistics.
15318 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15320 * ggc-common.c (struct ggc_usage): Add operator== and use
15321 it in operator< and compare function.
15322 * mem-stats.h (struct mem_usage): Likewise.
15323 * vec.c (struct vec_usage): Remove operator< and compare
15324 function. Can be simply inherited.
15326 2018-01-12 Martin Jambor <mjambor@suse.cz>
15329 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15330 * tree-ssa-math-opts.c: Include domwalk.h.
15331 (convert_mult_to_fma_1): New function.
15332 (fma_transformation_info): New type.
15333 (fma_deferring_state): Likewise.
15334 (cancel_fma_deferring): New function.
15335 (result_of_phi): Likewise.
15336 (last_fma_candidate_feeds_initial_phi): Likewise.
15337 (convert_mult_to_fma): Added deferring logic, split actual
15338 transformation to convert_mult_to_fma_1.
15339 (math_opts_dom_walker): New type.
15340 (math_opts_dom_walker::after_dom_children): New method, body moved
15341 here from pass_optimize_widening_mul::execute, added deferring logic
15343 (pass_optimize_widening_mul::execute): Moved most of code to
15344 math_opts_dom_walker::after_dom_children.
15345 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15346 * config/i386/i386.c (ix86_option_override_internal): Added
15347 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15349 2018-01-12 Richard Biener <rguenther@suse.de>
15352 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15353 inline instance vars.
15355 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15358 * config/rx/rx.c (rx_is_restricted_memory_address):
15359 Handle SUBREG case.
15361 2018-01-12 Richard Biener <rguenther@suse.de>
15363 PR tree-optimization/80846
15364 * target.def (split_reduction): New target hook.
15365 * targhooks.c (default_split_reduction): New function.
15366 * targhooks.h (default_split_reduction): Declare.
15367 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15368 target requests first reduce vectors by combining low and high
15370 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15371 (get_vectype_for_scalar_type_and_size): Export.
15372 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15373 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15374 * doc/tm.texi: Regenerate.
15375 * config/i386/i386.c (ix86_split_reduction): Implement
15376 TARGET_VECTORIZE_SPLIT_REDUCTION.
15378 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15381 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15382 in PIC mode except for TARGET_VXWORKS_RTP.
15383 * config/sparc/sparc.c: Include cfgrtl.h.
15384 (TARGET_INIT_PIC_REG): Define.
15385 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15386 (sparc_pic_register_p): New predicate.
15387 (sparc_legitimate_address_p): Use it.
15388 (sparc_legitimize_pic_address): Likewise.
15389 (sparc_delegitimize_address): Likewise.
15390 (sparc_mode_dependent_address_p): Likewise.
15391 (gen_load_pcrel_sym): Remove 4th parameter.
15392 (load_got_register): Adjust call to above. Remove obsolete stuff.
15393 (sparc_expand_prologue): Do not call load_got_register here.
15394 (sparc_flat_expand_prologue): Likewise.
15395 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15396 (sparc_use_pseudo_pic_reg): New function.
15397 (sparc_init_pic_reg): Likewise.
15398 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15399 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15401 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15403 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15404 Add item for branch_cost.
15406 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15408 PR rtl-optimization/83565
15409 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15410 not extend the result to a larger mode for rotate operations.
15411 (num_sign_bit_copies1): Likewise.
15413 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15416 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15418 Use values-Xc.o for -pedantic.
15419 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15421 2018-01-12 Martin Liska <mliska@suse.cz>
15424 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15426 (possible_polymorphic_call_targets): Use it.
15427 (ipa_devirt): Likewise.
15429 2018-01-12 Martin Liska <mliska@suse.cz>
15431 * profile-count.h (enum profile_quality): Use 0 as invalid
15432 enum value of profile_quality.
15434 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15436 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15437 -mext-string options.
15439 2018-01-12 Richard Biener <rguenther@suse.de>
15441 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15442 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15443 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15445 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15447 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15449 * configure.ac (--with-long-double-format): Add support for the
15450 configuration option to change the default long double format on
15452 * config.gcc (powerpc*-linux*-*): Likewise.
15453 * configure: Regenerate.
15454 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15455 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15456 used without modification.
15458 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15460 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15461 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15462 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15463 MISC_BUILTIN_SPEC_BARRIER.
15464 (rs6000_init_builtins): Likewise.
15465 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15467 (speculation_barrier): New define_insn.
15468 * doc/extend.texi: Document __builtin_speculation_barrier.
15470 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15473 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15474 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15475 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15477 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15478 integral modes instead of "ss" and "sd".
15479 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15480 vectors with 32-bit and 64-bit elements.
15481 (vecdupssescalarmodesuffix): New mode attribute.
15482 (vec_dup<mode>): Use it.
15484 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15487 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15488 frame if argument is passed on stack.
15490 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15493 * ree.c (combine_reaching_defs): Optimize also
15494 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15495 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15497 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15499 PR middle-end/83189
15500 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15502 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15504 PR middle-end/83718
15505 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15506 after they are computed.
15508 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15510 PR tree-optimization/83695
15511 * gimple-loop-linterchange.cc
15512 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15513 reset cached scev information after interchange.
15514 (pass_linterchange::execute): Remove call to scev_reset_htab.
15516 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15518 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15519 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15520 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15521 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15522 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15523 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15524 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15525 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15526 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15527 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15528 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15529 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15530 (V_lane_reg): Likewise.
15531 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15533 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15534 (vfmal_lane_low<mode>_intrinsic,
15535 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15536 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15537 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15538 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15539 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15540 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15542 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15544 * config/arm/arm-cpus.in (fp16fml): New feature.
15545 (ALL_SIMD): Add fp16fml.
15546 (armv8.2-a): Add fp16fml as an option.
15547 (armv8.3-a): Likewise.
15548 (armv8.4-a): Add fp16fml as part of fp16.
15549 * config/arm/arm.h (TARGET_FP16FML): Define.
15550 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15552 * config/arm/arm-modes.def (V2HF): Define.
15553 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15554 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15555 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15556 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15557 vfmsl_low, vfmsl_high): New set of builtins.
15558 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15559 (vfml_op): New code attribute.
15560 (VFMLHALVES): New int iterator.
15561 (VFML, VFMLSEL): New mode attributes.
15562 (V_reg): Define mapping for V2HF.
15563 (V_hi, V_lo): New mode attributes.
15564 (VF_constraint): Likewise.
15565 (vfml_half, vfml_half_selector): New int attributes.
15566 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15568 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15569 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15571 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15572 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15573 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15574 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15576 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15577 Document new effective target and option set.
15579 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15581 * config/arm/arm-cpus.in (armv8_4): New feature.
15582 (ARMv8_4a): New fgroup.
15583 (armv8.4-a): New arch.
15584 * config/arm/arm-tables.opt: Regenerate.
15585 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15586 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15587 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15588 Add matching rules for -march=armv8.4-a and extensions.
15589 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15591 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15594 * config/rx/rx.md (BW): New mode attribute.
15595 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15597 2018-01-11 Richard Biener <rguenther@suse.de>
15599 PR tree-optimization/83435
15600 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15601 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15602 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15604 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15605 Alan Hayward <alan.hayward@arm.com>
15606 David Sherwood <david.sherwood@arm.com>
15608 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15610 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15611 (aarch64_print_address_internal): Use it to check for a zero offset.
15613 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15614 Alan Hayward <alan.hayward@arm.com>
15615 David Sherwood <david.sherwood@arm.com>
15617 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15618 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15619 Return a poly_int64 rather than a HOST_WIDE_INT.
15620 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15621 rather than a HOST_WIDE_INT.
15622 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15623 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15624 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15625 final_offset from HOST_WIDE_INT to poly_int64.
15626 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15627 to_constant when getting the number of units in an Advanced SIMD
15629 (aarch64_builtin_vectorized_function): Check for a constant number
15631 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15633 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15634 attribute instead of GET_MODE_NUNITS.
15635 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15636 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15637 GET_MODE_SIZE for fixed-size registers.
15638 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15639 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15640 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15641 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15642 (aarch64_print_operand, aarch64_print_address_internal)
15643 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15644 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15645 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15646 Handle polynomial GET_MODE_SIZE.
15647 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15648 wider than SImode without modification.
15649 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15650 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15651 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15652 passing and returning SVE modes.
15653 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15654 rather than GEN_INT.
15655 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15656 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15657 (aarch64_allocate_and_probe_stack_space): Likewise.
15658 (aarch64_layout_frame): Cope with polynomial offsets.
15659 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15660 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15661 polynomial offsets.
15662 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15663 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15664 poly_int64 rather than a HOST_WIDE_INT.
15665 (aarch64_get_separate_components, aarch64_process_components)
15666 (aarch64_expand_prologue, aarch64_expand_epilogue)
15667 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15668 (aarch64_anchor_offset): New function, split out from...
15669 (aarch64_legitimize_address): ...here.
15670 (aarch64_builtin_vectorization_cost): Handle polynomial
15671 TYPE_VECTOR_SUBPARTS.
15672 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15674 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15675 number of elements from the PARALLEL rather than the mode.
15676 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15677 rather than GET_MODE_BITSIZE.
15678 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15679 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15680 (aarch64_expand_vec_perm_const_1): Handle polynomial
15681 d->perm.length () and d->perm elements.
15682 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15683 Apply to_constant to d->perm elements.
15684 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15685 polynomial CONST_VECTOR_NUNITS.
15686 (aarch64_move_pointer): Take amount as a poly_int64 rather
15688 (aarch64_progress_pointer): Avoid temporary variable.
15689 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15690 the mode attribute instead of GET_MODE.
15692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15693 Alan Hayward <alan.hayward@arm.com>
15694 David Sherwood <david.sherwood@arm.com>
15696 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15697 x exists before using it.
15698 (aarch64_add_constant_internal): Rename to...
15699 (aarch64_add_offset_1): ...this. Replace regnum with separate
15700 src and dest rtxes. Handle the case in which they're different,
15701 including when the offset is zero. Replace scratchreg with an rtx.
15702 Use 2 additions if there is no spare register into which we can
15703 move a 16-bit constant.
15704 (aarch64_add_constant): Delete.
15705 (aarch64_add_offset): Replace reg with separate src and dest
15706 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15707 Use aarch64_add_offset_1.
15708 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15709 an rtx rather than an int. Take the delta as a poly_int64
15710 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15711 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15712 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15713 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15714 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15715 and aarch64_add_sp.
15716 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15717 aarch64_add_constant.
15719 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15721 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15722 Use scalar_float_mode.
15724 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15726 * config/aarch64/aarch64-simd.md
15727 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15728 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15729 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15730 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15731 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15732 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15733 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15734 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15735 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15736 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15738 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15741 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15742 targ_options->x_arm_arch_string is non NULL.
15744 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15746 * config/aarch64/aarch64.h
15747 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15749 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15752 * expmed.c (emit_store_flag_force): Swap if const op0
15753 and change VOIDmode to mode of op0.
15755 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15757 PR rtl-optimization/83761
15758 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15759 than bytes to mode_for_size.
15761 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15763 PR middle-end/83189
15764 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15765 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15768 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15770 PR middle-end/83575
15771 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15772 when in layout mode.
15773 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15774 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15777 2018-01-10 Michael Collison <michael.collison@arm.com>
15779 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15780 * config/aarch64/aarch64-option-extension.def: Add
15781 AARCH64_OPT_EXTENSION of 'fp16fml'.
15782 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15783 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15784 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15785 * config/aarch64/constraints.md (Ui7): New constraint.
15786 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15787 (VFMLA_SEL_W): Ditto.
15790 (VFMLA16_LOW): New int iterator.
15791 (VFMLA16_HIGH): Ditto.
15792 (UNSPEC_FMLAL): New unspec.
15793 (UNSPEC_FMLSL): Ditto.
15794 (UNSPEC_FMLAL2): Ditto.
15795 (UNSPEC_FMLSL2): Ditto.
15796 (f16mac): New code attribute.
15797 * config/aarch64/aarch64-simd-builtins.def
15798 (aarch64_fmlal_lowv2sf): Ditto.
15799 (aarch64_fmlsl_lowv2sf): Ditto.
15800 (aarch64_fmlalq_lowv4sf): Ditto.
15801 (aarch64_fmlslq_lowv4sf): Ditto.
15802 (aarch64_fmlal_highv2sf): Ditto.
15803 (aarch64_fmlsl_highv2sf): Ditto.
15804 (aarch64_fmlalq_highv4sf): Ditto.
15805 (aarch64_fmlslq_highv4sf): Ditto.
15806 (aarch64_fmlal_lane_lowv2sf): Ditto.
15807 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15808 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15809 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15810 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15811 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15812 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15813 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15814 (aarch64_fmlal_lane_highv2sf): Ditto.
15815 (aarch64_fmlsl_lane_highv2sf): Ditto.
15816 (aarch64_fmlal_laneq_highv2sf): Ditto.
15817 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15818 (aarch64_fmlalq_lane_highv4sf): Ditto.
15819 (aarch64_fmlsl_lane_highv4sf): Ditto.
15820 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15821 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15822 * config/aarch64/aarch64-simd.md:
15823 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15824 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15825 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15826 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15827 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15828 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15829 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15830 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15831 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15832 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15833 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15834 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15835 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15836 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15837 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15838 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15839 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15840 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15841 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15842 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15843 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15844 (vfmlsl_low_u32): Ditto.
15845 (vfmlalq_low_u32): Ditto.
15846 (vfmlslq_low_u32): Ditto.
15847 (vfmlal_high_u32): Ditto.
15848 (vfmlsl_high_u32): Ditto.
15849 (vfmlalq_high_u32): Ditto.
15850 (vfmlslq_high_u32): Ditto.
15851 (vfmlal_lane_low_u32): Ditto.
15852 (vfmlsl_lane_low_u32): Ditto.
15853 (vfmlal_laneq_low_u32): Ditto.
15854 (vfmlsl_laneq_low_u32): Ditto.
15855 (vfmlalq_lane_low_u32): Ditto.
15856 (vfmlslq_lane_low_u32): Ditto.
15857 (vfmlalq_laneq_low_u32): Ditto.
15858 (vfmlslq_laneq_low_u32): Ditto.
15859 (vfmlal_lane_high_u32): Ditto.
15860 (vfmlsl_lane_high_u32): Ditto.
15861 (vfmlal_laneq_high_u32): Ditto.
15862 (vfmlsl_laneq_high_u32): Ditto.
15863 (vfmlalq_lane_high_u32): Ditto.
15864 (vfmlslq_lane_high_u32): Ditto.
15865 (vfmlalq_laneq_high_u32): Ditto.
15866 (vfmlslq_laneq_high_u32): Ditto.
15867 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15868 (AARCH64_FL_FOR_ARCH8_4): New.
15869 (AARCH64_ISA_F16FML): New ISA flag.
15870 (TARGET_F16FML): New feature flag for fp16fml.
15871 (doc/invoke.texi): Document new fp16fml option.
15873 2018-01-10 Michael Collison <michael.collison@arm.com>
15875 * config/aarch64/aarch64-builtins.c:
15876 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15877 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15878 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15879 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15880 (AARCH64_ISA_SHA3): New ISA flag.
15881 (TARGET_SHA3): New feature flag for sha3.
15882 * config/aarch64/iterators.md (sha512_op): New int attribute.
15883 (CRYPTO_SHA512): New int iterator.
15884 (UNSPEC_SHA512H): New unspec.
15885 (UNSPEC_SHA512H2): Ditto.
15886 (UNSPEC_SHA512SU0): Ditto.
15887 (UNSPEC_SHA512SU1): Ditto.
15888 * config/aarch64/aarch64-simd-builtins.def
15889 (aarch64_crypto_sha512hqv2di): New builtin.
15890 (aarch64_crypto_sha512h2qv2di): Ditto.
15891 (aarch64_crypto_sha512su0qv2di): Ditto.
15892 (aarch64_crypto_sha512su1qv2di): Ditto.
15893 (aarch64_eor3qv8hi): Ditto.
15894 (aarch64_rax1qv2di): Ditto.
15895 (aarch64_xarqv2di): Ditto.
15896 (aarch64_bcaxqv8hi): Ditto.
15897 * config/aarch64/aarch64-simd.md:
15898 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15899 (aarch64_crypto_sha512su0qv2di): Ditto.
15900 (aarch64_crypto_sha512su1qv2di): Ditto.
15901 (aarch64_eor3qv8hi): Ditto.
15902 (aarch64_rax1qv2di): Ditto.
15903 (aarch64_xarqv2di): Ditto.
15904 (aarch64_bcaxqv8hi): Ditto.
15905 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15906 (vsha512h2q_u64): Ditto.
15907 (vsha512su0q_u64): Ditto.
15908 (vsha512su1q_u64): Ditto.
15909 (veor3q_u16): Ditto.
15910 (vrax1q_u64): Ditto.
15911 (vxarq_u64): Ditto.
15912 (vbcaxq_u16): Ditto.
15913 * config/arm/types.md (crypto_sha512): New type attribute.
15914 (crypto_sha3): Ditto.
15915 (doc/invoke.texi): Document new sha3 option.
15917 2018-01-10 Michael Collison <michael.collison@arm.com>
15919 * config/aarch64/aarch64-builtins.c:
15920 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15921 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15922 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15923 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15924 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15925 (AARCH64_ISA_SM4): New ISA flag.
15926 (TARGET_SM4): New feature flag for sm4.
15927 * config/aarch64/aarch64-simd-builtins.def
15928 (aarch64_sm3ss1qv4si): Ditto.
15929 (aarch64_sm3tt1aq4si): Ditto.
15930 (aarch64_sm3tt1bq4si): Ditto.
15931 (aarch64_sm3tt2aq4si): Ditto.
15932 (aarch64_sm3tt2bq4si): Ditto.
15933 (aarch64_sm3partw1qv4si): Ditto.
15934 (aarch64_sm3partw2qv4si): Ditto.
15935 (aarch64_sm4eqv4si): Ditto.
15936 (aarch64_sm4ekeyqv4si): Ditto.
15937 * config/aarch64/aarch64-simd.md:
15938 (aarch64_sm3ss1qv4si): Ditto.
15939 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15940 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15941 (aarch64_sm4eqv4si): Ditto.
15942 (aarch64_sm4ekeyqv4si): Ditto.
15943 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15944 (sm3part_op): Ditto.
15945 (CRYPTO_SM3TT): Ditto.
15946 (CRYPTO_SM3PART): Ditto.
15947 (UNSPEC_SM3SS1): New unspec.
15948 (UNSPEC_SM3TT1A): Ditto.
15949 (UNSPEC_SM3TT1B): Ditto.
15950 (UNSPEC_SM3TT2A): Ditto.
15951 (UNSPEC_SM3TT2B): Ditto.
15952 (UNSPEC_SM3PARTW1): Ditto.
15953 (UNSPEC_SM3PARTW2): Ditto.
15954 (UNSPEC_SM4E): Ditto.
15955 (UNSPEC_SM4EKEY): Ditto.
15956 * config/aarch64/constraints.md (Ui2): New constraint.
15957 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15958 * config/arm/types.md (crypto_sm3): New type attribute.
15959 (crypto_sm4): Ditto.
15960 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15961 (vsm3tt1aq_u32): Ditto.
15962 (vsm3tt1bq_u32): Ditto.
15963 (vsm3tt2aq_u32): Ditto.
15964 (vsm3tt2bq_u32): Ditto.
15965 (vsm3partw1q_u32): Ditto.
15966 (vsm3partw2q_u32): Ditto.
15967 (vsm4eq_u32): Ditto.
15968 (vsm4ekeyq_u32): Ditto.
15969 (doc/invoke.texi): Document new sm4 option.
15971 2018-01-10 Michael Collison <michael.collison@arm.com>
15973 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15974 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15975 (AARCH64_FL_FOR_ARCH8_4): New.
15976 (AARCH64_FL_V8_4): New flag.
15977 (doc/invoke.texi): Document new armv8.4-a option.
15979 2018-01-10 Michael Collison <michael.collison@arm.com>
15981 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15982 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15983 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15984 * config/aarch64/aarch64-option-extension.def: Add
15985 AARCH64_OPT_EXTENSION of 'sha2'.
15986 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15987 (crypto): Disable sha2 and aes if crypto disabled.
15988 (crypto): Enable aes and sha2 if enabled.
15989 (simd): Disable sha2 and aes if simd disabled.
15990 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15992 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15993 (TARGET_SHA2): New feature flag for sha2.
15994 (TARGET_AES): New feature flag for aes.
15995 * config/aarch64/aarch64-simd.md:
15996 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15997 conditional on TARGET_AES.
15998 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15999 (aarch64_crypto_sha1hsi): Make pattern conditional
16001 (aarch64_crypto_sha1hv4si): Ditto.
16002 (aarch64_be_crypto_sha1hv4si): Ditto.
16003 (aarch64_crypto_sha1su1v4si): Ditto.
16004 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16005 (aarch64_crypto_sha1su0v4si): Ditto.
16006 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16007 (aarch64_crypto_sha256su0v4si): Ditto.
16008 (aarch64_crypto_sha256su1v4si): Ditto.
16009 (doc/invoke.texi): Document new aes and sha2 options.
16011 2018-01-10 Martin Sebor <msebor@redhat.com>
16013 PR tree-optimization/83781
16014 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16017 2018-01-11 Martin Sebor <msebor@gmail.com>
16018 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16020 PR tree-optimization/83501
16021 PR tree-optimization/81703
16023 * tree-ssa-strlen.c (get_string_cst): Rename...
16024 (get_string_len): ...to this. Handle global constants.
16025 (handle_char_store): Adjust.
16027 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16028 Jim Wilson <jimw@sifive.com>
16030 * config/riscv/riscv-protos.h (riscv_output_return): New.
16031 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16032 (riscv_attribute_table, riscv_output_return),
16033 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16034 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16035 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16037 (riscv_expand_prologue): Add early return for naked function.
16038 (riscv_expand_epilogue): Likewise.
16039 (riscv_function_ok_for_sibcall): Return false for naked function.
16040 (riscv_set_current_function): New.
16041 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16042 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16043 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16044 * doc/extend.texi (RISC-V Function Attributes): New.
16046 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16048 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16049 check for 128-bit long double before checking TCmode.
16050 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16051 128-bit long doubles before checking TFmode or TCmode.
16052 (FLOAT128_IBM_P): Likewise.
16054 2018-01-10 Martin Sebor <msebor@redhat.com>
16056 PR tree-optimization/83671
16057 * builtins.c (c_strlen): Unconditionally return zero for the empty
16059 Use -Warray-bounds for warnings.
16060 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16061 for non-constant array indices with COMPONENT_REF, arrays of
16062 arrays, and pointers to arrays.
16063 (gimple_fold_builtin_strlen): Determine and set length range for
16064 non-constant character arrays.
16066 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16068 PR middle-end/81897
16069 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16072 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16074 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16076 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16079 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16080 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16081 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16082 indexed_or_indirect_operand predicate.
16083 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16084 (*vsx_le_perm_load_v8hi): Likewise.
16085 (*vsx_le_perm_load_v16qi): Likewise.
16086 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16087 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16088 (*vsx_le_perm_store_v8hi): Likewise.
16089 (*vsx_le_perm_store_v16qi): Likewise.
16090 (eight unnamed splitters): Likewise.
16092 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16094 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16095 * config/rs6000/emmintrin.h: Likewise.
16096 * config/rs6000/mmintrin.h: Likewise.
16097 * config/rs6000/xmmintrin.h: Likewise.
16099 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16102 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16104 * tree.c (tree_nop_conversion): Return true for location wrapper
16106 (maybe_wrap_with_location): New function.
16107 (selftest::check_strip_nops): New function.
16108 (selftest::test_location_wrappers): New function.
16109 (selftest::tree_c_tests): Call it.
16110 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16111 (maybe_wrap_with_location): New decl.
16112 (EXPR_LOCATION_WRAPPER_P): New macro.
16113 (location_wrapper_p): New inline function.
16114 (tree_strip_any_location_wrapper): New inline function.
16116 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16119 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16120 stack_realign_offset for the largest alignment of stack slot
16122 (ix86_find_max_used_stack_alignment): New function.
16123 (ix86_finalize_stack_frame_flags): Use it. Set
16124 max_used_stack_alignment if we don't realign stack.
16125 * config/i386/i386.h (machine_function): Add
16126 max_used_stack_alignment.
16128 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16130 * config/arm/arm.opt (-mbranch-cost): New option.
16131 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16134 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16137 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16138 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16140 2018-01-10 Richard Biener <rguenther@suse.de>
16143 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16144 early out so it also covers the case where we have a non-NULL
16147 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16149 PR tree-optimization/83753
16150 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16151 for non-strided grouped accesses if the number of elements is 1.
16153 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16156 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16157 * i386.h (TARGET_USE_GATHER): Define.
16158 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16160 2018-01-10 Martin Liska <mliska@suse.cz>
16163 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16164 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16166 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16167 CLEANUP_NO_PARTITIONING is not set.
16169 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16171 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16172 for vectors, as a partial revert of r254296.
16173 * rtl.h (const_vec_p): Delete.
16174 (const_vec_duplicate_p): Don't test for vector CONSTs.
16175 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16176 * expmed.c (make_tree): Likewise.
16179 * common.md (E, F): Use CONSTANT_P instead of checking for
16181 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16182 checking for CONST_VECTOR.
16184 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16186 PR middle-end/83575
16187 * predict.c (force_edge_cold): Handle in more sane way edges
16188 with no prediction.
16190 2018-01-09 Carl Love <cel@us.ibm.com>
16192 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16194 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16195 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16196 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16197 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16198 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16199 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16200 * config/rs6000/rs6000-protos.h: Add extern defition for
16201 rs6000_generate_float2_double_code.
16202 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16204 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16205 (float2_v2df): Add define_expand.
16207 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16210 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16211 op_mode in the force_to_mode call.
16213 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16215 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16216 instead of checking each element individually.
16217 (aarch64_evpc_uzp): Likewise.
16218 (aarch64_evpc_zip): Likewise.
16219 (aarch64_evpc_ext): Likewise.
16220 (aarch64_evpc_rev): Likewise.
16221 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16222 instead of checking each element individually. Return true without
16224 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16225 whether all selected elements come from the same input, instead of
16226 checking each element individually. Remove calls to gen_rtx_REG,
16227 start_sequence and end_sequence and instead assert that no rtl is
16230 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16232 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16233 order of HIGH and CONST checks.
16235 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16237 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16238 if the destination isn't an SSA_NAME.
16240 2018-01-09 Richard Biener <rguenther@suse.de>
16242 PR tree-optimization/83668
16243 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16245 (canonicalize_loop_form): ... here, renamed from ...
16246 (canonicalize_loop_closed_ssa_form): ... this and amended to
16247 swap successor edges for loop exit blocks to make us use
16248 the RPO order we need for initial schedule generation.
16250 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16252 PR tree-optimization/64811
16253 * match.pd: When optimizing comparisons with Inf, avoid
16254 introducing or losing exceptions from comparisons with NaN.
16256 2018-01-09 Martin Liska <mliska@suse.cz>
16259 * asan.c (shadow_mem_size): Add gcc_assert.
16261 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16263 Don't save registers in main().
16266 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16267 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16268 * config/avr/avr.c (avr_set_current_function): Don't error if
16269 naked, OS_task or OS_main are specified at the same time.
16270 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16272 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16274 * common/config/avr/avr-common.c (avr_option_optimization_table):
16275 Switch on -mmain-is-OS_task for optimizing compilations.
16277 2018-01-09 Richard Biener <rguenther@suse.de>
16279 PR tree-optimization/83572
16280 * graphite.c: Include cfganal.h.
16281 (graphite_transform_loops): Connect infinite loops to exit
16282 and remove fake edges at the end.
16284 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16286 * ipa-inline.c (edge_badness): Revert accidental checkin.
16288 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16291 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16292 symbols; not inline clones.
16294 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16297 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16298 hard registers. Formatting fixes.
16300 PR preprocessor/83722
16301 * gcc.c (try_generate_repro): Pass
16302 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16303 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16306 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16307 Kito Cheng <kito.cheng@gmail.com>
16309 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16310 (riscv_leaf_function_p): Delete.
16311 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16313 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16315 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16317 (do_ifelse): New function.
16318 (do_isel): New function.
16319 (do_sub3): New function.
16320 (do_add3): New function.
16321 (do_load_mask_compare): New function.
16322 (do_overlap_load_compare): New function.
16323 (expand_compare_loop): New function.
16324 (expand_block_compare): Call expand_compare_loop() when appropriate.
16325 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16326 option description.
16327 (-mblock-compare-inline-loop-limit): New option.
16329 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16332 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16333 Reverse order of second and third operands in first alternative.
16334 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16335 of first and second elements in UNSPEC_VPERMR vector.
16336 (altivec_expand_vec_perm_le): Likewise.
16338 2018-01-08 Jeff Law <law@redhat.com>
16340 PR rtl-optimizatin/81308
16341 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16342 (process_switch): If group_case_labels makes a change, then set
16344 (pass_convert_switch::execute): If a switch is converted, then
16345 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16347 PR rtl-optimization/81308
16348 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16351 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16353 PR target/83663 - Revert r255946
16354 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16355 generation for cases where splatting a value is not useful.
16356 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16357 across a vec_duplicate and a paradoxical subreg forming a vector
16358 mode to a vec_concat.
16360 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16362 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16363 -march=armv8.3-a variants.
16364 * config/arm/t-multilib: Likewise.
16365 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16367 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16369 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16371 (cceq_ior_compare_complement): Give it a name so I can use it, and
16372 change boolean_or_operator predicate to boolean_operator so it can
16373 be used to generate a crand.
16374 (eqne): New code iterator.
16375 (bd/bd_neg): New code_attrs.
16376 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16377 a single define_insn.
16378 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16379 decrement (bdnzt/bdnzf/bdzt/bdzf).
16380 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16381 with the new names of the branch decrement patterns, and added the
16382 names of the branch decrement conditional patterns.
16384 2018-01-08 Richard Biener <rguenther@suse.de>
16386 PR tree-optimization/83563
16387 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16390 2018-01-08 Richard Biener <rguenther@suse.de>
16392 PR middle-end/83713
16393 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16395 2018-01-08 Richard Biener <rguenther@suse.de>
16397 PR tree-optimization/83685
16398 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16399 references to abnormals.
16401 2018-01-08 Richard Biener <rguenther@suse.de>
16404 * dwarf2out.c (output_indirect_strings): Handle empty
16405 skeleton_debug_str_hash.
16406 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16408 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16410 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16411 (emit_store_direct): Likewise.
16412 (arc_trampoline_adjust_address): Likewise.
16413 (arc_asm_trampoline_template): New function.
16414 (arc_initialize_trampoline): Use asm_trampoline_template.
16415 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16416 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16417 * config/arc/arc.md (flush_icache): Delete pattern.
16419 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16421 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16422 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16425 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16428 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16429 by not USED_FOR_TARGET.
16430 (make_pass_resolve_sw_modes): Likewise.
16432 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16434 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16437 2018-01-08 Richard Biener <rguenther@suse.de>
16439 PR middle-end/83580
16440 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16442 2018-01-08 Richard Biener <rguenther@suse.de>
16444 PR middle-end/83517
16445 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16447 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16449 PR middle-end/81897
16450 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16451 basic blocks with a small number of successors.
16452 (convert_control_dep_chain_into_preds): Improve handling of
16454 (dump_predicates): Split apart into...
16455 (dump_pred_chain): ...here...
16456 (dump_pred_info): ...and here.
16457 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16458 (can_chain_union_be_invalidated_p): Improve check for invalidation
16460 (uninit_uses_cannot_happen): Avoid unnecessary if
16461 convert_control_dep_chain_into_preds yielded nothing.
16463 2018-01-06 Martin Sebor <msebor@redhat.com>
16465 PR tree-optimization/83640
16466 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16467 subtracting negative offset from size.
16468 (builtin_access::overlap): Adjust offset bounds of the access to fall
16469 within the size of the object if possible.
16471 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16473 PR rtl-optimization/83699
16474 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16475 extract_bit_field_as_subreg to cases in which the extracted
16476 value is also a vector.
16478 * lra-constraints.c (process_alt_operands): Test for the equivalence
16479 substitutions when detecting a possible reload cycle.
16481 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16484 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16485 by default if flag_selective_schedling{,2}. Formatting fixes.
16487 PR rtl-optimization/83682
16488 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16489 if it has non-VECTOR_MODE element mode.
16490 (vec_duplicate_p): Likewise.
16492 PR middle-end/83694
16493 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16494 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16496 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16499 * config/i386/i386-builtin.def
16500 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16501 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16502 Require also OPTION_MASK_ISA_AVX512F in addition to
16503 OPTION_MASK_ISA_GFNI.
16504 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16505 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16506 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16507 to OPTION_MASK_ISA_GFNI.
16508 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16509 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16510 OPTION_MASK_ISA_AVX512BW.
16511 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16512 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16513 addition to OPTION_MASK_ISA_GFNI.
16514 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16515 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16516 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16517 to OPTION_MASK_ISA_GFNI.
16518 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16519 a requirement for all ISAs rather than any of them with a few
16521 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16523 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16524 bitmasks to be enabled with 3 exceptions, instead of requiring any
16525 enabled ISA with lots of exceptions.
16526 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16527 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16528 Change avx512bw in isa attribute to avx512f.
16529 * config/i386/sgxintrin.h: Add license boilerplate.
16530 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16531 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16532 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16533 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16535 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16536 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16537 temporarily sse2 rather than sse if not enabled already.
16540 * config/i386/sse.md (VI248_VLBW): Rename to ...
16541 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16542 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16543 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16544 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16545 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16546 mode iterator instead of VI248_VLBW.
16548 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16550 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16551 (record_modified): Skip clobbers; add debug output.
16552 (param_change_prob): Use sreal frequencies.
16554 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16556 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16557 punt for user-aligned variables.
16559 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16561 * tree-chrec.c (chrec_contains_symbols): Return true for
16564 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16567 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16568 of (x|y) == x for BICS pattern.
16570 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16572 PR tree-optimization/83605
16573 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16574 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16577 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16579 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16580 * config/epiphany/rtems.h: New file.
16582 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16583 Uros Bizjak <ubizjak@gmail.com>
16586 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16587 QIreg_operand instead of register_operand predicate.
16588 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16589 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16590 comments instead of -fmitigate[-_]rop.
16592 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16595 * cgraphunit.c (symbol_table::compile): Switch to text_section
16596 before calling assembly_start debug hook.
16597 * run-rtl-passes.c (run_rtl_passes): Likewise.
16600 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16602 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16603 range_int_cst_p rather than !symbolic_range_p before calling
16604 extract_range_from_multiplicative_op_1.
16606 2018-01-04 Jeff Law <law@redhat.com>
16608 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16609 redundant test in assertion.
16611 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16613 * doc/rtl.texi: Document machine_mode wrapper classes.
16615 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16617 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16618 using tree_to_uhwi.
16620 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16622 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16623 the VEC_PERM_EXPR fold to fail.
16625 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16628 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16629 to switched_sections.
16631 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16634 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16635 test for d.testing.
16637 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16640 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16641 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16643 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16646 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16647 is BLKmode and bitpos not zero or mode change is needed.
16649 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16652 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16655 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16658 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16659 instead of MULT rtx. Update all corresponding splitters.
16660 (*saddl_se): Ditto.
16661 (*ssub<modesuffix>): Ditto.
16662 (*ssubl_se): Ditto.
16663 (*cmp_sadd_di): Update split patterns.
16664 (*cmp_sadd_si): Ditto.
16665 (*cmp_sadd_sidi): Ditto.
16666 (*cmp_ssub_di): Ditto.
16667 (*cmp_ssub_si): Ditto.
16668 (*cmp_ssub_sidi): Ditto.
16669 * config/alpha/predicates.md (const23_operand): New predicate.
16670 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16671 Look for ASHIFT, not MULT inner operand.
16672 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16674 2018-01-04 Martin Liska <mliska@suse.cz>
16676 PR gcov-profile/83669
16677 * gcov.c (output_intermediate_file): Add version to intermediate
16679 * doc/gcov.texi: Document new field 'version' in intermediate
16680 file format. Fix location of '-k' option of gcov command.
16682 2018-01-04 Martin Liska <mliska@suse.cz>
16685 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16687 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16689 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16691 2018-01-03 Martin Sebor <msebor@redhat.com>
16693 PR tree-optimization/83655
16694 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16695 checking calls with invalid arguments.
16697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16699 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16700 (vectorizable_mask_load_store): Delete.
16701 (vectorizable_call): Return false for masked loads and stores.
16702 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16703 instead of gimple_assign_rhs1.
16704 (vectorizable_load): Handle IFN_MASK_LOAD.
16705 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16709 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16711 (vectorizable_mask_load_store): ...here.
16712 (vectorizable_load): ...and here.
16714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16716 * tree-vect-stmts.c (vect_build_all_ones_mask)
16717 (vect_build_zero_merge_argument): New functions, split out from...
16718 (vectorizable_load): ...here.
16720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16722 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16724 (vectorizable_mask_load_store): ...here.
16725 (vectorizable_store): ...and here.
16727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16729 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16731 (vectorizable_mask_load_store): ...here.
16733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16735 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16736 (vect_model_store_cost): Take a vec_load_store_type instead of a
16738 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16739 (vect_model_store_cost): Take a vec_load_store_type instead of a
16741 (vectorizable_mask_load_store): Update accordingly.
16742 (vectorizable_store): Likewise.
16743 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16747 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16748 IFN_MASK_LOAD calls here rather than...
16749 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16752 Alan Hayward <alan.hayward@arm.com>
16753 David Sherwood <david.sherwood@arm.com>
16755 * expmed.c (extract_bit_field_1): For vector extracts,
16756 fall back to extract_bit_field_as_subreg if vec_extract
16759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16760 Alan Hayward <alan.hayward@arm.com>
16761 David Sherwood <david.sherwood@arm.com>
16763 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16764 they are variable or constant sized.
16765 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16766 slots for constant-sized data.
16768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16769 Alan Hayward <alan.hayward@arm.com>
16770 David Sherwood <david.sherwood@arm.com>
16772 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16773 handling COND_EXPRs with boolean comparisons, try to find a better
16774 basis for the mask type than the boolean itself.
16776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16778 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16779 is calculated and how it can be overridden.
16780 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16781 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16783 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16787 Alan Hayward <alan.hayward@arm.com>
16788 David Sherwood <david.sherwood@arm.com>
16790 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16791 Remove the mode argument.
16792 (aarch64_simd_valid_immediate): Remove the mode and inverse
16794 * config/aarch64/iterators.md (bitsize): New iterator.
16795 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16796 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16797 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16798 aarch64_simd_valid_immediate.
16799 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16800 (aarch64_reg_or_bic_imm): Likewise.
16801 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16802 with an insn_type enum and msl with a modifier_type enum.
16803 Replace element_width with a scalar_mode. Change the shift
16804 to unsigned int. Add constructors for scalar_float_mode and
16805 scalar_int_mode elements.
16806 (aarch64_vect_float_const_representable_p): Delete.
16807 (aarch64_can_const_movi_rtx_p)
16808 (aarch64_simd_scalar_immediate_valid_for_move)
16809 (aarch64_simd_make_constant): Update call to
16810 aarch64_simd_valid_immediate.
16811 (aarch64_advsimd_valid_immediate_hs): New function.
16812 (aarch64_advsimd_valid_immediate): Likewise.
16813 (aarch64_simd_valid_immediate): Remove mode and inverse
16814 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16815 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16816 and aarch64_float_const_representable_p on the result.
16817 (aarch64_output_simd_mov_immediate): Remove mode argument.
16818 Update call to aarch64_simd_valid_immediate and use of
16819 simd_immediate_info.
16820 (aarch64_output_scalar_simd_mov_immediate): Update call
16823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16824 Alan Hayward <alan.hayward@arm.com>
16825 David Sherwood <david.sherwood@arm.com>
16827 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16828 (mode_nunits): Likewise CONST_MODE_NUNITS.
16829 * machmode.def (ADJUST_NUNITS): Document.
16830 * genmodes.c (mode_data::need_nunits_adj): New field.
16831 (blank_mode): Update accordingly.
16832 (adj_nunits): New variable.
16833 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16835 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16836 listed in adj_nunits.
16837 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16838 listed in adj_nunits. Don't emit case statements for such modes.
16839 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16840 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16841 nothing if adj_nunits is nonnull.
16842 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16843 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16844 (emit_mode_fbit): Update use of print_maybe_const_decl.
16845 (emit_move_size): Likewise. Treat the array as non-const
16847 (emit_mode_adjustments): Handle adj_nunits.
16849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16851 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16852 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16853 (VECTOR_MODES): Use it.
16854 (make_vector_modes): Take the prefix as an argument.
16856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16857 Alan Hayward <alan.hayward@arm.com>
16858 David Sherwood <david.sherwood@arm.com>
16860 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16861 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16862 for MODE_VECTOR_BOOL.
16863 * machmode.def (VECTOR_BOOL_MODE): Document.
16864 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16865 (make_vector_bool_mode): New function.
16866 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16868 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16869 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16871 * stor-layout.c (int_mode_for_mode): Likewise.
16872 * tree.c (build_vector_type_for_mode): Likewise.
16873 * varasm.c (output_constant_pool_2): Likewise.
16874 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16875 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16876 for MODE_VECTOR_BOOL.
16877 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16878 of mode class checks.
16879 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16880 instead of a list of mode class checks.
16881 (expand_vector_scalar_condition): Likewise.
16882 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16885 Alan Hayward <alan.hayward@arm.com>
16886 David Sherwood <david.sherwood@arm.com>
16888 * machmode.h (mode_size): Change from unsigned short to
16890 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16891 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16892 or if measurement_type is not polynomial.
16893 (fixed_size_mode::includes_p): Check for constant-sized modes.
16894 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16895 return a poly_uint16 rather than an unsigned short.
16896 (emit_mode_size): Change the type of mode_size from unsigned short
16897 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16898 (emit_mode_adjustments): Cope with polynomial vector sizes.
16899 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16901 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16903 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16904 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16905 * caller-save.c (setup_save_areas): Likewise.
16906 (replace_reg_with_saved_mem): Likewise.
16907 * calls.c (emit_library_call_value_1): Likewise.
16908 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16909 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16910 (gen_lowpart_for_combine): Likewise.
16911 * convert.c (convert_to_integer_1): Likewise.
16912 * cse.c (equiv_constant, cse_insn): Likewise.
16913 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16914 (cselib_subst_to_values): Likewise.
16915 * dce.c (word_dce_process_block): Likewise.
16916 * df-problems.c (df_word_lr_mark_ref): Likewise.
16917 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16918 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16919 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16920 (rtl_for_decl_location): Likewise.
16921 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16922 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16923 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16924 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16925 (expand_expr_real_1): Likewise.
16926 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16927 (pad_below): Likewise.
16928 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16929 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16930 * ira.c (get_subreg_tracking_sizes): Likewise.
16931 * ira-build.c (ira_create_allocno_objects): Likewise.
16932 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16933 (ira_sort_regnos_for_alter_reg): Likewise.
16934 * ira-costs.c (record_operand_costs): Likewise.
16935 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16936 (resolve_simple_move): Likewise.
16937 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16938 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16939 (lra_constraints): Likewise.
16940 (CONST_POOL_OK_P): Reject variable-sized modes.
16941 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16942 (add_pseudo_to_slot, lra_spill): Likewise.
16943 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16944 * optabs-query.c (get_best_extraction_insn): Likewise.
16945 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16946 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16947 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16948 * recog.c (offsettable_address_addr_space_p): Likewise.
16949 * regcprop.c (maybe_mode_change): Likewise.
16950 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16951 * regrename.c (build_def_use): Likewise.
16952 * regstat.c (dump_reg_info): Likewise.
16953 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16954 (find_reloads, find_reloads_subreg_address): Likewise.
16955 * reload1.c (eliminate_regs_1): Likewise.
16956 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16957 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16958 (simplify_binary_operation_1, simplify_subreg): Likewise.
16959 * targhooks.c (default_function_arg_padding): Likewise.
16960 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16961 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16962 (verify_gimple_assign_ternary): Likewise.
16963 * tree-inline.c (estimate_move_cost): Likewise.
16964 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16965 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16966 (get_address_cost_ainc): Likewise.
16967 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16968 (vect_supportable_dr_alignment): Likewise.
16969 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16970 (vectorizable_reduction): Likewise.
16971 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16972 (vectorizable_operation, vectorizable_load): Likewise.
16973 * tree.c (build_same_sized_truth_vector_type): Likewise.
16974 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16975 * var-tracking.c (emit_note_insn_var_location): Likewise.
16976 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16977 (ADDR_VEC_ALIGN): Likewise.
16979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16980 Alan Hayward <alan.hayward@arm.com>
16981 David Sherwood <david.sherwood@arm.com>
16983 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16985 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16986 or if measurement_type is polynomial.
16987 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16988 * combine.c (make_extraction): Likewise.
16989 * dse.c (find_shift_sequence): Likewise.
16990 * dwarf2out.c (mem_loc_descriptor): Likewise.
16991 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16992 (extract_bit_field, extract_low_bits): Likewise.
16993 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16994 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16995 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16996 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16997 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16998 * reload.c (find_reloads): Likewise.
16999 * reload1.c (alter_reg): Likewise.
17000 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17001 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17002 * tree-if-conv.c (predicate_mem_writes): Likewise.
17003 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17004 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17005 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17006 * valtrack.c (dead_debug_insert_temp): Likewise.
17007 * varasm.c (mergeable_constant_section): Likewise.
17008 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17011 Alan Hayward <alan.hayward@arm.com>
17012 David Sherwood <david.sherwood@arm.com>
17014 * expr.c (expand_assignment): Cope with polynomial mode sizes
17015 when assigning to a CONCAT.
17017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17018 Alan Hayward <alan.hayward@arm.com>
17019 David Sherwood <david.sherwood@arm.com>
17021 * machmode.h (mode_precision): Change from unsigned short to
17023 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17025 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17026 or if measurement_type is not polynomial.
17027 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17028 in which the mode is already known to be a scalar_int_mode.
17029 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17030 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17032 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17033 for GET_MODE_PRECISION.
17034 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17035 for GET_MODE_PRECISION.
17036 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17038 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17039 (expand_field_assignment, make_extraction): Likewise.
17040 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17041 (get_last_value): Likewise.
17042 * convert.c (convert_to_integer_1): Likewise.
17043 * cse.c (cse_insn): Likewise.
17044 * expr.c (expand_expr_real_1): Likewise.
17045 * lra-constraints.c (simplify_operand_subreg): Likewise.
17046 * optabs-query.c (can_atomic_load_p): Likewise.
17047 * optabs.c (expand_atomic_load): Likewise.
17048 (expand_atomic_store): Likewise.
17049 * ree.c (combine_reaching_defs): Likewise.
17050 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17051 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17052 * tree.h (type_has_mode_precision_p): Likewise.
17053 * ubsan.c (instrument_si_overflow): Likewise.
17055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17056 Alan Hayward <alan.hayward@arm.com>
17057 David Sherwood <david.sherwood@arm.com>
17059 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17060 polynomial numbers of units.
17061 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17062 (valid_vector_subparts_p): New function.
17063 (build_vector_type): Remove temporary shim and take the number
17064 of units as a poly_uint64 rather than an int.
17065 (build_opaque_vector_type): Take the number of units as a
17066 poly_uint64 rather than an int.
17067 * tree.c (build_vector_from_ctor): Handle polynomial
17068 TYPE_VECTOR_SUBPARTS.
17069 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17070 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17071 (build_vector_from_val): If the number of units is variable,
17072 use build_vec_duplicate_cst for constant operands and
17073 VEC_DUPLICATE_EXPR otherwise.
17074 (make_vector_type): Remove temporary is_constant ().
17075 (build_vector_type, build_opaque_vector_type): Take the number of
17076 units as a poly_uint64 rather than an int.
17077 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17079 * cfgexpand.c (expand_debug_expr): Likewise.
17080 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17081 (store_constructor, expand_expr_real_1): Likewise.
17082 (const_scalar_mask_from_tree): Likewise.
17083 * fold-const-call.c (fold_const_reduction): Likewise.
17084 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17085 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17086 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17087 (fold_relational_const): Likewise.
17088 (native_interpret_vector): Likewise. Change the size from an
17089 int to an unsigned int.
17090 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17091 TYPE_VECTOR_SUBPARTS.
17092 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17093 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17094 duplicating a non-constant operand into a variable-length vector.
17095 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17096 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17097 * ipa-icf.c (sem_variable::equals): Likewise.
17098 * match.pd: Likewise.
17099 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17100 * print-tree.c (print_node): Likewise.
17101 * stor-layout.c (layout_type): Likewise.
17102 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17103 * tree-cfg.c (verify_gimple_comparison): Likewise.
17104 (verify_gimple_assign_binary): Likewise.
17105 (verify_gimple_assign_ternary): Likewise.
17106 (verify_gimple_assign_single): Likewise.
17107 * tree-pretty-print.c (dump_generic_node): Likewise.
17108 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17109 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17110 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17111 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17112 (vect_shift_permute_load_chain): Likewise.
17113 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17114 (expand_vector_condition, optimize_vector_constructor): Likewise.
17115 (lower_vec_perm, get_compute_type): Likewise.
17116 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17117 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17118 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17119 (vect_recog_mask_conversion_pattern): Likewise.
17120 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17121 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17122 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17123 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17124 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17125 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17126 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17127 (supportable_widening_operation): Likewise.
17128 (supportable_narrowing_operation): Likewise.
17129 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17131 * varasm.c (output_constant): Likewise.
17133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17134 Alan Hayward <alan.hayward@arm.com>
17135 David Sherwood <david.sherwood@arm.com>
17137 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17138 so that both the length == 3 and length != 3 cases set up their
17139 own permute vectors. Add comments explaining why we know the
17140 number of elements is constant.
17141 (vect_permute_load_chain): Likewise.
17143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17144 Alan Hayward <alan.hayward@arm.com>
17145 David Sherwood <david.sherwood@arm.com>
17147 * machmode.h (mode_nunits): Change from unsigned char to
17149 (ONLY_FIXED_SIZE_MODES): New macro.
17150 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17151 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17152 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17154 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17155 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17156 or if measurement_type is not polynomial.
17157 * genmodes.c (ZERO_COEFFS): New macro.
17158 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17160 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17161 Use ZERO_COEFFS when emitting initializers.
17162 * data-streamer.h (bp_pack_poly_value): New function.
17163 (bp_unpack_poly_value): Likewise.
17164 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17165 for GET_MODE_NUNITS.
17166 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17167 for GET_MODE_NUNITS.
17168 * tree.c (make_vector_type): Remove temporary shim and make
17169 the real function take the number of units as a poly_uint64
17170 rather than an int.
17171 (build_vector_type_for_mode): Handle polynomial nunits.
17172 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17173 * emit-rtl.c (const_vec_series_p_1): Likewise.
17174 (gen_rtx_CONST_VECTOR): Likewise.
17175 * fold-const.c (test_vec_duplicate_folding): Likewise.
17176 * genrecog.c (validate_pattern): Likewise.
17177 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17178 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17179 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17180 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17181 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17182 * rtlanal.c (subreg_get_info): Likewise.
17183 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17184 (vect_grouped_load_supported): Likewise.
17185 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17186 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17187 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17188 (simplify_const_unary_operation, simplify_binary_operation_1)
17189 (simplify_const_binary_operation, simplify_ternary_operation)
17190 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17191 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17192 instead of CONST_VECTOR_NUNITS.
17193 * varasm.c (output_constant_pool_2): Likewise.
17194 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17195 explicit-encoded elements in the XVEC for variable-length vectors.
17197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17199 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17202 Alan Hayward <alan.hayward@arm.com>
17203 David Sherwood <david.sherwood@arm.com>
17205 * coretypes.h (fixed_size_mode): Declare.
17206 (fixed_size_mode_pod): New typedef.
17207 * builtins.h (target_builtins::x_apply_args_mode)
17208 (target_builtins::x_apply_result_mode): Change type to
17209 fixed_size_mode_pod.
17210 * builtins.c (apply_args_size, apply_result_size, result_vector)
17211 (expand_builtin_apply_args_1, expand_builtin_apply)
17212 (expand_builtin_return): Update accordingly.
17214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17216 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17217 * cselib.c (cselib_hash_rtx): Likewise.
17218 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17219 CONST_VECTOR encoding.
17221 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17222 Jeff Law <law@redhat.com>
17225 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17226 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17227 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17228 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17231 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17232 explicitly probe *sp in a noreturn function if there were any callee
17233 register saves or frame pointer is needed.
17235 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17238 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17239 BLKmode for ternary, binary or unary expressions.
17242 * var-tracking.c (delete_vta_debug_insn): New inline function.
17243 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17244 insns from get_insns () to NULL instead of each bb separately.
17245 Use delete_vta_debug_insn. No longer static.
17246 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17247 delete_vta_debug_insns callers.
17248 * rtl.h (delete_vta_debug_insns): Declare.
17249 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17250 instead of variable_tracking_main.
17252 2018-01-03 Martin Sebor <msebor@redhat.com>
17254 PR tree-optimization/83603
17255 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17256 arguments past the endof the argument list in functions declared
17257 without a prototype.
17258 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17259 Avoid checking when arguments are null.
17261 2018-01-03 Martin Sebor <msebor@redhat.com>
17264 * doc/extend.texi (attribute const): Fix a typo.
17265 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17266 issuing -Wsuggest-attribute for void functions.
17268 2018-01-03 Martin Sebor <msebor@redhat.com>
17270 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17271 offset_int::from instead of wide_int::to_shwi.
17272 (maybe_diag_overlap): Remove assertion.
17273 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17274 * gimple-ssa-sprintf.c (format_directive): Same.
17275 (parse_directive): Same.
17276 (sprintf_dom_walker::compute_format_length): Same.
17277 (try_substitute_return_value): Same.
17279 2018-01-03 Jeff Law <law@redhat.com>
17281 PR middle-end/83654
17282 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17283 non-constant residual for zero at runtime and avoid probing in
17284 that case. Reorganize code for trailing problem to mirror handling
17287 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17289 PR tree-optimization/83501
17290 * tree-ssa-strlen.c (get_string_cst): New.
17291 (handle_char_store): Call get_string_cst.
17293 2018-01-03 Martin Liska <mliska@suse.cz>
17295 PR tree-optimization/83593
17296 * tree-ssa-strlen.c: Include tree-cfg.h.
17297 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17298 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17299 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17301 (strlen_dom_walker::before_dom_children): Call
17302 gimple_purge_dead_eh_edges. Dump tranformation with details
17304 (strlen_dom_walker::before_dom_children): Update call by adding
17305 new argument cleanup_eh.
17306 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17308 2018-01-03 Martin Liska <mliska@suse.cz>
17311 * cif-code.def (VARIADIC_THUNK): New enum value.
17312 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17315 2018-01-03 Jan Beulich <jbeulich@suse.com>
17317 * sse.md (mov<mode>_internal): Tighten condition for when to use
17318 vmovdqu<ssescalarsize> for TI and OI modes.
17320 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17322 Update copyright years.
17324 2018-01-03 Martin Liska <mliska@suse.cz>
17327 * ipa-visibility.c (function_and_variable_visibility): Skip
17328 functions with noipa attribure.
17330 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17332 * gcc.c (process_command): Update copyright notice dates.
17333 * gcov-dump.c (print_version): Ditto.
17334 * gcov.c (print_version): Ditto.
17335 * gcov-tool.c (print_version): Ditto.
17336 * gengtype.c (create_file): Ditto.
17337 * doc/cpp.texi: Bump @copying's copyright year.
17338 * doc/cppinternals.texi: Ditto.
17339 * doc/gcc.texi: Ditto.
17340 * doc/gccint.texi: Ditto.
17341 * doc/gcov.texi: Ditto.
17342 * doc/install.texi: Ditto.
17343 * doc/invoke.texi: Ditto.
17345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17347 * vector-builder.h (vector_builder::m_full_nelts): Change from
17348 unsigned int to poly_uint64.
17349 (vector_builder::full_nelts): Update prototype accordingly.
17350 (vector_builder::new_vector): Likewise.
17351 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17352 (vector_builder::operator ==): Likewise.
17353 (vector_builder::finalize): Likewise.
17354 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17355 Take the number of elements as a poly_uint64 rather than an
17357 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17358 from unsigned int to poly_uint64.
17359 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17360 (vec_perm_indices::new_vector): Likewise.
17361 (vec_perm_indices::length): Likewise.
17362 (vec_perm_indices::nelts_per_input): Likewise.
17363 (vec_perm_indices::input_nelts): Likewise.
17364 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17365 number of elements per input as a poly_uint64 rather than an
17366 unsigned int. Use the original encoding for variable-length
17367 vectors, rather than clamping each individual element.
17368 For the second and subsequent elements in each pattern,
17369 clamp the step and base before clamping their sum.
17370 (vec_perm_indices::series_p): Handle polynomial element counts.
17371 (vec_perm_indices::all_in_range_p): Likewise.
17372 (vec_perm_indices_to_tree): Likewise.
17373 (vec_perm_indices_to_rtx): Likewise.
17374 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17375 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17376 (tree_vector_builder::new_binary_operation): Handle polynomial
17377 element counts. Return false if we need to know the number
17378 of elements at compile time.
17379 * fold-const.c (fold_vec_perm): Punt if the number of elements
17380 isn't known at compile time.
17382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17384 * vec-perm-indices.h (vec_perm_builder): Change element type
17385 from HOST_WIDE_INT to poly_int64.
17386 (vec_perm_indices::element_type): Update accordingly.
17387 (vec_perm_indices::clamp): Handle polynomial element_types.
17388 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17389 (vec_perm_indices::all_in_range_p): Likewise.
17390 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17392 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17393 polynomial vec_perm_indices element types.
17394 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17395 * fold-const.c (fold_vec_perm): Likewise.
17396 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17397 * tree-vect-generic.c (lower_vec_perm): Likewise.
17398 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17399 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17400 element type to HOST_WIDE_INT.
17402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17403 Alan Hayward <alan.hayward@arm.com>
17404 David Sherwood <david.sherwood@arm.com>
17406 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17407 rather than an int. Use plus_constant.
17408 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17409 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17412 Alan Hayward <alan.hayward@arm.com>
17413 David Sherwood <david.sherwood@arm.com>
17415 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17416 a HOST_WIDE_INT to a poly_int64.
17418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17419 Alan Hayward <alan.hayward@arm.com>
17420 David Sherwood <david.sherwood@arm.com>
17422 * calls.c (load_register_parameters): Cope with polynomial
17423 mode sizes. Require a constant size for BLKmode parameters
17424 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17425 forces a parameter to be padded at the lsb end in order to
17426 fill a complete number of words, require the parameter size
17427 to be ordered wrt UNITS_PER_WORD.
17429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17430 Alan Hayward <alan.hayward@arm.com>
17431 David Sherwood <david.sherwood@arm.com>
17433 * reload1.c (spill_stack_slot_width): Change element type
17434 from unsigned int to poly_uint64_pod.
17435 (alter_reg): Treat mode sizes as polynomial.
17437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17438 Alan Hayward <alan.hayward@arm.com>
17439 David Sherwood <david.sherwood@arm.com>
17441 * reload.c (complex_word_subreg_p): New function.
17442 (reload_inner_reg_of_subreg, push_reload): Use it.
17444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17445 Alan Hayward <alan.hayward@arm.com>
17446 David Sherwood <david.sherwood@arm.com>
17448 * lra-constraints.c (process_alt_operands): Reject matched
17449 operands whose sizes aren't ordered.
17450 (match_reload): Refer to this check here.
17452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17453 Alan Hayward <alan.hayward@arm.com>
17454 David Sherwood <david.sherwood@arm.com>
17456 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17457 that the mode size is in the set {1, 2, 4, 8, 16}.
17459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17460 Alan Hayward <alan.hayward@arm.com>
17461 David Sherwood <david.sherwood@arm.com>
17463 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17464 Use plus_constant instead of gen_rtx_PLUS.
17466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17467 Alan Hayward <alan.hayward@arm.com>
17468 David Sherwood <david.sherwood@arm.com>
17470 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17471 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17472 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17473 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17474 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17475 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17476 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17477 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17478 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17479 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17481 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17482 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17483 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17484 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17485 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17486 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17487 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17488 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17489 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17490 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17492 * expr.c (emit_move_resolve_push): Treat the input and result
17493 of PUSH_ROUNDING as a poly_int64.
17494 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17495 (emit_push_insn): Likewise.
17496 * lra-eliminations.c (mark_not_eliminable): Likewise.
17497 * recog.c (push_operand): Likewise.
17498 * reload1.c (elimination_effects): Likewise.
17499 * rtlanal.c (nonzero_bits1): Likewise.
17500 * calls.c (store_one_arg): Likewise. Require the padding to be
17501 known at compile time.
17503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17504 Alan Hayward <alan.hayward@arm.com>
17505 David Sherwood <david.sherwood@arm.com>
17507 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17508 Use plus_constant instead of gen_rtx_PLUS.
17510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17511 Alan Hayward <alan.hayward@arm.com>
17512 David Sherwood <david.sherwood@arm.com>
17514 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17515 rather than an int.
17517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17518 Alan Hayward <alan.hayward@arm.com>
17519 David Sherwood <david.sherwood@arm.com>
17521 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17522 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17523 via stack temporaries. Treat the mode size as polynomial too.
17525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17526 Alan Hayward <alan.hayward@arm.com>
17527 David Sherwood <david.sherwood@arm.com>
17529 * expr.c (expand_expr_real_2): When handling conversions involving
17530 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17531 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17532 as a poly_uint64 too.
17534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17535 Alan Hayward <alan.hayward@arm.com>
17536 David Sherwood <david.sherwood@arm.com>
17538 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17541 Alan Hayward <alan.hayward@arm.com>
17542 David Sherwood <david.sherwood@arm.com>
17544 * combine.c (can_change_dest_mode): Handle polynomial
17545 REGMODE_NATURAL_SIZE.
17546 * expmed.c (store_bit_field_1): Likewise.
17547 * expr.c (store_constructor): Likewise.
17548 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17549 and polynomial REGMODE_NATURAL_SIZE.
17550 (gen_lowpart_common): Likewise.
17551 * reginfo.c (record_subregs_of_mode): Likewise.
17552 * rtlanal.c (read_modify_subreg_p): Likewise.
17554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17555 Alan Hayward <alan.hayward@arm.com>
17556 David Sherwood <david.sherwood@arm.com>
17558 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17559 numbers of elements.
17561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17562 Alan Hayward <alan.hayward@arm.com>
17563 David Sherwood <david.sherwood@arm.com>
17565 * match.pd: Cope with polynomial numbers of vector elements.
17567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17568 Alan Hayward <alan.hayward@arm.com>
17569 David Sherwood <david.sherwood@arm.com>
17571 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17572 in a POINTER_PLUS_EXPR.
17574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17575 Alan Hayward <alan.hayward@arm.com>
17576 David Sherwood <david.sherwood@arm.com>
17578 * omp-simd-clone.c (simd_clone_subparts): New function.
17579 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17580 (ipa_simd_modify_function_body): Likewise.
17582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17583 Alan Hayward <alan.hayward@arm.com>
17584 David Sherwood <david.sherwood@arm.com>
17586 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17587 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17588 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17589 (expand_vector_condition, vector_element): Likewise.
17590 (subparts_gt): New function.
17591 (get_compute_type): Use subparts_gt.
17592 (count_type_subparts): Delete.
17593 (expand_vector_operations_1): Use subparts_gt instead of
17594 count_type_subparts.
17596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17597 Alan Hayward <alan.hayward@arm.com>
17598 David Sherwood <david.sherwood@arm.com>
17600 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17601 (vect_compile_time_alias): ...this new function. Do the calculation
17602 on poly_ints rather than trees.
17603 (vect_prune_runtime_alias_test_list): Update call accordingly.
17605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17606 Alan Hayward <alan.hayward@arm.com>
17607 David Sherwood <david.sherwood@arm.com>
17609 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17611 (vect_schedule_slp_instance): Likewise.
17613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17614 Alan Hayward <alan.hayward@arm.com>
17615 David Sherwood <david.sherwood@arm.com>
17617 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17618 constant and extern definitions for variable-length vectors.
17619 (vect_get_constant_vectors): Note that the number of units
17620 is known to be constant.
17622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17623 Alan Hayward <alan.hayward@arm.com>
17624 David Sherwood <david.sherwood@arm.com>
17626 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17627 of units as polynomial. Choose between WIDE and NARROW based
17630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17631 Alan Hayward <alan.hayward@arm.com>
17632 David Sherwood <david.sherwood@arm.com>
17634 * tree-vect-stmts.c (simd_clone_subparts): New function.
17635 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17638 Alan Hayward <alan.hayward@arm.com>
17639 David Sherwood <david.sherwood@arm.com>
17641 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17642 vectors as polynomial. Use build_index_vector for
17643 IFN_GOMP_SIMD_LANE.
17645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17646 Alan Hayward <alan.hayward@arm.com>
17647 David Sherwood <david.sherwood@arm.com>
17649 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17650 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17651 for variable-length vectors.
17652 (vectorizable_mask_load_store): Treat the number of units as
17653 polynomial, asserting that it is constant if the condition has
17654 already been enforced.
17655 (vectorizable_store, vectorizable_load): Likewise.
17657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17658 Alan Hayward <alan.hayward@arm.com>
17659 David Sherwood <david.sherwood@arm.com>
17661 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17662 of units as polynomial. Punt if we can't tell at compile time
17663 which vector contains the final result.
17665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17666 Alan Hayward <alan.hayward@arm.com>
17667 David Sherwood <david.sherwood@arm.com>
17669 * tree-vect-loop.c (vectorizable_induction): Treat the number
17670 of units as polynomial. Punt on SLP inductions. Use an integer
17671 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17672 cast of such a series for variable-length floating-point
17675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17676 Alan Hayward <alan.hayward@arm.com>
17677 David Sherwood <david.sherwood@arm.com>
17679 * tree.h (build_index_vector): Declare.
17680 * tree.c (build_index_vector): New function.
17681 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17682 of units as polynomial, forcibly converting it to a constant if
17683 vectorizable_reduction has already enforced the condition.
17684 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17685 to create a {1,2,3,...} vector.
17686 (vectorizable_reduction): Treat the number of units as polynomial.
17687 Choose vectype_in based on the largest scalar element size rather
17688 than the smallest number of units. Enforce the restrictions
17691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17692 Alan Hayward <alan.hayward@arm.com>
17693 David Sherwood <david.sherwood@arm.com>
17695 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17696 number of units as polynomial.
17698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17699 Alan Hayward <alan.hayward@arm.com>
17700 David Sherwood <david.sherwood@arm.com>
17702 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17703 * target.def (autovectorize_vector_sizes): Return the vector sizes
17704 by pointer, using vector_sizes rather than a bitmask.
17705 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17706 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17707 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17709 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17710 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17711 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17712 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17713 * omp-general.c (omp_max_vf): Likewise.
17714 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17715 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17716 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17717 * tree-vect-slp.c (vect_slp_bb): Likewise.
17718 * doc/tm.texi: Regenerate.
17719 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17721 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17722 the vector size as a poly_uint64 rather than an unsigned int.
17723 (current_vector_size): Change from an unsigned int to a poly_uint64.
17724 (get_vectype_for_scalar_type): Update accordingly.
17725 * tree.h (build_truth_vector_type): Take the size and number of
17726 units as a poly_uint64 rather than an unsigned int.
17727 (build_vector_type): Add a temporary overload that takes
17728 the number of units as a poly_uint64 rather than an unsigned int.
17729 * tree.c (make_vector_type): Likewise.
17730 (build_truth_vector_type): Take the number of units as a poly_uint64
17731 rather than an unsigned int.
17733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17734 Alan Hayward <alan.hayward@arm.com>
17735 David Sherwood <david.sherwood@arm.com>
17737 * target.def (get_mask_mode): Take the number of units and length
17738 as poly_uint64s rather than unsigned ints.
17739 * targhooks.h (default_get_mask_mode): Update accordingly.
17740 * targhooks.c (default_get_mask_mode): Likewise.
17741 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17742 * doc/tm.texi: Regenerate.
17744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17745 Alan Hayward <alan.hayward@arm.com>
17746 David Sherwood <david.sherwood@arm.com>
17748 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17749 * omp-general.c (omp_max_vf): Likewise.
17750 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17751 (expand_omp_simd): Handle polynomial safelen.
17752 * omp-low.c (omplow_simd_context): Add a default constructor.
17753 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17754 (lower_rec_simd_input_clauses): Update accordingly.
17755 (lower_rec_input_clauses): Likewise.
17757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17758 Alan Hayward <alan.hayward@arm.com>
17759 David Sherwood <david.sherwood@arm.com>
17761 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17762 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17763 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17764 (vect_analyze_slp_cost): Likewise.
17765 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17766 (vect_model_load_cost): Likewise.
17768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17769 Alan Hayward <alan.hayward@arm.com>
17770 David Sherwood <david.sherwood@arm.com>
17772 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17773 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17774 from an unsigned int * to a poly_uint64_pod *.
17775 (calculate_unrolling_factor): New function.
17776 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17779 Alan Hayward <alan.hayward@arm.com>
17780 David Sherwood <david.sherwood@arm.com>
17782 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17783 from an unsigned int to a poly_uint64.
17784 (_loop_vec_info::slp_unrolling_factor): Likewise.
17785 (_loop_vec_info::vectorization_factor): Change from an int
17787 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17788 (vect_get_num_vectors): New function.
17789 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17790 (vect_get_num_copies): Use vect_get_num_vectors.
17791 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17792 to an unsigned int *.
17793 (vect_analyze_data_refs): Change min_vf from an int * to a
17795 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17796 than an unsigned HOST_WIDE_INT.
17797 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17798 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17799 to an unsigned int *.
17800 (vect_analyze_data_ref_dependences): Likewise.
17801 (vect_compute_data_ref_alignment): Handle polynomial vf.
17802 (vect_enhance_data_refs_alignment): Likewise.
17803 (vect_prune_runtime_alias_test_list): Likewise.
17804 (vect_shift_permute_load_chain): Likewise.
17805 (vect_supportable_dr_alignment): Likewise.
17806 (dependence_distance_ge_vf): Take the vectorization factor as a
17807 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17808 (vect_analyze_data_refs): Change min_vf from an int * to a
17810 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17811 vfm1 as a poly_uint64 rather than an int. Make the same change
17812 for the returned bound_scalar.
17813 (vect_gen_vector_loop_niters): Handle polynomial vf.
17814 (vect_do_peeling): Likewise. Update call to
17815 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17816 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17818 * tree-vect-loop.c (vect_determine_vectorization_factor)
17819 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17820 (vect_get_known_peeling_cost): Likewise.
17821 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17822 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17823 (vect_transform_loop): Likewise. Use the lowest possible VF when
17824 updating the upper bounds of the loop.
17825 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17826 rather than an int.
17827 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17828 polynomial unroll factors.
17829 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17830 (vect_make_slp_decision): Likewise.
17831 (vect_supported_load_permutation_p): Likewise, and polynomial
17833 (vect_analyze_slp_cost): Handle polynomial vf.
17834 (vect_slp_analyze_node_operations): Likewise.
17835 (vect_slp_analyze_bb_1): Likewise.
17836 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17837 than an unsigned HOST_WIDE_INT.
17838 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17839 (vectorizable_load): Handle polynomial vf.
17840 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17842 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17845 Alan Hayward <alan.hayward@arm.com>
17846 David Sherwood <david.sherwood@arm.com>
17848 * match.pd: Handle bit operations involving three constants
17849 and try to fold one pair.
17851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17853 * tree-vect-loop-manip.c: Include gimple-fold.h.
17854 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17855 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17856 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17857 Add a path that uses a step of VF instead of 1, but disable it
17859 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17860 and niters_no_overflow parameters. Update calls to
17861 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17862 Create a new SSA name if the latter choses to use a ste other
17863 than zero, and return it via niters_vector_mult_vf_var.
17864 * tree-vect-loop.c (vect_transform_loop): Update calls to
17865 vect_do_peeling, vect_gen_vector_loop_niters and
17866 slpeel_make_loop_iterate_ntimes.
17867 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17868 (vect_gen_vector_loop_niters): Update declarations after above changes.
17870 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17872 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17873 128-bit round to integer instructions.
17874 (ceil<mode>2): Likewise.
17875 (btrunc<mode>2): Likewise.
17876 (round<mode>2): Likewise.
17878 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17880 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17881 unaligned VSX load/store on P8/P9.
17882 (expand_block_clear): Allow the use of unaligned VSX
17883 load/store on P8/P9.
17885 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17887 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17889 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17890 swap associated with both a load and a store.
17892 2018-01-02 Andrew Waterman <andrew@sifive.com>
17894 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17895 * config/riscv/riscv.md (clear_cache): Use it.
17897 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17899 * web.c: Remove out-of-date comment.
17901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17903 * expr.c (fixup_args_size_notes): Check that any existing
17904 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17905 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17906 (emit_single_push_insn): ...here.
17908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17910 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17911 (const_vector_encoded_nelts): New function.
17912 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17913 (const_vector_int_elt, const_vector_elt): Declare.
17914 * emit-rtl.c (const_vector_int_elt_1): New function.
17915 (const_vector_elt): Likewise.
17916 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17917 of CONST_VECTOR_ELT.
17919 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17921 * expr.c: Include rtx-vector-builder.h.
17922 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17923 directly on the tree encoding.
17924 (const_vector_from_tree): Likewise.
17925 * optabs.c: Include rtx-vector-builder.h.
17926 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17927 sequence of "u" values.
17928 * vec-perm-indices.c: Include rtx-vector-builder.h.
17929 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17930 directly on the vec_perm_indices encoding.
17932 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17934 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17935 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17936 * rtx-vector-builder.h: New file.
17937 * rtx-vector-builder.c: Likewise.
17938 * rtl.h (rtx_def::u2): Add a const_vector field.
17939 (CONST_VECTOR_NPATTERNS): New macro.
17940 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17941 (CONST_VECTOR_DUPLICATE_P): Likewise.
17942 (CONST_VECTOR_STEPPED_P): Likewise.
17943 (CONST_VECTOR_ENCODED_ELT): Likewise.
17944 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17945 (unwrap_const_vec_duplicate): Likewise.
17946 (const_vec_series_p): Check for a non-duplicated vector encoding.
17947 Say that the function only returns true for integer vectors.
17948 * emit-rtl.c: Include rtx-vector-builder.h.
17949 (gen_const_vec_duplicate_1): Delete.
17950 (gen_const_vector): Call gen_const_vec_duplicate instead of
17951 gen_const_vec_duplicate_1.
17952 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17953 (gen_const_vec_duplicate): Use rtx_vector_builder.
17954 (gen_const_vec_series): Likewise.
17955 (gen_rtx_CONST_VECTOR): Likewise.
17956 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17957 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17958 Build a new vector rather than modifying a CONST_VECTOR in-place.
17959 (handle_special_swappables): Update call accordingly.
17960 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17961 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17962 Build a new vector rather than modifying a CONST_VECTOR in-place.
17963 (handle_special_swappables): Update call accordingly.
17965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17967 * simplify-rtx.c (simplify_const_binary_operation): Use
17968 CONST_VECTOR_ELT instead of XVECEXP.
17970 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17972 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17973 the selector elements to be different from the data elements
17974 if the selector is a VECTOR_CST.
17975 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17976 ssizetype for the selector.
17978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17980 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17981 before testing each element individually.
17982 * tree-vect-generic.c (lower_vec_perm): Likewise.
17984 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17986 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17987 * selftest-run-tests.c (selftest::run_tests): Call it.
17988 * vector-builder.h (vector_builder::operator ==): New function.
17989 (vector_builder::operator !=): Likewise.
17990 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17991 (vec_perm_indices::all_from_input_p): New function.
17992 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17993 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17994 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17995 instead of reading the VECTOR_CST directly. Detect whether both
17996 vector inputs are the same before constructing the vec_perm_indices,
17997 and update the number of inputs argument accordingly. Use the
17998 utility functions added above. Only construct sel2 if we need to.
18000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18002 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18003 the broadcast of the low byte.
18004 (expand_mult_highpart): Use an explicit encoding for the permutes.
18005 * optabs-query.c (can_mult_highpart_p): Likewise.
18006 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18007 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18008 (vectorizable_bswap): Likewise.
18009 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18010 explicit encoding for the power-of-2 permutes.
18011 (vect_permute_store_chain): Likewise.
18012 (vect_grouped_load_supported): Likewise.
18013 (vect_permute_load_chain): Likewise.
18015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18017 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18018 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18019 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18020 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18021 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18022 (vect_gen_perm_mask_any): Likewise.
18024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18026 * int-vector-builder.h: New file.
18027 * vec-perm-indices.h: Include int-vector-builder.h.
18028 (vec_perm_indices): Redefine as an int_vector_builder.
18029 (auto_vec_perm_indices): Delete.
18030 (vec_perm_builder): Redefine as a stand-alone class.
18031 (vec_perm_indices::vec_perm_indices): New function.
18032 (vec_perm_indices::clamp): Likewise.
18033 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18034 (vec_perm_indices::new_vector): New function.
18035 (vec_perm_indices::new_expanded_vector): Update for new
18036 vec_perm_indices class.
18037 (vec_perm_indices::rotate_inputs): New function.
18038 (vec_perm_indices::all_in_range_p): Operate directly on the
18039 encoded form, without computing elided elements.
18040 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18041 encoding. Update for new vec_perm_indices class.
18042 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18043 the given vec_perm_builder.
18044 (expand_vec_perm_var): Update vec_perm_builder constructor.
18045 (expand_mult_highpart): Use vec_perm_builder instead of
18046 auto_vec_perm_indices.
18047 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18048 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18049 or double series encoding as appropriate.
18050 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18051 vec_perm_indices instead of auto_vec_perm_indices.
18052 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18053 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18054 (vect_permute_store_chain): Likewise.
18055 (vect_grouped_load_supported): Likewise.
18056 (vect_permute_load_chain): Likewise.
18057 (vect_shift_permute_load_chain): Likewise.
18058 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18059 (vect_transform_slp_perm_load): Likewise.
18060 (vect_schedule_slp_instance): Likewise.
18061 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18062 (vectorizable_mask_load_store): Likewise.
18063 (vectorizable_bswap): Likewise.
18064 (vectorizable_store): Likewise.
18065 (vectorizable_load): Likewise.
18066 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18067 vec_perm_indices instead of auto_vec_perm_indices. Use
18068 tree_to_vec_perm_builder to read the vector from a tree.
18069 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18070 vec_perm_builder instead of a vec_perm_indices.
18071 (have_whole_vector_shift): Use vec_perm_builder and
18072 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18073 truncation to calc_vec_perm_mask_for_shift.
18074 (vect_create_epilog_for_reduction): Likewise.
18075 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18076 from auto_vec_perm_indices to vec_perm_indices.
18077 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18078 instead of changing individual elements.
18079 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18080 the vector in d.perm.
18081 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18082 from auto_vec_perm_indices to vec_perm_indices.
18083 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18084 instead of changing individual elements.
18085 (arm_vectorize_vec_perm_const): Use new_vector to install
18086 the vector in d.perm.
18087 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18088 Update vec_perm_builder constructor.
18089 (rs6000_expand_interleave): Likewise.
18090 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18091 (rs6000_expand_interleave): Likewise.
18093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18095 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18096 to qimode could truncate the indices.
18097 * optabs.c (expand_vec_perm_var): Likewise.
18099 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18101 * Makefile.in (OBJS): Add vec-perm-indices.o.
18102 * vec-perm-indices.h: New file.
18103 * vec-perm-indices.c: Likewise.
18104 * target.h (vec_perm_indices): Replace with a forward class
18106 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18107 * optabs.h: Include vec-perm-indices.h.
18108 (expand_vec_perm): Delete.
18109 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18110 (expand_vec_perm_const): Declare.
18111 * target.def (vec_perm_const_ok): Replace with...
18112 (vec_perm_const): ...this new hook.
18113 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18114 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18115 * doc/tm.texi: Regenerate.
18116 * optabs.def (vec_perm_const): Delete.
18117 * doc/md.texi (vec_perm_const): Likewise.
18118 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18119 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18120 expand_vec_perm for constant permutation vectors. Assert that
18121 the mode of variable permutation vectors is the integer equivalent
18122 of the mode that is being permuted.
18123 * optabs-query.h (selector_fits_mode_p): Declare.
18124 * optabs-query.c: Include vec-perm-indices.h.
18125 (selector_fits_mode_p): New function.
18126 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18127 is defined, instead of checking whether the vec_perm_const_optab
18128 exists. Use targetm.vectorize.vec_perm_const instead of
18129 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18130 fit in the vector mode before using a variable permute.
18131 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18132 vec_perm_indices instead of an rtx.
18133 (expand_vec_perm): Replace with...
18134 (expand_vec_perm_const): ...this new function. Take the selector
18135 as a vec_perm_indices rather than an rtx. Also take the mode of
18136 the selector. Update call to shift_amt_for_vec_perm_mask.
18137 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18138 Use vec_perm_indices::new_expanded_vector to expand the original
18139 selector into bytes. Check whether the indices fit in the vector
18140 mode before using a variable permute.
18141 (expand_vec_perm_var): Make global.
18142 (expand_mult_highpart): Use expand_vec_perm_const.
18143 * fold-const.c: Includes vec-perm-indices.h.
18144 * tree-ssa-forwprop.c: Likewise.
18145 * tree-vect-data-refs.c: Likewise.
18146 * tree-vect-generic.c: Likewise.
18147 * tree-vect-loop.c: Likewise.
18148 * tree-vect-slp.c: Likewise.
18149 * tree-vect-stmts.c: Likewise.
18150 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18152 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18153 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18154 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18155 (aarch64_vectorize_vec_perm_const): ...this new function.
18156 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18157 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18158 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18159 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18160 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18161 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18162 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18164 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18165 check for NEON modes.
18166 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18167 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18168 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18169 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18171 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18172 the old VEC_PERM_CONST conditions.
18173 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18174 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18175 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18176 (ia64_vectorize_vec_perm_const_ok): Merge into...
18177 (ia64_vectorize_vec_perm_const): ...this new function.
18178 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18179 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18180 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18181 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18182 * config/mips/mips.c (mips_expand_vec_perm_const)
18183 (mips_vectorize_vec_perm_const_ok): Merge into...
18184 (mips_vectorize_vec_perm_const): ...this new function.
18185 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18186 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18187 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18188 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18189 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18190 (rs6000_expand_vec_perm_const): Delete.
18191 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18193 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18194 (altivec_expand_vec_perm_const_le): Take each operand individually.
18195 Operate on constant selectors rather than rtxes.
18196 (altivec_expand_vec_perm_const): Likewise. Update call to
18197 altivec_expand_vec_perm_const_le.
18198 (rs6000_expand_vec_perm_const): Delete.
18199 (rs6000_vectorize_vec_perm_const_ok): Delete.
18200 (rs6000_vectorize_vec_perm_const): New function.
18201 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18202 an element count and rtx array.
18203 (rs6000_expand_extract_even): Update call accordingly.
18204 (rs6000_expand_interleave): Likewise.
18205 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18206 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18207 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18208 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18209 (rs6000_expand_vec_perm_const): Delete.
18210 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18211 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18212 (altivec_expand_vec_perm_const_le): Take each operand individually.
18213 Operate on constant selectors rather than rtxes.
18214 (altivec_expand_vec_perm_const): Likewise. Update call to
18215 altivec_expand_vec_perm_const_le.
18216 (rs6000_expand_vec_perm_const): Delete.
18217 (rs6000_vectorize_vec_perm_const_ok): Delete.
18218 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18219 reference to the SPE evmerge intructions.
18220 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18221 an element count and rtx array.
18222 (rs6000_expand_extract_even): Update call accordingly.
18223 (rs6000_expand_interleave): Likewise.
18224 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18225 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18227 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18231 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18232 vector mode and that that mode matches the mode of the data
18234 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18235 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18236 directly using expand_vec_perm_1 when forcing selectors into
18238 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18242 * optabs-query.h (can_vec_perm_p): Delete.
18243 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18244 * optabs-query.c (can_vec_perm_p): Split into...
18245 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18246 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18247 particular selector is valid.
18248 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18249 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18250 (vect_grouped_load_supported): Likewise.
18251 (vect_shift_permute_load_chain): Likewise.
18252 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18253 (vect_transform_slp_perm_load): Likewise.
18254 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18255 (vectorizable_bswap): Likewise.
18256 (vect_gen_perm_mask_checked): Likewise.
18257 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18258 implementations of variable permutation vectors into account
18259 when deciding which selector to use.
18260 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18261 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18262 with a false third argument.
18263 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18264 to test whether the constant selector is valid and can_vec_perm_var_p
18265 to test whether a variable selector is valid.
18267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18269 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18270 * optabs-query.c (can_vec_perm_p): Likewise.
18271 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18272 instead of vec_perm_indices.
18273 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18274 (vect_gen_perm_mask_checked): Likewise,
18275 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18276 (vect_gen_perm_mask_checked): Likewise,
18278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18280 * optabs-query.h (qimode_for_vec_perm): Declare.
18281 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18282 (qimode_for_vec_perm): ...this new function.
18283 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18285 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18287 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18288 does not have a conditional at the top.
18290 2018-01-02 Richard Biener <rguenther@suse.de>
18292 * ipa-inline.c (big_speedup_p): Fix expression.
18294 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18297 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18300 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18304 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18305 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18306 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18307 cond_taken_branch_cost 3->4.
18309 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18311 PR tree-optimization/83581
18312 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18313 TODO_cleanup_cfg if any changes have been made.
18315 PR middle-end/83608
18316 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18317 convert_modes if target mode has the right side, but different mode
18320 PR middle-end/83609
18321 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18322 last argument when extracting from CONCAT. If either from_real or
18323 from_imag is NULL, use expansion through memory. If result is not
18324 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18325 the parts directly to inner mode, if even that fails, use expansion
18328 PR middle-end/83623
18329 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18330 check for bswap in mode rather than HImode and use that in expand_unop
18333 Copyright (C) 2018 Free Software Foundation, Inc.
18335 Copying and distribution of this file, with or without modification,
18336 are permitted in any medium without royalty provided the copyright
18337 notice and this notice are preserved.