1 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
4 goto_locus of each outgoing edge of each basic block.
6 2018-06-28 Richard Biener <rguenther@suse.de>
8 * dwarf2out.c (decl_scope_table): Remove.
9 (push_decl_scope): Likewise.
10 (pop_decl_scope): Likewise.
11 (gen_type_die_for_member): Do not call push/pop_decl_scope.
12 (gen_struct_or_union_type_die): Likewise.
13 (gen_tagged_type_die): Likewise.
14 (dwarf2out_init): Do not initialize decl_scope_table.
15 (dwarf2out_c_finalize): Do not free it.
17 2018-06-28 Richard Biener <rguenther@suse.de>
19 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
20 deciding whether to not re-use a DIE.
22 2018-06-28 Richard Biener <rguenther@suse.de>
24 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
25 DW_AT_abstract_origin attribute.
27 2018-06-28 Martin Liska <mliska@suse.cz>
29 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
30 Use newly introduced constants.
31 * tree-switch-conversion.h (struct jump_table_cluster):
32 Define max_ratio_for_size and max_ratio_for_speed.
34 2018-06-28 Martin Liska <mliska@suse.cz>
36 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
37 Add new checking assert to catch invalid state.
38 (jump_table_cluster::can_be_handled): Handle single case
40 (jump_table_cluster::is_beneficial): Bail out for such case.
41 (bit_test_cluster::find_bit_tests):
42 Add new checking assert to catch invalid state.
43 (bit_test_cluster::can_be_handled): Handle single case
45 (bit_test_cluster::is_beneficial): Bail out for such case.
46 (switch_decision_tree::analyze_switch_statement):
49 2018-06-28 Martin Liska <mliska@suse.cz>
51 * common.opt: Introduce -completion option.
52 * gcc.c (driver_handle_option): Handle it.
53 (driver::main): Print completions if completion
55 * opt-suggestions.c (option_proposer::get_completions):
57 (option_proposer::suggest_completion): Likewise.
58 (option_proposer::find_param_completions): Likewise.
59 (verify_autocompletions): Likewise.
60 (test_completion_valid_options): Likewise.
61 (test_completion_valid_params): Likewise.
62 (in_completion_p): Likewise.
63 (empty_completion_p): Likewise.
64 (test_completion_partial_match): Likewise.
65 (test_completion_garbage): Likewise.
66 (opt_proposer_c_tests): Likewise.
67 * opt-suggestions.h: Declare new functions.
68 * opts.c (common_handle_option): Handle OPT__completion_.
69 * selftest-run-tests.c (selftest::run_tests): Add
71 * selftest.c (assert_str_startswith): New.
72 * selftest.h (assert_str_startswith): Likewise.
73 (opt_proposer_c_tests): New.
74 (ASSERT_STR_STARTSWITH): Likewise.
76 2018-06-28 Martin Liska <mliska@suse.cz>
78 * Makefile.in: Add opt-suggestions.o.
79 * gcc-main.c: Include opt-suggestions.h.
80 * gcc.c (driver::driver): Likewise.
81 (driver::~driver): Remove m_option_suggestions.
82 (driver::build_option_suggestions): Moved to option_proposer.
83 (driver::suggest_option): Likewise.
84 (driver::handle_unrecognized_options): Use option_proposer.
85 * gcc.h (class driver): Add new memver m_option_proposer.
86 * opt-suggestions.c: New file.
87 * opt-suggestions.h: New file.
89 2018-06-28 Martin Liska <mliska@suse.cz>
91 * vec.h (class auto_string_vec): New (moved from auto_argvec).
92 (auto_string_vec::~auto_string_vec): Likewise.
94 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
96 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
97 statement if id->reset_location is true.
98 (copy_edges_for_bb): Do not set goto_locus on the new edges if
99 id->reset_location is true.
100 (copy_phis_for_bb): Force input_location on the arguments if
101 id->reset_location is true.
102 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
103 is set on the function to be inlined.
104 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
105 prevent_decl_creation_for_types fields up and add reset_location field.
107 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
109 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
111 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
113 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
115 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
116 hard registers for the clobbered pseudo.
118 2018-06-27 Paul Koning <ni1d@arrl.net>
120 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
121 mutually exclusive options.
122 * config/pdp11/constraints.md (h): New constraint.
123 (O): Update definition to match shift code generation.
125 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
127 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
129 (output_jump): Change arguments.
130 (pdp11_fixed_cc_regs): New function.
131 (pdp11_cc_mode): Ditto.
132 (pdp11_expand_shift): Ditto.
133 (pdp11_assemble_shift): Ditto.
134 (pdp11_small_shift): Ditto.
135 (pdp11_branch_cost): Remove.
136 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
138 (pdp11_register_move_cost): Update for CC registers.
139 (pdp11_rtx_costs): Add case for LSHIFTRT.
140 (pdp11_output_jump): Add CCNZ mode conditional branches.
141 (notice_update_cc_on_set): Remove.
142 (pdp11_cc_mode): New function.
143 (simple_memory_operand): Correct pre/post decrement case.
144 (no_side_effect_operand): New function.
145 (pdp11_regno_reg_class): Add CC_REGS class.
146 (pdp11_fixed_cc_regs): New function.
147 (pdp11_small_shift): New function.
148 (pdp11_expand_shift): New function to expand shift insns.
149 (pdp11_assemble_shift): New function to output shifts.
150 (pdp11_branch_cost): Remove.
151 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
152 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
154 (PTRDIFF_TYPE): Ditto.
155 (ADJUST_INSN_LENGTH): New macro.
156 (FIXED_REGISTERS): Add CC registers.
157 (CALL_USED_REGISTERS): Ditto.
159 (REG_CLASS_NAMES): Ditto.
160 (REG_CLASS_CONTENTS): Ditto.
161 (SELECT_CC_MODE): Use new function.
162 (TARGET_FLAGS_REGNUM): New macro.
163 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
164 (cc0_reg_rtx): Remove.
165 (CC_STATUS_MDEP): Remove.
166 (CC_STATUS_MDEFP_INIT): Remove.
168 (NOTICE_UPDATE_CC): Remove.
169 (REGISTER_NAMES): Add CC registers.
170 (BRANCH_COST): Change to constant 1.
171 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
173 * config/pdp11/pdp11.opt (mbcopy): Remove.
174 (mbcopy-builtin): Remove.
175 (mbranch-cheap): Remove.
176 (mbranch-expensive): Remove.
177 * config/pdp11/predicates.md (expand_shift_operand): Update to
178 match shift code generation.
179 (ccnz_operator): New predicate.
180 * doc/invoke.texi (PDP-11 Options): Remove deleted options
181 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
182 Remove non-existent option -mabshi, -mno-abshi. Document mutually
184 * doc/md.texi (PDP-11): Document new D and h constraints. Update
185 description of O constraint.
187 2018-06-27 Jeff Law <law@redhat.com>
188 Austin Law <austinklaw@gmail.com>
190 * config/v850/v850.md (addsi3_set_flags): New pattern.
191 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
192 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
193 (zero_extendhisi2_v850_set_flags): Likewise.
194 (zero_extendqisi2_v850_set_flags): Likewise.
195 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
196 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
197 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
199 * config/v850/v850-protos.h (notice_update_cc): Remove.
200 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
201 (v850_print_operand): Handle 'D' and "d".
202 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
203 Add handling of arithmetic/logical operations compared against zero.
204 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
205 Do not look at v850_compare_op, instead get mode from last argument.
206 (v850_gen_compare): Remove
207 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
208 after reload for prologue insns.
209 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
211 (construct_save_jarl): Likewise.
212 (TARGET_FLAGS_REGNUM): Define.
213 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
214 (NOTICE_UPDATE_CC): Remove.
215 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
216 than cc0. Conditionalize on reload_completed.
217 (cmpsi_insn, setfcc_insn): Likewise.
218 (tst1 splitter): Turn into define_and_split which sets the flags
220 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
221 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
222 (cstoresf4, cstoredf4): Clobber the flags.
223 (cmpsi, cmpsf, cmpdf): Remove expanders.
224 (setf_insn): Remove pattern.
225 (addsi3): Turn into define_and_split which clobbers the flags after
226 reload and a suitable pattern (addsi3_clobber_flags) for use after
228 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
229 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
230 (ashrsi3, ashrsi3_v850e2): Likewise.
231 (bins): Clobber the flags.
232 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
233 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
234 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
235 (call_value_internal_short, call_value_internal_long): Likewise.
236 (callt_save_interrupt, callt_return_interrupt): Likewise.
237 (save_interrupt, return_interrupt): Likewise.
238 (callt_save_all_interrupt, save_all_interrupt): Likewise.
239 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
240 (restore_all_interrupt, _restore_all_interrupt): Likewise.
241 (All FP comparisons): Only allow after reload has completed.
243 (divh, divhu): Tweak output template.
244 (branch_z_normal, branch_z_invert): Remove
245 (branch_nz_normal, branch_nz_invert): Likewise.
246 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
248 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
249 * config/v850/v850.c (notice_update_cc): Remove.
250 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
251 (CC_NO_CARRY): Likewise.
252 (NOTICE_UPDATE_CC): Define to nothing.
253 * config/v850/v850.md: Remove block comment on cc0 handling
254 Remove "cc" attribute from all patterns. Remove cc_status handling
255 from all patterns. Minor formatting fixes.
257 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
259 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
260 (cortex-a76.cortex-a55): Likewise.
261 * config/aarch64/aarch64-tune.md: Regenerate.
262 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
263 cortex-a76.cortex-a55.
265 2018-06-27 Jeff Law <law@redhat.com>
267 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
268 (MULTILIB_DIRNAMES): Similarly.
270 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
272 * gimple.h (gimple_return_retbnd): Delete.
273 (gimple_return_set_retbnd): Likewise.
274 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
275 gimple_return_set_retbnd.
276 * gimple-pretty-print.c (dump_gimple_return): Remove call to
277 gimple_return_retbnd and adjust.
278 * tree-inline.h (struct copy_body_data): Remove retbnd field.
279 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
280 Explicitly return NULL in a couple more cases. Move assertion
281 on debug statements and remove unreachable code.
282 (reset_debug_binding): Do not test id->retbnd.
283 (expand_call_inline): Do not set it.
285 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
287 * configure.ac: Add --disable-gcov option.
288 * configure: Regenerate.
289 * Makefile.in: Honour @enable_gcov@.
290 * doc/install.texi: Document --disable-gcov.
292 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
294 * config/arm/arm-cpus.in (cortex-a76): New entry.
295 (cortex-a76.cortex-a55): Likewise.
296 * config/arm/arm-tables.opt: Regenerate.
297 * config/arm/arm-tune.md: Likewise.
298 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
299 * doc/invoke.texi (ARM Options): Document cortex-a76 and
300 cortex-a76.cortex-a55.
302 2018-06-27 Tamar Christina <tamar.christina@arm.com>
305 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
307 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
309 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
311 (EPILOGUE_USES): Likewise.
313 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
315 * tree-inline.c (remap_location): New function extracted from...
316 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
317 (copy_phis_for_bb): ...here. Call remap_location.
318 (copy_cfg_body): Adjust call to copy_edges_for_bb.
320 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
322 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
323 unaligned vsx for 16B memset.
325 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
328 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
329 ieee128_float_type_node to long_double_type_node unless
330 TARGET_LONG_DOUBLE_128 is set.
332 2018-06-26 David Malcolm <dmalcolm@redhat.com>
334 * cfgloop.c (get_loop_location): Convert return type from
335 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
336 by implicit construction from rtx_insn *, and using
337 dump_user_location_t::from_function_decl for the fallback case.
338 * cfgloop.h (get_loop_location): Convert return type from
339 location_t to dump_user_location_t.
340 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
341 dump_printf_loc to pass in a dump_location_t rather than a
342 location_t, via the gimple stmt.
343 * coverage.c (get_coverage_counts): Update calls to
344 dump_printf_loc to pass in dump_location_t rather than a
346 * doc/optinfo.texi (Dump types): Convert example of
347 dump_printf_loc from taking "locus" to taking "insn". Update
348 description of the "_loc" calls to cover dump_location_t.
349 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
351 (dump_user_location_t::dump_user_location_t): New constructors,
352 from gimple *stmt and rtx_insn *.
353 (dump_user_location_t::from_function_decl): New function.
354 (dump_loc): Make static.
355 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
356 const dump_location_t &.
357 (dump_generic_expr_loc): Delete.
358 (dump_printf_loc): Convert param "loc" from location_t to
359 const dump_location_t &.
360 (selftest::test_impl_location): New function.
361 (selftest::dumpfile_c_tests): New function.
362 * dumpfile.h: Include "profile-count.h".
363 (class dump_user_location_t): New class.
364 (struct dump_impl_location_t): New struct.
365 (class dump_location_t): New class.
366 (dump_printf_loc): Convert 2nd param from source_location to
367 const dump_location_t &.
368 (dump_generic_expr_loc): Delete.
369 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
370 const dump_location_t &.
371 * gimple-fold.c (fold_gimple_assign): Update call to
372 dump_printf_loc to pass in a dump_location_t rather than a
373 location_t, via the gimple stmt.
374 (gimple_fold_call): Likewise.
375 * gimple-loop-interchange.cc
376 (loop_cand::analyze_iloop_reduction_var): Update for change to
377 check_reduction_path.
378 (tree_loop_interchange::interchange): Update for change to
380 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
381 change in return-type of find_loop_location.
382 (graphite_regenerate_ast_isl): Likewise.
383 * graphite-optimize-isl.c (optimize_isl): Likewise.
384 * graphite.c (graphite_transform_loops): Likewise.
385 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
386 pass in a dump_location_t rather than a location_t, via the
388 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
389 * ipa.c (walk_polymorphic_call_targets): Likewise.
390 * loop-unroll.c (report_unroll): Convert "locus" param from
391 location_t to dump_location_t.
392 (decide_unrolling): Update for change to get_loop_location's
394 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
395 location_t to dump_user_location_t.
396 (grid_find_single_omp_among_assignments_1): Updates calls to
397 dump_printf_loc to pass in a dump_location_t rather than a
398 location_t, via the gimple stmt.
399 (grid_parallel_clauses_gridifiable): Convert "tloc" from
400 location_t to dump_location_t. Updates calls to dump_printf_loc
401 to pass in a dump_location_t rather than a location_t, via the
403 (grid_inner_loop_gridifiable_p): Likewise.
404 (grid_dist_follows_simple_pattern): Likewise.
405 (grid_gfor_follows_tiling_pattern): Likewise.
406 (grid_target_follows_gridifiable_pattern): Likewise.
407 (grid_attempt_target_gridification): Convert initialization
408 of local "grid" from memset to zero-initialization; FIXME: does
409 this require C++11? Update call to dump_printf_loc to pass in a
410 optinfo_location rather than a location_t, via the gimple stmt.
411 * profile.c (read_profile_edge_counts): Updates call to
412 dump_printf_loc to pass in a dump_location_t rather than a
414 (compute_branch_probabilities): Likewise.
415 * selftest-run-tests.c (selftest::run_tests): Call
417 * selftest.h (dumpfile_c_tests): New decl.
418 * tree-loop-distribution.c (pass_loop_distribution::execute):
419 Update for change in return type of find_loop_location.
420 * tree-parloops.c (parallelize_loops): Likewise.
421 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
422 "locus" from location_t to dump_user_location_t.
423 (canonicalize_loop_induction_variables): Likewise.
424 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
425 for change in return type of find_loop_location.
426 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
427 to dump_printf_loc to pass in a dump_location_t rather than a
428 location_t, via the stmt.
429 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
431 * tree-vect-loop-manip.c (find_loop_location): Convert return
432 type from source_location to dump_user_location_t.
433 (vect_do_peeling): Update for above change.
434 (vect_loop_versioning): Update for change in type of
436 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
437 from location_t to dump_user_location_t.
438 (vect_estimate_min_profitable_iters): Update for change in type
440 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
441 location_t to dump_location_t.
442 (vect_slp_bb): Update for change in type of vect_location.
443 * tree-vectorizer.c (vect_location): Convert from source_location
444 to dump_user_location_t.
445 (try_vectorize_loop_1): Update for change in vect_location's type.
446 (vectorize_loops): Likewise.
447 (increase_alignment): Likewise.
448 * tree-vectorizer.h (vect_location): Convert from source_location
449 to dump_user_location_t.
450 (find_loop_location): Convert return type from source_location to
451 dump_user_location_t.
452 (check_reduction_path): Convert 1st param from location_t to
453 dump_user_location_t.
454 * value-prof.c (check_counter): Update call to dump_printf_loc to
455 pass in a dump_user_location_t rather than a location_t; update
456 call to error_at for change in type of "locus".
457 (check_ic_target): Update call to dump_printf_loc to
458 pass in a dump_user_location_t rather than a location_t, via the
461 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
463 * config/s390/s390.h (enum processor_flags): Do not use
464 default tune parameter when -march was specified.
466 2018-06-26 Jakub Jelinek <jakub@redhat.com>
469 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
470 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
473 2018-06-26 Richard Biener <rguenther@suse.de>
475 PR tree-optimization/86287
477 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
478 (vect_analyze_loop): Initialize n_stmts.
480 2018-06-26 Richard Biener <rguenther@suse.de>
483 * fold-const.c (fold_convertible_p): Pointer extension
486 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
489 * dwarf2out.c (loc_list_has_views): Adjust comments.
490 (dw_loc_list): Split single cross-partition range with
493 2018-06-25 Jeff Law <law@redhat.com>
495 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
496 on -mbig-switch by default.
498 * config/v850/predicates.md (const_float_1_operand): Fix match_code
500 (const_float_0_operand): Remove unused predicate.
501 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
502 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
503 (recipsf2): New expander. Original pattern now called
505 (recipdf2, recipdf2_insn): Similarly.
506 (rsqrtsf2, rsqrtsf2_insn): Similarly
507 (rsqrtdf2, rsqrtdf2_insn): Similarly
509 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
511 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
512 Simplify logic for FreeBSD (twice).
514 2018-06-25 Martin Sebor <msebor@redhat.com>
516 PR tree-optimization/86204
517 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
518 a strnlen result if it's less than the length of the string.
520 2018-06-25 Martin Sebor <msebor@redhat.com>
522 PR tree-optimization/85700
523 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
524 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
525 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
527 2018-06-25 Martin Sebor <msebor@redhat.com>
529 * doc/extend.texi (Zero-length arrays): Update and clarify.
531 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
533 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
534 added IEEE/IBM long double multilib support on PowerPC little
535 endian Linux systems.
536 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
537 (MULTILIB_DEFAULTS): Likewise.
538 * config/rs6000/rs6000.c (rs6000_option_override_internal):
540 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
541 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
542 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
544 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
547 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
548 (REORDER_45): Likewise.
550 2018-06-25 Jeff Law <law@redhat.com>
552 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
553 dividend to 32 bits. Adjust length.
554 (udivmodhi4): Cleanup output template. Fix length.
556 2018-06-25 Carl Love <cel@us.ibm.com>
558 * config/rs6000/vsx.md: Change word selector to prefered location.
560 2018-06-25 Richard Biener <rguenther@suse.de>
562 PR tree-optimization/86304
563 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
564 epilogue-if-converted loops as well.
566 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
568 * lto-section-out.c (lto_begin_section): Do not print section
569 name for noaddr and unnumbered dumps.
571 2018-06-25 Richard Biener <rguenther@suse.de>
573 * tree-vectorizer.h (struct vec_info_shared): New structure
574 with parts split out from struct vec_info and loop_nest from
575 struct _loop_vec_info.
576 (struct vec_info): Adjust accordingly.
577 (struct _loop_vec_info): Likewise.
578 (LOOP_VINFO_LOOP_NEST): Adjust.
579 (LOOP_VINFO_DATAREFS): Likewise.
580 (LOOP_VINFO_DDRS): Likewise.
581 (struct _bb_vec_info): Likewise.
582 (BB_VINFO_DATAREFS): Likewise.
583 (BB_VINFO_DDRS): Likewise.
584 (struct _stmt_vec_info): Add dr_aux member.
585 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
586 (DR_MISALIGNMENT_UNINITIALIZED): New.
587 (set_dr_misalignment): Adjust.
588 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
589 (vect_analyze_loop): Adjust prototype.
590 (vect_analyze_loop_form): Likewise.
591 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
592 Compute dependences lazily.
593 (vect_record_base_alignments): Use shared datarefs/ddrs.
594 (vect_verify_datarefs_alignment): Likewise.
595 (vect_analyze_data_refs_alignment): Likewise.
596 (vect_analyze_data_ref_accesses): Likewise.
597 (vect_analyze_data_refs): Likewise.
598 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
599 constructor parameter for shared part.
600 (vect_analyze_loop_form): Pass in shared part and adjust.
601 (vect_analyze_loop_2): Pass in storage for the number of
602 stmts. Move loop nest finding to the caller. Compute
604 (vect_analyze_loop): Pass in shared part.
605 (vect_transform_loop): Verify shared datarefs are unchanged.
606 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
607 constructor parameter for shared part.
608 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
609 (vect_slp_bb): Verify shared datarefs are unchanged before
611 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
613 (new_stmt_vec_info): Initialize DR_AUX misalignment to
614 DR_MISALIGNMENT_UNINITIALIZED.
615 * tree-vectorizer.c (vec_info::vec_info): Add constructor
616 parameter for shared part.
617 (vec_info::~vec_info): Adjust.
618 (vec_info_shared::vec_info_shared): New.
619 (vec_info_shared::~vec_info_shared): Likewise.
620 (vec_info_shared::save_datarefs): Likewise.
621 (vec_info_shared::check_datarefs): Likewise.
622 (try_vectorize_loop_1): Construct shared part live for analyses
623 of a single loop for multiple vector sizes.
624 * tree-parloops.c (gather_scalar_reductions): Adjust.
626 2018-06-25 Richard Biener <rguenther@suse.de>
628 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
629 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
630 (vect_analyze_data_refs): Remove similar code from here and
631 simplify accordingly.
633 2018-06-25 Richard Biener <rguenther@suse.de>
635 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
636 for reverse storage order accesses rather than asserting
637 they cannot happen here.
639 2018-06-25 Tom de Vries <tdevries@suse.de>
642 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
643 Use data16 instead of .byte for insn prefix.
645 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
648 * parser.c (make_char_string_pack): Pass this literal chars
649 through cpp_interpret_string.
650 (cp_parser_userdef_numeric_literal): Check the result of
651 make_char_string_pack.
653 2018-06-24 Maya Rashish <coypu@sdf.org>
655 * ginclude/stddef.h: Simplify conditions around avoiding
656 re-definition of __size_t.
658 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
660 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
661 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
663 2018-06-22 Maya Rashish <coypu@sdf.org>
665 * doc/invoke.texi (mno-fancy-math-387): Update for changes
666 made to OpenBSD and NetBSD through the years.
668 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
670 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
671 behavior of vec_pack (vector double, vector double) to match
672 behavior of vec_float2 (vector double, vector double).
674 2018-06-22 Olivier Hainque <hainque@adacore.com>
676 * gimplify.c (gimplify_function_tree): Prevent creation
677 of a trampoline for the address of the current function
678 passed to entry/exit instrumentation hooks.
680 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
683 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
686 2018-06-22 Martin Liska <mliska@suse.cz>
688 PR tree-optimization/86263
689 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
690 Bail out if is_enabled is false.
691 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
693 (jump_table_cluster::is_enabled): New function.
695 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
697 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
698 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
699 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
700 (lto_input_ts_binfo_tree_pointers): Likewise.
701 * tree-streamer-out.c (streamer_write_tree_bitfields,
702 write_ts_binfo_tree_pointers): Likewise.
703 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
705 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
707 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
709 2018-06-22 Martin Liska <mliska@suse.cz>
711 * symbol-summary.h (get): Make it pure and inline move
712 functionality from ::get function.
713 (get): Remove and inline into ::get and ::get_create.
714 (get_create): Move code from ::get function.
716 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
719 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
720 -x assembler-with-cpp.
722 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
724 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
725 _FILE_OFFSET_BITS=64 for C++.
727 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
729 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
730 conversion insn that shows up when pr85657-3.c is compiled using
731 IEEE 128-bit long double.
732 (neg<mode>2_internal): Use the correct mode to check whether the
733 mode is IBM extended.
734 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
735 multiply and divide external functions from being created more
738 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
740 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
742 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
743 the edge can be forwarded.
744 (cfg_layout_merge_blocks): Likewise.
746 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
748 * except.c (finish_eh_generation): Commit edge insertions only after
749 the EH edges have been redirected from post-landing to landing pads.
751 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
753 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
754 create_tmp_var_for to create the FRAME decl.
755 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
757 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
759 * tree-inline.c (copy_edges_for_bb): Minor tweak.
760 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
761 debug statement when resetting its value.
762 (expand_call_inline): Copy the locus of the call onto the assignment
763 of the return value, if any. Use local variable in more cases.
765 2018-06-21 Martin Liska <mliska@suse.cz>
767 * ipa-pure-const.c (propagate_nothrow): Use
768 funct_state_summaries->get.
769 (dump_malloc_lattice): Likewise.
770 (propagate_malloc): Likewise.
772 2018-06-21 Richard Biener <rguenther@suse.de>
774 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
775 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
776 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
778 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
779 BLOCK_ABSTRACT_ORIGIN unconditionally.
781 2018-06-21 David Malcolm <dmalcolm@redhat.com>
783 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
785 * ipa-reference.c (ipa_reference_c_finalize): Delete
786 ipa_ref_opt_sum_summaries and set it to NULL.
788 2018-06-21 Tom de Vries <tdevries@suse.de>
790 PR tree-optimization/85859
791 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
792 test with comment from bb_no_side_effects_p.
794 2018-06-21 Richard Biener <rguenther@suse.de>
796 PR tree-optimization/86232
797 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
798 max for constant niter.
800 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
802 * config/aarch64/aarch64-simd.md
803 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
805 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
807 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
808 Make opernads of the unspec commutative.
810 2018-06-21 Richard Biener <rguenther@suse.de>
812 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
813 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
814 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
815 (vect_analyze_data_ref_dependence): Re-order checks to deal with
817 (vect_record_base_alignments): Do not record base alignment
818 for gathers or scatters.
819 (vect_compute_data_ref_alignment): Drop return value that is always
820 true. Bail out early for gathers or scatters.
821 (vect_enhance_data_refs_alignment): Bail out early for gathers
823 (vect_find_same_alignment_drs): Likewise.
824 (vect_analyze_data_refs_alignment): Remove dead code.
825 (vect_slp_analyze_and_verify_node_alignment): Likewise.
826 (vect_analyze_data_refs): For possible gathers or scatters do
827 not create an alternate DR, just check their possible validity
828 and mark them. Adjust DECL_NONALIASED handling to not rely
830 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
831 update inits of gathers or scatters.
832 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
833 Also copy gather/scatter flag to pattern vinfo.
835 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
837 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
838 behavior of vec_packsu (vector unsigned long long, vector unsigned
839 long long) to match behavior of vec_packs with same signature.
841 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
842 Thomas Schwinge <thomas@codesourcery.com>
843 Cesar Philippidis <cesar@codesourcery.com>
845 * gimplify.c (gimplify_scan_omp_clauses): Add support for
846 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
847 (gimplify_adjust_omp_clauses): Likewise.
848 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
849 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
850 (gimplify_omp_target_update): Update handling of acc update and
852 * omp-low.c (install_var_field): Remove unused parameter
853 base_pointers_restrict.
854 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
855 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
857 (omp_target_base_pointers_restrict_p): Delete.
858 (scan_omp_target): Update call to scan_sharing_clauses.
859 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
861 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
862 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
863 (convert_local_omp_clauses): Likewise.
864 * tree-pretty-print.c (dump_omp_clause): Likewise.
865 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
867 (omp_clause_code_name): Likewise.
869 2018-06-20 Jakub Jelinek <jakub@redhat.com>
872 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
875 PR tree-optimization/86231
876 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
877 anti-range don't overwrite *vr0min before using it to compute *vr0max.
879 2018-06-20 Tom de Vries <tdevries@suse.de>
881 PR tree-optimization/86097
882 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
883 iv type if signedness of iv type is not the same as that of *nit.
885 2018-06-20 Jakub Jelinek <jakub@redhat.com>
887 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
888 EDGE_EH edges, verify they are all EDGE_EH.
890 2018-06-20 Maya Rashish <coypu@sdf.org>
892 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
894 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
896 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
897 * config/aarch64/aarch64.c (xgene1_tunings): Add
898 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
899 (aarch64_mode_valid_for_sched_fusion_p):
901 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
902 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
904 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
906 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
907 * config/aarch64/iterators.md (VQ2): New mode iterator.
909 2018-06-20 Martin Liska <mliska@suse.cz>
911 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
912 Change default ratio from 10 to 8.
914 2018-06-20 Martin Liska <mliska@suse.cz>
916 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
918 (bit_test_cluster::find_bit_tests): Likewise.
919 (switch_decision_tree::analyze_switch_statement): Find clusters.
920 * tree-switch-conversion.h (struct jump_table_cluster): Document
923 2018-06-20 Martin Liska <mliska@suse.cz>
925 * tree-switch-conversion.c (switch_conversion::collect):
926 Record m_uniq property.
927 (switch_conversion::expand): Bail out for special conditions.
928 (group_cluster::~group_cluster): New.
929 (group_cluster::group_cluster): Likewise.
930 (group_cluster::dump): Likewise.
931 (jump_table_cluster::emit): New.
932 (switch_decision_tree::fix_phi_operands_for_edges): New.
933 (struct case_node): Remove struct.
934 (jump_table_cluster::can_be_handled): New.
935 (case_values_threshold): Moved to header.
936 (reset_out_edges_aux): Likewise.
937 (jump_table_cluster::is_beneficial): New.
938 (bit_test_cluster::can_be_handled): Likewise.
939 (add_case_node): Remove.
940 (bit_test_cluster::is_beneficial): New.
941 (case_bit_test::cmp): New.
942 (bit_test_cluster::emit): New.
943 (expand_switch_as_decision_tree_p): Remove.
944 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
945 (fix_phi_operands_for_edge): Likewise.
946 (switch_decision_tree::analyze_switch_statement): New.
947 (compute_cases_per_edge): Move ...
948 (switch_decision_tree::compute_cases_per_edge): ... here.
949 (try_switch_expansion): Likewise.
950 (switch_decision_tree::try_switch_expansion): Likewise.
951 (record_phi_operand_mapping): Likewise.
952 (switch_decision_tree::record_phi_operand_mapping): Likewise.
953 (emit_case_decision_tree): Likewise.
954 (switch_decision_tree::emit): Likewise.
955 (balance_case_nodes): Likewise.
956 (switch_decision_tree::balance_case_nodes): Likewise.
957 (dump_case_nodes): Likewise.
958 (switch_decision_tree::dump_case_nodes): Likewise.
959 (emit_jump): Likewise.
960 (switch_decision_tree::emit_jump): Likewise.
961 (emit_cmp_and_jump_insns): Likewise.
962 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
963 (emit_case_nodes): Likewise.
964 (switch_decision_tree::emit_case_nodes): Likewise.
965 (conditional_probability): Remove.
966 * tree-switch-conversion.h (enum cluster_type): New.
968 (struct cluster): Likewise.
969 (cluster::cluster): Likewise.
970 (struct simple_cluster): Likewise.
971 (simple_cluster::simple_cluster): Likewise.
972 (struct group_cluster): Likewise.
973 (struct jump_table_cluster): Likewise.
974 (struct bit_test_cluster): Likewise.
975 (struct min_cluster_item): Likewise.
976 (struct case_tree_node): Likewise.
977 (case_tree_node::case_tree_node): Likewise.
978 (jump_table_cluster::case_values_threshold): Likewise.
979 (struct case_bit_test): Likewise.
980 (struct switch_decision_tree): Likewise.
981 (struct switch_conversion): Likewise.
982 (switch_decision_tree::reset_out_edges_aux): Likewise.
984 2018-06-20 Martin Liska <mliska@suse.cz>
986 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
987 (hoist_edge_and_branch_if_true): Likewise.
988 (expand_switch_using_bit_tests_p): Likewise.
989 (struct case_bit_test): Likewise.
990 (case_bit_test_cmp): Likewise.
991 (emit_case_bit_tests): Likewise.
992 (switch_conversion::switch_conversion): New class.
993 (struct switch_conv_info): Remove old struct.
994 (collect_switch_conv_info): More to ...
995 (switch_conversion::collect): ... this.
996 (check_range): Likewise.
997 (switch_conversion::check_range): Likewise.
998 (check_all_empty_except_final): Likewise.
999 (switch_conversion::check_all_empty_except_final): Likewise.
1000 (check_final_bb): Likewise.
1001 (switch_conversion::check_final_bb): Likewise.
1002 (create_temp_arrays): Likewise.
1003 (switch_conversion::create_temp_arrays): Likewise.
1004 (free_temp_arrays): Likewise.
1005 (gather_default_values): Likewise.
1006 (switch_conversion::gather_default_values): Likewise.
1007 (build_constructors): Likewise.
1008 (switch_conversion::build_constructors): Likewise.
1009 (constructor_contains_same_values_p): Likewise.
1010 (switch_conversion::contains_same_values_p): Likewise.
1011 (array_value_type): Likewise.
1012 (switch_conversion::array_value_type): Likewise.
1013 (build_one_array): Likewise.
1014 (switch_conversion::build_one_array): Likewise.
1015 (build_arrays): Likewise.
1016 (switch_conversion::build_arrays): Likewise.
1017 (gen_def_assigns): Likewise.
1018 (switch_conversion::gen_def_assigns): Likewise.
1019 (prune_bbs): Likewise.
1020 (switch_conversion::prune_bbs): Likewise.
1021 (fix_phi_nodes): Likewise.
1022 (switch_conversion::fix_phi_nodes): Likewise.
1023 (gen_inbound_check): Likewise.
1024 (switch_conversion::gen_inbound_check): Likewise.
1025 (process_switch): Use the newly created class.
1026 (switch_conversion::expand): New.
1027 (switch_conversion::~switch_conversion): New.
1028 * tree-switch-conversion.h: New file.
1030 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1032 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1033 tree-vect-patterns.c.
1034 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1035 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1036 (vect_recog_sad_pattern): Likewise.
1037 (vect_recog_widen_sum_pattern): Likewise.
1038 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1039 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1040 (vect_recog_rotate_pattern): Likewise.
1041 (vect_recog_mult_pattern): Likewise.
1042 (vect_recog_vector_vector_shift_pattern): Likewise.
1043 (vect_recog_divmod_pattern): Likewise.
1044 (vect_recog_mixed_size_cond_pattern): Likewise.
1045 (vect_recog_bool_pattern): Likewise.
1046 (vect_recog_mask_conversion_pattern): Likewise.
1047 (vect_try_gather_scatter_pattern): Likewise.
1048 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1049 (vect_recog_over_widening_pattern): Likewise.
1050 (vect_recog_gather_scatter_pattern): Likewise.
1051 (vect_recog_func_ptr): Move from tree-vectorizer.h
1052 (vect_vect_recog_func_ptrs): Move further down the file.
1053 (vect_recog_func): Likewise. Remove the third argument.
1054 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1055 (vect_pattern_recog_1): Expect the pattern function to do any
1056 necessary target tests. Also expect it to provide a vector type.
1057 Remove the type_in handling.
1059 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1061 * tree-vect-patterns.c (vect_pattern_detected): New function.
1062 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1063 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1064 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1065 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1066 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1067 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1068 (vect_recog_mask_conversion_pattern)
1069 (vect_try_gather_scatter_pattern): Likewise.
1071 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1073 * tree-vect-patterns.c (vect_get_internal_def): New function.
1074 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1075 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1076 (search_type_for_mask_1): Use it.
1078 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1080 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1081 redundant WIDEN_SUM_EXPR handling.
1082 (vect_recog_sad_pattern): Likewise.
1084 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1086 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1087 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1088 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1089 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1090 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1092 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1094 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1095 use the stmt_vec_info of the original bb statement for the
1096 new zero assignment, even if the call is part of a pattern.
1098 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1100 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1101 that the sequence is attached to the original statement rather
1102 than the pattern statement.
1103 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1104 PATTERN_DEF_SEQ from the original statement rather than
1105 the main pattern statement.
1106 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1107 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1108 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1110 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1112 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1113 definition statements before the early exit for statements that aren't
1115 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1117 (vect_transform_loop): ...here. Process pattern definition
1118 statements without first checking whether the main pattern
1119 statement is live or relevant.
1121 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1123 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1124 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1126 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1128 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1129 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1130 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1131 (expand_block_compare): Change select_block_compare_mode call.
1132 (expand_strncmp_align_check): Use new functions, fix comment.
1133 (emit_final_str_compare_gpr): New function.
1134 (expand_strn_compare): Refactor and clean up code.
1135 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1137 2018-06-19 Tony Reix <tony.reix@atos.com>
1138 Damien Bergamini <damien.bergamini@atos.com>
1139 David Edelsohn <dje.gcc@gmail.com>
1141 * collect2.c (static_obj): New variable.
1142 (static_libs): New variable.
1143 (is_in_list): Uncomment declaration.
1144 (main): Track AIX libraries linked statically.
1145 (is_in_list): Uncomment definition.
1146 (scan_prog_file): Don't add AIX shared libraries initializer
1147 to constructor list if linking statically.
1149 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
1151 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
1153 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
1155 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
1157 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
1160 2018-06-19 Martin Liska <mliska@suse.cz>
1162 * config/i386/i386.c (ix86_can_inline_p): Do not use
1163 ipa_fn_summaries::get_create.
1164 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
1166 (devirtualization_time_bonus): Likewise.
1167 (ipcp_propagate_stage): Likewise.
1168 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1169 (edge_set_predicate): Likewise.
1170 (evaluate_conditions_for_known_args): Likewise.
1171 (evaluate_properties_for_edge): Likewise.
1172 (ipa_call_summary::reset): Tranform to ...
1173 (ipa_call_summary::~ipa_call_summary): ... this.
1174 (ipa_fn_summary::reset): Transform to ...
1175 (ipa_fn_summary::~ipa_fn_summary): ... this.
1176 (ipa_fn_summary_t::remove): Rename to ...
1177 (ipa_fn_summary_t::remove_callees): ... this.
1178 (ipa_fn_summary_t::duplicate): Use placement new
1179 instead of memory copy.
1180 (ipa_call_summary_t::duplicate): Likewise.
1181 (ipa_call_summary_t::remove): Remove.
1182 (dump_ipa_call_summary): Change get_create to get.
1183 (ipa_dump_fn_summary): Dump only when summary exists.
1184 (analyze_function_body): Use symbol_summary::get instead
1186 (compute_fn_summary): Likewise.
1187 (estimate_edge_devirt_benefit): Likewise.
1188 (estimate_edge_size_and_time): Likewise.
1189 (inline_update_callee_summaries): Likewise.
1190 (remap_edge_change_prob): Likewise.
1191 (remap_edge_summaries): Likewise.
1192 (ipa_merge_fn_summary_after_inlining): Likewise.
1193 (write_ipa_call_summary): Likewise.
1194 (ipa_fn_summary_write): Likewise.
1195 (ipa_free_fn_summary): Likewise.
1196 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
1197 (struct ipa_call_summary): Likewise.
1198 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
1200 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
1201 (estimate_size_after_inlining): Likewise.
1202 (estimate_growth): Likewise.
1203 (growth_likely_positive): Likewise.
1204 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1205 (inline_call): Likewise.
1206 * ipa-inline.c (caller_growth_limits): Likewise.
1207 (can_inline_edge_p): Likewise.
1208 (can_inline_edge_by_limits_p): Likewise.
1209 (compute_uninlined_call_time): Likewise.
1210 (compute_inlined_call_time): Likewise.
1211 (want_inline_small_function_p): Likewise.
1212 (edge_badness): Likewise.
1213 (update_caller_keys): Likewise.
1214 (update_callee_keys): Likewise.
1215 (inline_small_functions): Likewise.
1216 (inline_to_all_callers_1): Likewise.
1217 (dump_overall_stats): Likewise.
1218 (early_inline_small_functions): Likewise.
1219 (early_inliner): Likewise.
1220 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1221 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1222 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1223 * ipa-split.c (execute_split_functions): Likewise.
1224 * symbol-summary.h: Likewise.
1225 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1227 2018-06-19 Richard Biener <rguenther@suse.de>
1229 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
1230 (vectorize_loops): ... here. Fix dbgcnt handling.
1231 (try_vectorize_loop): Wrap try_vectorize_loop_1.
1233 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
1236 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
1237 ieee128 argument takes up only one (vector) register, not two (floating
1240 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1242 * gimplify.c (gimplify_init_constructor): Really never clear for an
1243 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
1245 2018-06-19 Richard Biener <rguenther@suse.de>
1247 PR tree-optimization/86179
1248 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
1249 after failed recognition.
1251 2018-06-18 Martin Sebor <msebor@redhat.com>
1254 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
1255 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
1256 Handle integer subtraction.
1257 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
1258 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
1260 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1262 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
1263 param from rtx to rtx_insn *.
1264 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
1266 (frv_ifcvt_modify_insn): Likwise.
1267 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
1268 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
1269 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
1270 as_a <rtx_insn *> cast to local "unprotected_region" once
1271 it's been established that it's not NULL or pc_rtx.
1272 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
1273 param "sethi" from rtx to rtx_insn *.
1274 (nds32_group_float_insns): Likewise for param "insn".
1275 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
1277 (vax_output_int_subtract): Likewise.
1278 * config/vax/vax.c (vax_output_int_add): Likewise for param
1280 (vax_output_int_subtract): Likewise.
1281 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
1282 (emit_pattern_after): Likewise for param "after".
1283 (emit_insn_after): Likewise.
1284 (emit_jump_insn_after): Likewise.
1285 (emit_call_insn_after): Likewise.
1286 (emit_debug_insn_after): Likewise.
1287 (emit_pattern_before): Likewise for param "before".
1288 (emit_insn_before): Likewise.
1289 (emit_jump_insn_before): Likewise.
1290 * final.c (get_insn_template): Likewise for param "insn", removing
1292 * output.h (get_insn_template): Likewise for 2nd param.
1293 * rtl.h (emit_insn_before): Likewise.
1294 (emit_jump_insn_before): Likewise.
1295 (emit_debug_insn_before_noloc): Likewise.
1296 (emit_insn_after): Likewise.
1297 (emit_jump_insn_after): Likewise.
1298 (emit_call_insn_after): Likewise.
1299 (emit_debug_insn_after): Likewise.
1300 (set_insn_deleted): Likewise for param.
1302 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
1305 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
1306 floating point modes, so that IFmode is numerically greater than
1307 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
1308 to declare the ordering. This prevents IFmode from being
1309 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
1310 machine. Include rs6000-modes.h to share the fractional values
1311 between genmodes* and the rest of the compiler.
1315 * config/rs6000/rs6000-modes.h: New file.
1316 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
1317 meaning of rs6000_long_double_size so that 126..128 selects an
1318 appropriate 128-bit floating point type.
1319 (rs6000_option_override_internal): Likewise.
1320 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
1321 (TARGET_LONG_DOUBLE_128): Change the meaning of
1322 rs6000_long_double_size so that 126..128 selects an appropriate
1323 128-bit floating point type.
1324 (LONG_DOUBLE_TYPE_SIZE): Update comment.
1325 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
1326 source and destination to match the standard usage.
1327 (truncifkf2): Likewise.
1328 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
1329 ISA 2.07 to use an explicit clobber, instead of passing in a
1331 (copysign<mode>3_soft): Likewise.
1333 2018-06-18 David Malcolm <dmalcolm@redhat.com>
1335 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1336 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
1337 (vect_slp_analyze_instance_dependence): Likewise.
1338 (vect_enhance_data_refs_alignment): Likewise.
1339 (vect_analyze_data_refs_alignment): Likewise.
1340 (vect_slp_analyze_and_verify_instance_alignment
1341 (vect_analyze_data_ref_accesses): Likewise.
1342 (vect_prune_runtime_alias_test_list): Likewise.
1343 (vect_analyze_data_refs): Likewise.
1344 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
1345 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
1346 (vect_analyze_scalar_cycles_1): Likewise.
1347 (vect_get_loop_niters): Likewise.
1348 (vect_analyze_loop_form_1): Likewise.
1349 (vect_update_vf_for_slp): Likewise.
1350 (vect_analyze_loop_operations): Likewise.
1351 (vect_analyze_loop): Likewise.
1352 (vectorizable_induction): Likewise.
1353 (vect_transform_loop): Likewise.
1354 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
1355 * tree-vect-slp.c (vect_analyze_slp): Likewise.
1356 (vect_make_slp_decision): Likewise.
1357 (vect_detect_hybrid_slp): Likewise.
1358 (vect_slp_analyze_operations): Likewise.
1359 (vect_slp_bb): Likewise.
1360 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
1361 (vectorizable_bswap): Likewise.
1362 (vectorizable_call): Likewise.
1363 (vectorizable_simd_clone_call): Likewise.
1364 (vectorizable_conversion): Likewise.
1365 (vectorizable_assignment): Likewise.
1366 (vectorizable_shift): Likewise.
1367 (vectorizable_operation): Likewise.
1368 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
1370 2018-06-18 Martin Sebor <msebor@redhat.com>
1372 PR tree-optimization/81384
1373 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
1374 * builtins.c (expand_builtin_strnlen): New function.
1375 (expand_builtin): Call it.
1376 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
1377 * builtins.def (BUILT_IN_STRNLEN): New.
1378 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
1379 Warn for bounds in excess of maximum object size.
1380 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
1381 single-value ranges. Handle strnlen.
1382 (handle_builtin_strlen): Handle strnlen.
1383 (strlen_check_and_optimize_stmt): Same.
1384 * doc/extend.texi (Other Builtins): Document strnlen.
1386 2018-06-18 Maya Rashish <coypu@sdf.org>
1388 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
1389 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
1390 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
1392 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
1394 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
1396 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1398 * tree.c (escaped_string::escape): Replace cast to char * by
1399 const_cast<char *> (unescaped).
1401 2018-06-18 Nick Clifton <nickc@redhat.com>
1404 * tree.c (escaped_string): New class. Converts an unescaped
1405 string into its escaped equivalent.
1406 (warn_deprecated_use): Use the new class to convert the
1407 deprecation message, if present.
1408 (test_escaped_strings): New self test.
1409 (test_c_tests): Add test_escaped_strings.
1410 * doc/extend.texi (deprecated): Add a note that the
1411 deprecation message is affected by the -fmessage-length
1412 option, and that control characters will be escaped.
1413 (#pragma GCC error): Document this pragma.
1414 (#pragma GCC warning): Likewise.
1415 * doc/invoke.texi (-fmessage-length): Document this option's
1416 effect on the #warning and #error preprocessor directives and
1417 the deprecated attribute.
1419 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
1421 * tree.c (decl_value_expr_lookup): Revert latest change.
1422 (decl_value_expr_insert): Likewise.
1424 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
1426 * gimplify.c (nonlocal_vlas): Delete.
1427 (nonlocal_vla_vars): Likewise.
1428 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
1430 (gimplify_body): Do not create and destroy nonlocal_vlas.
1431 * tree-nested.c: Include diagnostic.h.
1432 (use_pointer_in_frame): Tweak.
1433 (lookup_field_for_decl): Add assertion and declare the transformation.
1434 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
1435 internal error when the reference is in a wrong context. Do not
1436 create a debug decl by default.
1437 (note_nonlocal_block_vlas): Delete.
1438 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
1439 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
1440 create a debug decl by default.
1441 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
1442 call is in a wrong context.
1443 (fixup_vla_decls): New function.
1444 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
1445 debug variables were created.
1446 * tree.c (decl_value_expr_lookup): Add checking assertion.
1447 (decl_value_expr_insert): Likewise.
1449 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1452 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
1453 * tree-scalar-evolution.c (interpret_expr): Likewise.
1454 (expression_expensive_p): Likewise.
1455 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
1456 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
1457 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
1458 (ssa_defined_by_minus_one_stmt_p): New.
1460 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
1463 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
1464 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
1465 * dojump.c (do_jump): Likewise.
1466 * expr.c (expand_expr_real_2): Check operand type's sign.
1467 * fold-const.c (const_unop): Handle ABSU_EXPR.
1468 (fold_abs_const): Likewise.
1469 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
1470 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
1471 (strip_sign_op_1): Likesise.
1472 * match.pd: Add new pattern to generate ABSU_EXPR.
1473 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
1474 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1475 * tree-eh.c (operation_could_trap_helper_p): Likewise.
1476 * tree-inline.c (estimate_operator_cost): Likewise.
1477 * tree-pretty-print.c (dump_generic_node): Likewise.
1478 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
1479 * tree.def (ABSU_EXPR): New.
1481 2018-06-16 Jakub Jelinek <jakub@redhat.com>
1484 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
1485 documented as preserved for backward compatibility only.
1486 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
1488 PR rtl-optimization/86108
1489 * bb-reorder.c (create_forwarder_block): Renamed to ...
1490 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
1491 jump from new landing pad to the second part.
1492 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
1495 2018-06-15 Jakub Jelinek <jakub@redhat.com>
1498 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
1499 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
1500 Only call store_expr for halves if the mode is the same.
1503 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
1504 Fix up comment formatting.
1506 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
1508 * typed-splay-tree.h (typed_splay_tree::remove): New function.
1509 (typed_splay_tree::closure,
1510 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
1511 (typed_splay_tree::typed_splay_tree,
1512 typed_splay_tree::operator =): Declared private.
1513 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
1514 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
1515 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
1516 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
1517 typed_splay_tree::splay_tree_splay,
1518 typed_splay_tree::splay_tree_foreach_helper,
1519 typed_splay_tree::splay_tree_insert,
1520 typed_splay_tree::splay_tree_remove,
1521 typed_splay_tree::splay_tree_lookup,
1522 typed_splay_tree::splay_tree_predecessor,
1523 typed_splay_tree::splay_tree_successor,
1524 typed_splay_tree::splay_tree_min,
1525 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1526 (typed_splay_tree::root, typed_splay_tree::comp,
1527 typed_splay_tree::delete_key,
1528 typed_splay_tree::delete_value): New data members.
1529 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1530 typed_splay_tree::remove.
1532 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1534 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1535 -mginv and -mno-ginv to the assembler.
1536 * config/mips/mips.opt (-mcrc): New option.
1538 * doc/invoke.text (-mcrc): Document.
1541 2018-06-15 Nick Clifton <nickc@redhat.com>
1544 * tree.c (escaped_string): New class. Converts an unescaped
1545 string into its escaped equivalent.
1546 (warn_deprecated_use): Use the new class to convert the
1547 deprecation message, if present.
1548 (test_escaped_strings): New self test.
1549 (test_c_tests): Add test_escaped_strings.
1550 * doc/extend.texi (deprecated): Add a note that the
1551 deprecation message is affected by the -fmessage-length
1552 option, and that control characters will be escaped.
1553 (#pragma GCC error): Document this pragma.
1554 (#pragma GCC warning): Likewise.
1555 * doc/invoke.texi (-fmessage-length): Document this option's
1556 effect on the #warning and #error preprocessor directives and
1557 the deprecated attribute.
1559 2018-06-15 Richard Biener <rguenther@suse.de>
1561 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1562 here, also noting vector size used.
1563 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1564 size used in MSG_OPTIMIZED_LOCATIONS dump.
1565 (pass_slp_vectorize::execute): Adjust.
1567 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1570 * config/arc/arc.c (arc_return_address_register): Fix
1573 2018-06-15 Richard Biener <rguenther@suse.de>
1576 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1577 leave useless conversion stripping to force_gimple_operand_gsi.
1578 (gimplify_build2): Likewise.
1579 (gimplify_build1): Likewise.
1581 2018-06-15 Richard Biener <rguenther@suse.de>
1584 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1585 before adjusting their block.
1587 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1589 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1590 multilibs for *-*-rtems*.
1591 * config/riscv/t-rtems: New file.
1593 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1596 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1597 unsigned_type_for returns NULL.
1600 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1601 subregs of multi-word pseudos unless the float mode has word size.
1603 2018-06-14 Richard Biener <rguenther@suse.de>
1606 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1607 duplicate and harmful type_hash_canon.
1608 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1610 2018-06-14 Richard Biener <rguenther@suse.de>
1613 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1616 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1618 * config/rtems.h (STDINT_LONG32): Define.
1620 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1621 Prachi Godbole <prachi.godbole@imgtec.com>
1623 * config/mips/mips-cpus.def: Define P6600.
1624 * config/mips/mips-tables.opt: Regenerate.
1625 * config/mips/mips.c (mips_ucbranch_type): New enum.
1626 (mips_rtx_cost_data): Add support for P6600.
1627 (mips_issue_rate): Likewise.
1628 (mips_multipass_dfa_lookahead): Likewise.
1629 (mips_avoid_hazard): Likewise.
1630 (mips_reorg_process_insns): Likewise.
1631 (mips_classify_branch_p6600): New function.
1632 * config/mips/mips.h (TUNE_P6600): New define.
1633 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1634 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1635 * config/mips/mips.md: Include p6600.md.
1636 (processor): Add p6600.
1637 * config/mips/p6600.md: New file.
1638 * doc/invoke.texi: Add p6600 to supported architectures.
1640 2018-06-13 Martin Sebor <msebor@redhat.com>
1642 PR tree-optimization/86114
1643 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1645 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1647 2018-06-13 Richard Biener <rguenther@suse.de>
1649 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1650 Properly set vector type of the intermediate stmt.
1651 * tree-vect-stmts.c (vectorizable_operation): The destination
1652 var always has vectype_out type.
1654 2018-06-13 Jeff Law <law@redhat.com>
1656 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1657 integer 0 for argument to print_rtl_with_bb.
1658 (rl78_reorg): Likewise.
1660 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1662 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1663 from rtx to rtx_insn *.
1664 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1666 (add_sched_insns_for_speculation): Likewise for local "target",
1667 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1668 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1669 from rtx_insn ** to rtx_code_label **.
1670 (reorg_emit_nops): Likewise.
1671 (c6x_reorg): Likewise for local "call_labels".
1672 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1674 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1675 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1676 the loops over LABEL_REFS.
1677 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1679 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1680 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1681 (split_branches): Strengthen local "olabel" from rtx to
1682 rtx_insn *, adding a safe_as_a cast.
1683 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1685 (add_insn_after): Likewise for first two params.
1686 (add_insn_before): Likewise.
1687 (remove_insn): Likewise for param.
1688 (emit_pattern_before_noloc): Likewise for second and third params.
1689 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1690 (emit_call_insn_before_noloc): Likewise.
1691 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1693 (emit_barrier_before): Likewise.
1694 (emit_label_before): Strengthen "label" param from "rtx" to
1695 "rtx_code_label *". Strengthen "before" param from "rtx" to
1697 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1699 (emit_pattern_after_noloc): Likewise.
1700 (emit_insn_after_noloc): Likewise.
1701 (emit_jump_insn_after_noloc): Likewise.
1702 (emit_call_insn_after_noloc): Likewise.
1703 (emit_debug_insn_after_noloc): Likewise.
1704 (emit_barrier_after): Likewise.
1705 (emit_label_after): Likewise for both params.
1706 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1707 "loc" param from "int" to "location_t".
1708 (emit_insn_after_setloc): Likewise.
1709 (emit_jump_insn_after_setloc): Likewise.
1710 (emit_call_insn_after_setloc): Likewise.
1711 (emit_debug_insn_after_setloc): Likewise.
1712 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1713 "loc" param from "int" to "location_t".
1714 (emit_pattern_before): Convert NULL_RTX to NULL.
1715 (emit_insn_before_setloc): Convert "loc" param from "int" to
1717 (emit_jump_insn_before_setloc): Likewise.
1718 (emit_call_insn_before_setloc): Likewise.
1719 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1720 rtx_insn *. Convert "loc" param from "int" to "location_t".
1721 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1722 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1723 Convert 3rd param from "int" to "location_t".
1724 (emit_barrier_before, emit_barrier_after, next_real_insn):
1725 Strengthen param from rtx to rtx_insn *.
1726 (emit_label_before): Strengthen 1st param from "rtx" to
1727 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1729 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1730 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1731 Strengthen 2nd param from "rtx" to "rtx_insn *".
1732 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1733 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1734 Likewise. Convert 3rd param from "int" to "location_t".
1735 (emit_label_after): Strengthen 1st param from "rtx" to
1737 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1739 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1740 from "rtx" to "rtx_insn *".
1742 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1744 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1745 bodies streamed in with -Q.
1746 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1747 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1748 * gimple-streamer-out.c: Include gimple-pretty-print.h
1749 (output_bb): Dump stmts streamed.
1750 * lto-section-out.c: Include print-tree.h
1751 (lto_begin_section): Dump sections created.
1752 (lto_output_decl_index): Dump decl encoded.
1753 * lto-streamer-out.c: Include print-tree.h
1754 (create_output_block): Dump output block created.
1755 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1756 (output_function): Dump function output.
1757 (output_constructor): Dump constructor streamed.
1758 (write_global_stream): Output indexes encoded.
1759 (produce_asm_for_decls): Dump streams encoded.
1760 * lto-streamer.c (streamer_dump_file): New global var.
1761 * lto-streamer.h (streamer_dump_file): Declare.
1762 * passes.c (ipa_write_summaries): Initialize streamer dump.
1763 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1766 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1769 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1770 offsets for register save directives. Emit a second batch of save
1771 directives, if need be, when the function accesses prior frames.
1773 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1775 * config/arc/fpu.md (fmasf4): Force operand to register.
1776 (fnmasf4): Likewise.
1778 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1780 * config/arc/arc-protos.h (arc_pad_return): Remove.
1781 * config/arc/arc.c (machine_function): Remove force_short_suffix
1783 (arc_print_operand): Adjust printing of '&'.
1784 (arc_verify_short): Remove conditional printing of short suffix.
1785 (arc_final_prescan_insn): Remove reference to size_reason.
1786 (pad_return): New function.
1787 (arc_reorg): Call pad_return.
1788 (arc_pad_return): Remove.
1789 (arc_init_machine_status): Remove reference to force_short_suffix.
1790 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1791 (attr length): When attribute iscompact is true force to 2
1792 regardless; in the case of maybe check if we want to force the
1793 instruction to have 4 bytes length.
1794 (nopv): Change it to generate 4 byte long nop as well.
1795 (blockage): New pattern.
1796 (simple_return): Remove call to arc_pad_return.
1797 (p_return_i): Likewise.
1799 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1801 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1803 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1805 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1808 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1810 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1811 for ARC700 and ARCv2.
1813 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1816 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1817 operands[2] instead of operands[1].
1820 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1822 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1823 case, check whether the outer register overlaps an unallocatable
1824 register, not just whether it fits the required class.
1826 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1828 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1830 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1831 (memrefs_conflict_p): Likewise.
1832 (init_alias_analysis): Likewise.
1833 * cfgexpand.c (expand_debug_expr): Likewise.
1834 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1835 * cse.c (fold_rtx): Likewise.
1836 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1837 * expr.c (emit_block_move_hints): Likewise.
1838 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1839 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1840 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1841 (emit_group_store): Likewise.
1842 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1843 to read the PRE/POST_MODIFY increment.
1844 * calls.c (store_one_arg): Use strip_offset.
1845 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1847 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1849 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1850 (simplify_binary_operation_1): Extend CONST_INT handling to
1852 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1853 than a HOST_WIDE_INT.
1854 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1856 (adjust_mems, add_stores): Update accodingly.
1857 (vt_canonicalize_addr): Track polynomial offsets.
1858 (emit_note_insn_var_location): Likewise.
1859 (vt_add_function_parameter): Likewise.
1860 (vt_initialize): Likewise.
1862 2018-06-12 Jeff Law <law@redhat.com>
1864 * config.gcc (alpha*-*-freebsd*): Remove.
1865 * config/alpha/freebsd.h: Remove.
1867 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1870 * spellcheck-tree.c (levenshtein_distance): Rename to...
1871 (get_edit_distance): ...this, and update for underlying renaming.
1872 * spellcheck-tree.h (levenshtein_distance): Rename to...
1873 (get_edit_distance): ...this.
1874 * spellcheck.c (levenshtein_distance): Rename to...
1875 (get_edit_distance): ...this. Convert from Levenshtein distance
1876 to Damerau-Levenshtein distance by supporting transpositions of
1877 adjacent characters. Rename "v1" to "v_next" and "v0" to
1879 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1880 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1881 update for underlying renaming.
1882 (selftest::levenshtein_distance_unit_test): Rename to...
1883 (selftest::test_get_edit_distance_unit): ...this, and update for
1884 underlying renaming.
1885 (selftest::test_find_closest_string): Add example from PR 69968
1886 where transposition helps
1887 (selftest::test_metric_conditions): Update for renaming.
1888 (selftest::test_metric_conditions): Likewise.
1889 (selftest::spellcheck_c_tests): Likewise.
1890 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1891 (get_edit_distance): ...this.
1892 (best_match::consider): Update for renaming.
1894 2018-06-12 Martin Sebor <msebor@redhat.com>
1896 PR tree-optimization/85259
1897 * builtins.c (compute_objsize): Handle constant offsets.
1898 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1899 true iff a warning has been issued.
1900 * gimple.h (gimple_nonartificial_location): New function.
1901 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1902 gimple_nonartificial_location and handle -Wno-system-headers.
1903 (handle_builtin_stxncpy): Same.
1905 2018-06-12 Martin Sebor <msebor@redhat.com>
1908 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1910 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1912 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1913 BUILTIN_VEC_XST entries for pointer to double and long long.
1915 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1918 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1920 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1923 2018-06-12 Martin Liska <mliska@suse.cz>
1925 * doc/options.texi: Document IntegerRange.
1927 2018-06-12 Martin Liska <mliska@suse.cz>
1929 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1930 * opt-functions.awk: Handle Deprecated flag.
1931 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1933 (read_cmdline_option): Report warning for a deprecated option.
1934 * opts.h (struct cl_option): Add new field cl_deprecated.
1935 (CL_ERR_DEPRECATED): New.
1937 2018-06-12 Martin Liska <mliska@suse.cz>
1939 * doc/options.texi: Document Deprecated option flag.
1941 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1943 * config/arc/arc-arch.h (arc_extras): New enum.
1944 (arc_cpu_t):Add field extra.
1945 (arc_cpu_types): Consider the extras.
1946 * config/arc/arc-cpus.def: Add extras info.
1947 * config/arc/arc-opts.h (processor_type): Consider extra field.
1948 * config/arc/arc.c (arc_override_options): Handle extra field.
1950 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1952 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1953 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1955 * config/arc/arc.c (arc_init): Update pic warning.
1956 (irq_range): Update irq range parsing warnings.
1957 (arc_override_options): Update various warning messages.
1958 (arc_handle_aux_attribute): Likewise.
1960 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1962 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1964 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1966 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1967 compiler messages directives.
1969 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1971 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1972 * config/mips/mips-tables.opt: Regenerate.
1973 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1975 * doc/invoke.texi: Document -march=i6500.
1977 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1979 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1980 (i6400_gpmul): Add cpu_unit.
1981 (i6400_gpdiv): Likewise.
1982 (i6400_msa_add_d): Update reservations.
1983 (i6400_msa_int_add) Likewise.
1984 (i6400_msa_short_logic3) Likewise.
1985 (i6400_msa_short_logic2) Likewise.
1986 (i6400_msa_short_logic) Likewise.
1987 (i6400_msa_move) Likewise.
1988 (i6400_msa_cmp) Likewise.
1989 (i6400_msa_short_float2) Likewise.
1990 (i6400_msa_div_d) Likewise.
1991 (i6400_msa_long_logic1) Likewise.
1992 (i6400_msa_long_logic2) Likewise.
1993 (i6400_msa_mult) Likewise.
1994 (i6400_msa_long_float2) Likewise.
1995 (i6400_msa_long_float4) Likewise.
1996 (i6400_msa_long_float5) Likewise.
1997 (i6400_msa_long_float8) Likewise.
1998 (i6400_fpu_fadd): Include frint type.
1999 (i6400_fpu_store): New define_insn_reservation.
2000 (i6400_fpu_load): Likewise.
2001 (i6400_fpu_move): Likewise.
2002 (i6400_fpu_fcmp): Likewise.
2003 (i6400_fpu_fmadd): Likewise.
2004 (i6400_int_mult): Include imul3nc type and update reservation.
2005 (i6400_int_div): Include idiv3 type and update reservation.
2006 (i6400_int_load): Update to check type not move_type.
2007 (i6400_int_store): Likewise.
2008 (i6400_int_prefetch): Set zero latency.
2010 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2012 * gcc.c: Document new %@{...} sequence.
2013 (LINK_COMMAND_SPEC): Use it for the -L switches.
2014 (cpp_unique_options): Use it for the -I switches.
2015 (at_file_argbuf): New global variable.
2016 (in_at_file): Likewise.
2017 (alloc_args): Create at_file_argbuf.
2018 (clear_args): Truncate at_file_argbuf.
2019 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2020 (open_at_file): New function.
2021 (close_at_file): Likewise.
2022 (create_at_file): Delete.
2023 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2026 (validate_switches_from_spec): Deal with %@{...} sequence.
2027 (validate_switches): Likewise.
2028 (driver::finalize): Call clear_args.
2030 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2032 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2034 2018-06-11 Martin Sebor <msebor@redhat.com>
2036 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2037 (Wno-class-memaccess): Add @opindex.
2038 (Wno-templates, Wno-multiple-inheritance): Same.
2039 (Wno-virtual-inheritance, Wno-namespaces): Same.
2040 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2041 (Wno-format-overflow, Wno-format-truncation): Same.
2042 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2043 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2044 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2045 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2046 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2047 (Wno-misspelled-isr): Same.
2049 2018-06-11 Martin Sebor <msebor@redhat.com>
2051 * PR tree-optimization/86083
2052 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2054 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2056 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2058 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2061 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2062 on the correct operand.
2063 (*movdi_internal64): Ditto.
2065 2018-06-11 Martin Liska <mliska@suse.cz>
2067 PR tree-optimization/86089
2068 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2070 2018-06-11 Julia Koval <julia.koval@intel.com>
2072 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2073 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2074 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2076 2018-06-11 Olivier Hainque <hainque@adacore.com>
2078 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2079 for Ada with strict dwarf2.
2081 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2084 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2087 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2089 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2091 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2093 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2096 2018-06-08 Martin Liska <mliska@suse.cz>
2098 * tree-cfg.h (debug_function): Fix argument type to match
2101 2018-06-08 Martin Liska <mliska@suse.cz>
2103 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2104 Remove usage of MPX-related (and removed) fields.
2105 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2107 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2109 * cfg.c (debug): Use TDF_NONE rather than 0.
2110 * cfghooks.c (debug): Likewise.
2111 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2112 (struct dump_option_value_info): Convert to...
2113 (struct kv_pair): ...this template type.
2114 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2116 (optinfo_verbosity_options): Likewise.
2117 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2119 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2120 than int for "optgroup_flags" param.
2121 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2123 (dump_dec): Likewise.
2124 (dump_finish): Use TDF_NONE rather than 0.
2125 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2126 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2127 than 0. Update for change to option_ptr.
2128 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2129 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2130 0. Update for changes to optinfo_verbosity_options and
2132 (opt_info_switch_p): Convert optgroup_flags from int to
2134 (dump_basic_block): Use dump_flags_t rather than int
2135 for "dump_kind" param.
2136 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2137 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2138 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2139 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2140 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2141 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2142 TDF_NONE): Convert from macros to...
2143 (enum dump_flag): ...this new enum.
2144 (dump_flags_t): Update to use enum.
2145 (operator|, operator&, operator~, operator|=, operator&=):
2146 Implement for dump_flags_t.
2147 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2148 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2149 Convert from macros to...
2150 (enum optgroup_flag): ...this new enum.
2151 (optgroup_flags_t): New typedef.
2152 (operator|, operator|=): Implement for optgroup_flags_t.
2153 (struct dump_file_info): Convert field "alt_flags" to
2154 dump_flags_t. Convert field "optgroup_flags" to
2156 (dump_basic_block): Use dump_flags_t rather than int for param.
2157 (dump_generic_expr_loc): Likewise.
2158 (dump_dec): Likewise.
2159 (dump_register): Convert param "optgroup_flags" to
2161 (opt_info_enable_passes): Likewise.
2162 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
2164 * gimple-pretty-print.c (debug): Likewise.
2165 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
2166 (merged_store_group::apply_stores): Likewise.
2167 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
2168 * gimple.c (verify_gimple_pp): Likewise.
2169 * graphite-poly.c (print_pbb_body): Likewise.
2170 * passes.c (pass_manager::register_one_dump_file): Convert
2171 local "optgroup_flags" to optgroup_flags_t.
2172 * print-tree.c (print_node): Use TDF_NONE rather than 0.
2174 (debug_body): Likewise.
2175 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
2176 to optgroup_flags_t.
2177 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
2179 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
2180 (convert_mult_to_fma): Likewise.
2181 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
2182 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
2183 * tree-vect-data-refs.c (dump_lower_bound): Convert param
2184 "dump_kind" to dump_flags_t.
2186 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2188 * config/rs6000/rs6000.c (min, max): Delete.
2190 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
2192 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
2193 -mabi=spe and -mabi=no-spe.
2195 2018-06-08 Martin Liska <mliska@suse.cz>
2197 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
2198 where we expect an existing summary.
2200 2018-06-08 Martin Liska <mliska@suse.cz>
2202 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
2203 * ipa-inline.h (estimate_edge_growth): Likewise.
2205 2018-06-08 Martin Liska <mliska@suse.cz>
2207 * cgraph.c (function_version_hasher::hash): Use
2208 cgraph_node::get_uid ().
2209 (function_version_hasher::equal):
2210 * cgraph.h (cgraph_node::get_uid): New method.
2211 * ipa-inline.c (update_caller_keys): Use
2212 cgraph_node::get_uid ().
2213 (update_callee_keys): Likewise.
2214 * ipa-utils.c (searchc): Likewise.
2215 (ipa_reduced_postorder): Likewise.
2216 * lto-cgraph.c (input_node): Likewise.
2217 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
2218 * symbol-summary.h (symtab_insertion): Likewise.
2219 (symtab_removal): Likewise.
2220 (symtab_duplication): Likewise.
2221 * tree-pretty-print.c (dump_function_header): Likewise.
2222 * tree-sra.c (convert_callers_for_node): Likewise.
2224 2018-06-08 Martin Liska <mliska@suse.cz>
2226 * cgraph.c (symbol_table::create_edge): Always assign a new
2228 (symbol_table::free_edge): Do not recycle numbers.
2229 * cgraph.h (cgraph_edge::get): New method.
2230 * symbol-summary.h (symtab_removal): Use it.
2231 (symtab_duplication): Likewise.
2232 (call_summary::hashable_uid): Remove.
2234 2018-06-08 Martin Liska <mliska@suse.cz>
2236 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
2237 (initialize_growth_caches): Remove.
2238 (free_growth_caches): Likewise.
2239 (do_estimate_edge_time): Use edge_growth_cache.
2240 (do_estimate_edge_size): Likewise.
2241 (do_estimate_edge_hints): Likewise.
2242 * ipa-inline.c (reset_edge_caches): Likewise.
2243 (recursive_inlining): Likewise.
2244 (inline_small_functions): Likewise.
2245 * ipa-inline.h (initialize_growth_caches): Remove.
2246 (estimate_edge_size): Likewise.
2247 (estimate_edge_time): Likewise.
2248 (estimate_edge_hints): Likewise.
2249 (reset_edge_growth_cache): Likewise.
2250 * symbol-summary.h (call_summary::remove): New method.
2252 2018-06-08 Martin Liska <mliska@suse.cz>
2254 * ipa-cp.c (class edge_clone_summary): New summary.
2255 (grow_edge_clone_vectors): Remove.
2256 (ipcp_edge_duplication_hook): Remove.
2257 (class edge_clone_summary_t): New call_summary class.
2258 (ipcp_edge_removal_hook): Remove.
2259 (edge_clone_summary_t::duplicate): New function.
2260 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
2261 (create_specialized_node): Likewise.
2262 (ipcp_driver): Initialize edge_clone_summaries and do not
2265 2018-06-08 Martin Liska <mliska@suse.cz>
2267 * symbol-summary.h (get): New function.
2268 (call_summary::m_initialize_when_cloning): New class member.
2270 2018-06-08 Martin Liska <mliska@suse.cz>
2272 * cgraph.c (cgraph_node::remove): Do not recycle uid.
2273 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
2274 (symbol_table::allocate_cgraph_symbol): Do not set uid.
2275 * passes.c (uid_hash_t): Record removed_nodes by their uids.
2276 (remove_cgraph_node_from_order): Use the removed_nodes set.
2277 (do_per_function_toporder): Likwise.
2278 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
2279 instead of summary_uid.
2280 (symtab_removal): Likewise.
2281 (symtab_duplication): Likewise.
2283 2018-06-08 Martin Liska <mliska@suse.cz>
2285 * ipa-cp.c (ipcp_store_bits_results): Use
2286 ipcp_transformation_sum.
2287 (ipcp_store_vr_results): Likewise.
2288 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
2290 (ipcp_transformation_initialize): ... this.
2291 (ipa_set_node_agg_value_chain):
2292 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
2293 (write_ipcp_transformation_info): Likewise.
2294 (read_ipcp_transformation_info): Likewise.
2295 (ipcp_update_bits): Likewise.
2296 (ipcp_update_vr): Likewise.
2297 (ipcp_transform_function): Likewise.
2298 * ipa-prop.h: Rename ipcp_transformation_summary to
2299 ipcp_transformation.
2300 (class ipcp_transformation_t): New function summary.
2301 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
2302 (ipa_get_agg_replacements_for_node): Likewise.
2304 2018-06-08 Martin Liska <mliska@suse.cz>
2306 * ipa-pure-const.c (struct funct_state_d): Do it class instead
2308 (class funct_state_summary_t): New function_summary class.
2309 (has_function_state): Remove.
2310 (get_function_state): Likewise.
2311 (set_function_state): Likewise.
2312 (add_new_function): Likewise.
2313 (funct_state_summary_t::insert): New function.
2314 (duplicate_node_data): Remove.
2315 (remove_node_data): Remove.
2316 (funct_state_summary_t::duplicate): New function.
2317 (register_hooks): Create new funct_state_summaries.
2318 (pure_const_generate_summary): Use it.
2319 (pure_const_write_summary): Likewise.
2320 (pure_const_read_summary): Likewise.
2321 (propagate_pure_const): Likewise.
2322 (propagate_nothrow): Likewise.
2323 (dump_malloc_lattice): Likewise.
2324 (propagate_malloc): Likewise.
2325 (execute): Do not register hooks, just remove summary
2327 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
2330 2018-06-08 Martin Liska <mliska@suse.cz>
2332 * ipa-reference.c (remove_node_data): Remove.
2333 (duplicate_node_data): Likewise.
2334 (class ipa_ref_var_info_summary_t): New class.
2335 (class ipa_ref_opt_summary_t): Likewise.
2336 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
2337 (get_reference_optimization_summary): Use
2338 ipa_ref_opt_sum_summaries.
2339 (set_reference_vars_info): Remove.
2340 (set_reference_optimization_summary): Likewise.
2341 (ipa_init): Create summaries.
2342 (init_function_info): Use function summary.
2343 (ipa_ref_opt_summary_t::duplicate): New function.
2344 (ipa_ref_opt_summary_t::remove): New function.
2345 (get_read_write_all_from_node): Fix GNU coding style.
2346 (propagate): Use function summary.
2347 (write_node_summary_p): Fix GNU coding style.
2348 (stream_out_bitmap): Likewise.
2349 (ipa_reference_read_optimization_summary): Use function summary.
2350 (ipa_reference_c_finalize): Do not release hooks.
2352 2018-06-08 Martin Liska <mliska@suse.cz>
2354 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
2355 (analyze_function_body): Extract multiple calls of get_create.
2356 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2357 * ipa-inline.c (recursive_inlining): Use ::get method.
2358 * ipa-inline.h (estimate_edge_growth): Likewise.
2360 2018-06-08 Martin Liska <mliska@suse.cz>
2362 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
2364 (hsa_function_summary::hsa_function_summary): Use the new enum
2366 (hsa_gpu_implementation_p): Use hsa_summaries::get.
2367 * hsa-gen.c (hsa_get_host_function): Likewise.
2368 (get_brig_function_name): Likewise.
2369 * ipa-hsa.c (process_hsa_functions): Likewise.
2370 (ipa_hsa_write_summary): Likewise.
2371 * symbol-summary.h (symtab_duplication): Use ::get function/
2372 (get): New function.
2374 2018-06-08 Martin Liska <mliska@suse.cz>
2376 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
2378 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
2379 (hsa_register_kernel): Likewise.
2380 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
2381 * hsa-gen.c (hsa_get_host_function): Likewise.
2382 (get_brig_function_name): Likewise.
2383 (generate_hsa): Likewise.
2384 (pass_gen_hsail::execute): Likewise.
2385 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2386 (devirtualization_time_bonus): Likewise.
2387 (ipcp_propagate_stage): Likewise.
2388 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2389 (edge_set_predicate): Likewise.
2390 (evaluate_conditions_for_known_args): Likewise.
2391 (evaluate_properties_for_edge): Likewise.
2392 (ipa_fn_summary::reset): Likewise.
2393 (ipa_fn_summary_t::duplicate): Likewise.
2394 (dump_ipa_call_summary): Likewise.
2395 (ipa_dump_fn_summary): Likewise.
2396 (analyze_function_body): Likewise.
2397 (compute_fn_summary): Likewise.
2398 (estimate_edge_devirt_benefit): Likewise.
2399 (estimate_edge_size_and_time): Likewise.
2400 (estimate_calls_size_and_time): Likewise.
2401 (estimate_node_size_and_time): Likewise.
2402 (inline_update_callee_summaries): Likewise.
2403 (remap_edge_change_prob): Likewise.
2404 (remap_edge_summaries): Likewise.
2405 (ipa_merge_fn_summary_after_inlining): Likewise.
2406 (ipa_update_overall_fn_summary): Likewise.
2407 (read_ipa_call_summary): Likewise.
2408 (inline_read_section): Likewise.
2409 (write_ipa_call_summary): Likewise.
2410 (ipa_fn_summary_write): Likewise.
2411 (ipa_free_fn_summary): Likewise.
2412 * ipa-hsa.c (process_hsa_functions): Likewise.
2413 (ipa_hsa_write_summary): Likewise.
2414 (ipa_hsa_read_section): Likewise.
2415 * ipa-icf.c (sem_function::merge): Likewise.
2416 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
2417 (do_estimate_edge_time): Likewise.
2418 (estimate_size_after_inlining): Likewise.
2419 (estimate_growth): Likewise.
2420 (growth_likely_positive): Likewise.
2421 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2422 (inline_call): Likewise.
2423 * ipa-inline.c (caller_growth_limits): Likewise.
2424 (can_inline_edge_p): Likewise.
2425 (can_inline_edge_by_limits_p): Likewise.
2426 (compute_uninlined_call_time): Likewise.
2427 (compute_inlined_call_time): Likewise.
2428 (want_inline_small_function_p): Likewise.
2429 (edge_badness): Likewise.
2430 (update_caller_keys): Likewise.
2431 (update_callee_keys): Likewise.
2432 (recursive_inlining): Likewise.
2433 (inline_small_functions): Likewise.
2434 (inline_to_all_callers_1): Likewise.
2435 (dump_overall_stats): Likewise.
2436 (early_inline_small_functions): Likewise.
2437 (early_inliner): Likewise.
2438 * ipa-inline.h (estimate_edge_growth): Likewise.
2439 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2440 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2441 * ipa-prop.h (IPA_NODE_REF): Likewise.
2442 (IPA_EDGE_REF): Likewise.
2443 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2444 (propagate_malloc): Likewise.
2445 * ipa-split.c (execute_split_functions): Likewise.
2446 * symbol-summary.h: Rename get to get_create.
2448 (get_create): Likewise.
2449 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2451 2018-06-08 Martin Liska <mliska@suse.cz>
2453 * symbol-summary.h (release): Move definition out of class
2455 (symtab_removal): Likewise.
2456 (symtab_duplication): Likewise.
2458 2018-06-08 Martin Liska <mliska@suse.cz>
2460 * symbol-summary.h (function_summary): Move constructor
2461 implementation out of class declaration.
2462 (release): Likewise.
2463 (symtab_insertion): Likewise.
2464 (symtab_removal): Likewise.
2465 (symtab_duplication): Likewise.
2468 2018-06-08 Martin Liska <mliska@suse.cz>
2470 * Makefile.in: Remove support for MPX (macros, related functions,
2471 fields in cgraph_node, ...).
2472 * builtin-types.def (BT_BND): Likewise.
2473 (BT_FN_BND_CONST_PTR): Likewise.
2474 (BT_FN_CONST_PTR_BND): Likewise.
2475 (BT_FN_VOID_PTR_BND): Likewise.
2476 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
2477 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
2478 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
2479 (expand_builtin_mempcpy_with_bounds): Likewise.
2480 (expand_builtin_memset_with_bounds): Likewise.
2481 (expand_builtin_memset_args): Likewise.
2482 (std_expand_builtin_va_start): Likewise.
2483 (expand_builtin): Likewise.
2484 (expand_builtin_with_bounds): Likewise.
2485 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
2486 (DEF_LIB_BUILTIN_CHKP): Likewise.
2487 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
2488 (DEF_CHKP_BUILTIN): Likewise.
2489 (BUILT_IN_MEMCPY): Likewise.
2490 (BUILT_IN_MEMMOVE): Likewise.
2491 (BUILT_IN_MEMPCPY): Likewise.
2492 (BUILT_IN_MEMSET): Likewise.
2493 (BUILT_IN_STPCPY): Likewise.
2494 (BUILT_IN_STRCAT): Likewise.
2495 (BUILT_IN_STRCHR): Likewise.
2496 (BUILT_IN_STRCPY): Likewise.
2497 (BUILT_IN_STRLEN): Likewise.
2498 (BUILT_IN_MEMCPY_CHK): Likewise.
2499 (BUILT_IN_MEMMOVE_CHK): Likewise.
2500 (BUILT_IN_MEMPCPY_CHK): Likewise.
2501 (BUILT_IN_MEMSET_CHK): Likewise.
2502 (BUILT_IN_STPCPY_CHK): Likewise.
2503 (BUILT_IN_STRCAT_CHK): Likewise.
2504 (BUILT_IN_STRCPY_CHK): Likewise.
2505 * calls.c (store_bounds): Likewise.
2506 (emit_call_1): Likewise.
2507 (special_function_p): Likewise.
2508 (maybe_warn_nonstring_arg): Likewise.
2509 (initialize_argument_information): Likewise.
2510 (finalize_must_preallocate): Likewise.
2511 (compute_argument_addresses): Likewise.
2512 (expand_call): Likewise.
2513 * cfgexpand.c (expand_call_stmt): Likewise.
2514 (expand_return): Likewise.
2515 (expand_gimple_stmt_1): Likewise.
2516 (pass_expand::execute): Likewise.
2517 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2518 (cgraph_node::remove): Likewise.
2519 (cgraph_node::dump): Likewise.
2520 (cgraph_node::verify_node): Likewise.
2521 * cgraph.h (chkp_function_instrumented_p): Likewise.
2522 (symtab_node::get_alias_target): Likewise.
2523 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
2524 (cgraph_local_p): Likewise.
2525 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2526 (cgraph_edge::rebuild_references): Likewise.
2527 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2528 (walk_polymorphic_call_targets): Likewise.
2529 (cgraph_node::expand_thunk): Likewise.
2530 (symbol_table::output_weakrefs): Likewise.
2531 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2532 (ix86_handle_option): Likewise.
2533 * config/i386/constraints.md: Likewise.
2534 * config/i386/i386-builtin-types.def (BND): Likewise.
2538 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2539 (BDESC_FIRST): Likewise.
2541 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2542 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2543 * config/i386/i386.c (enum reg_class): Likewise.
2544 (ix86_target_string): Likewise.
2545 (ix86_option_override_internal): Likewise.
2546 (ix86_conditional_register_usage): Likewise.
2547 (ix86_valid_target_attribute_inner_p): Likewise.
2548 (ix86_set_indirect_branch_type): Likewise.
2549 (ix86_set_current_function): Likewise.
2550 (ix86_function_arg_regno_p): Likewise.
2551 (init_cumulative_args): Likewise.
2552 (ix86_function_arg_advance): Likewise.
2553 (ix86_function_arg): Likewise.
2554 (ix86_pass_by_reference): Likewise.
2555 (ix86_function_value_regno_p): Likewise.
2556 (ix86_function_value_1): Likewise.
2557 (ix86_function_value_bounds): Likewise.
2558 (ix86_return_in_memory): Likewise.
2559 (ix86_setup_incoming_vararg_bounds): Likewise.
2560 (ix86_va_start): Likewise.
2561 (indirect_thunk_need_prefix): Likewise.
2562 (print_reg): Likewise.
2563 (ix86_print_operand): Likewise.
2564 (ix86_expand_call): Likewise.
2565 (ix86_output_function_return): Likewise.
2566 (reg_encoded_number): Likewise.
2567 (BDESC_VERIFYS): Likewise.
2568 (ix86_init_mpx_builtins): Likewise.
2569 (ix86_init_builtins): Likewise.
2570 (ix86_emit_cmove): Likewise.
2571 (ix86_emit_move_max): Likewise.
2572 (ix86_expand_builtin): Likewise.
2573 (ix86_builtin_mpx_function): Likewise.
2574 (ix86_get_arg_address_for_bt): Likewise.
2575 (ix86_load_bounds): Likewise.
2576 (ix86_store_bounds): Likewise.
2577 (ix86_load_returned_bounds): Likewise.
2578 (ix86_store_returned_bounds): Likewise.
2579 (ix86_class_likely_spilled_p): Likewise.
2580 (ix86_hard_regno_mode_ok): Likewise.
2581 (x86_order_regs_for_local_alloc): Likewise.
2582 (ix86_mitigate_rop): Likewise.
2583 (ix86_bnd_prefixed_insn_p): Likewise.
2584 (ix86_mpx_bound_mode): Likewise.
2585 (ix86_make_bounds_constant): Likewise.
2586 (ix86_initialize_bounds): Likewise.
2587 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2588 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2589 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2590 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2591 (TARGET_CHKP_BOUND_MODE): Likewise.
2592 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2593 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2594 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2595 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2596 * config/i386/i386.h (TARGET_MPX): Likewise.
2597 (TARGET_MPX_P): Likewise.
2598 (VALID_BND_REG_MODE): Likewise.
2599 (FIRST_BND_REG): Likewise.
2600 (LAST_BND_REG): Likewise.
2601 (enum reg_class): Likewise.
2602 (BND_REG_P): Likewise.
2603 (BND_REGNO_P): Likewise.
2604 (BNDmode): Likewise.
2605 (ADJUST_INSN_LENGTH): Likewise.
2606 * config/i386/i386.md: Likewise.
2607 * config/i386/i386.opt: Likewise.
2608 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2609 (defined): Likewise.
2610 (LINK_MPX): Likewise.
2611 (MPX_SPEC): Likewise.
2612 (LIBMPX_SPEC): Likewise.
2613 (LIBMPXWRAPPERS_SPEC): Likewise.
2614 (CHKP_SPEC): Likewise.
2615 * config/i386/predicates.md: Likewise.
2616 * dbxout.c (dbxout_type): Likewise.
2617 * doc/extend.texi: Likewise.
2618 * doc/invoke.texi: Likewise.
2619 * doc/md.texi: Likewise.
2620 * doc/tm.texi: Likewise.
2621 * doc/tm.texi.in: Likewise.
2622 * dwarf2out.c (is_base_type): Likewise.
2623 (gen_formal_types_die): Likewise.
2624 (gen_subprogram_die): Likewise.
2625 (gen_type_die_with_usage): Likewise.
2626 (gen_decl_die): Likewise.
2627 (dwarf2out_late_global_decl): Likewise.
2628 * expr.c (expand_assignment): Likewise.
2629 (emit_storent_insn): Likewise.
2630 (store_expr_with_bounds): Likewise.
2631 (store_expr): Likewise.
2632 (expand_expr_real_1): Likewise.
2633 * expr.h (store_expr_with_bounds): Likewise.
2634 * function.c (use_register_for_decl): Likewise.
2635 (struct bounds_parm_data): Likewise.
2636 (assign_parms_augmented_arg_list): Likewise.
2637 (assign_parm_find_entry_rtl): Likewise.
2638 (assign_parm_is_stack_parm): Likewise.
2639 (assign_parm_load_bounds): Likewise.
2640 (assign_bounds): Likewise.
2641 (assign_parms): Likewise.
2642 (expand_function_start): Likewise.
2643 * gcc.c (CHKP_SPEC): Likewise.
2644 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2645 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2646 (wrestrict_dom_walker::check_call): Likewise.
2647 * gimple.c (gimple_build_call_from_tree): Likewise.
2648 * gimple.h (enum gf_mask): Likewise.
2649 (gimple_call_with_bounds_p): Likewise.
2650 (gimple_call_set_with_bounds): Likewise.
2651 * gimplify.c (gimplify_init_constructor): Likewise.
2652 * ipa-cp.c (initialize_node_lattices): Likewise.
2653 (propagate_constants_across_call): Likewise.
2654 (find_more_scalar_values_for_callers_subset): Likewise.
2655 * ipa-hsa.c (process_hsa_functions): Likewise.
2656 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2657 * ipa-icf.c (sem_function::merge): Likewise.
2658 * ipa-inline.c (early_inliner): Likewise.
2659 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2660 (warn_function_cold): Likewise.
2661 (propagate_pure_const): Likewise.
2662 * ipa-ref.h (enum GTY): Likewise.
2663 * ipa-split.c (find_retbnd): Likewise.
2664 (consider_split): Likewise.
2665 (split_function): Likewise.
2666 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2667 * ipa.c (walk_polymorphic_call_targets): Likewise.
2668 (symbol_table::remove_unreachable_nodes): Likewise.
2669 (process_references): Likewise.
2670 (cgraph_build_static_cdtor_1): Likewise.
2671 * lto-cgraph.c (lto_output_node): Likewise.
2672 (output_refs): Likewise.
2673 (compute_ltrans_boundary): Likewise.
2674 (input_overwrite_node): Likewise.
2675 (input_node): Likewise.
2676 (input_cgraph_1): Likewise.
2677 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2678 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2679 (class pass_chkp_instrumentation_passes): Likewise.
2680 (make_pass_chkp_instrumentation_passes): Likewise.
2681 * passes.def: Likewise.
2682 * rtl.h (struct GTY): Likewise.
2683 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2684 * stor-layout.c (layout_type): Likewise.
2685 * symtab.c: Likewise.
2686 * target.def: Likewise.
2687 * targhooks.c (default_chkp_bound_type): Likewise.
2688 (default_chkp_bound_mode): Likewise.
2689 (default_builtin_chkp_function): Likewise.
2690 (default_chkp_function_value_bounds): Likewise.
2691 (default_chkp_make_bounds_constant): Likewise.
2692 (default_chkp_initialize_bounds): Likewise.
2693 * targhooks.h (default_chkp_bound_type): Likewise.
2694 (default_chkp_bound_mode): Likewise.
2695 (default_builtin_chkp_function): Likewise.
2696 (default_chkp_function_value_bounds): Likewise.
2697 (default_chkp_make_bounds_constant): Likewise.
2698 (default_chkp_initialize_bounds): Likewise.
2699 * toplev.c (compile_file): Likewise.
2700 (process_options): Likewise.
2701 * tree-core.h (DEF_BUILTIN): Likewise.
2702 (DEF_BUILTIN_CHKP): Likewise.
2703 * tree-inline.c (declare_return_variable): Likewise.
2704 (remap_gimple_stmt): Likewise.
2705 (copy_bb): Likewise.
2706 (initialize_inlined_parameters): Likewise.
2707 (expand_call_inline): Likewise.
2708 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2709 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2710 (make_pass_ipa_chkp_produce_thunks): Likewise.
2711 (make_pass_chkp): Likewise.
2712 (make_pass_chkp_opt): Likewise.
2713 (make_pass_chkp_instrumentation_passes): Likewise.
2714 * tree-pretty-print.c (dump_generic_node): Likewise.
2715 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2716 * tree-ssa-dce.c (propagate_necessity): Likewise.
2717 (eliminate_unnecessary_stmts): Likewise.
2718 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2719 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2720 * tree-ssa-sccvn.h: Likewise.
2721 * tree-ssa-strlen.c (get_string_length): Likewise.
2722 (valid_builtin_call): Likewise.
2723 (adjust_last_stmt): Likewise.
2724 (handle_builtin_strchr): Likewise.
2725 (handle_builtin_strcpy): Likewise.
2726 (handle_builtin_stxncpy): Likewise.
2727 (handle_builtin_memcpy): Likewise.
2728 (handle_builtin_strcat): Likewise.
2729 (strlen_check_and_optimize_stmt): Likewise.
2730 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2731 * tree-streamer-in.c: Likewise.
2732 * tree-streamer.c (record_common_node): Likewise.
2733 * tree.c (tree_code_size): Likewise.
2734 (wide_int_to_tree_1): Likewise.
2735 (type_contains_placeholder_1): Likewise.
2736 (build_common_tree_nodes): Likewise.
2737 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2738 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2739 (POINTER_BOUNDS_P): Likewise.
2740 (BOUNDED_TYPE_P): Likewise.
2741 (BOUNDED_P): Likewise.
2742 (CALL_WITH_BOUNDS_P): Likewise.
2743 (pointer_bounds_type_node): Likewise.
2744 * value-prof.c (gimple_ic): Likewise.
2745 * var-tracking.c (vt_add_function_parameters): Likewise.
2746 * varasm.c (make_decl_rtl): Likewise.
2747 (assemble_start_function): Likewise.
2748 (output_constant): Likewise.
2749 (maybe_assemble_visibility): Likewise.
2750 * varpool.c (ctor_for_folding): Likewise.
2751 * chkp-builtins.def: Remove.
2752 * ipa-chkp.c: Remove.
2753 * ipa-chkp.h: Remove.
2754 * rtl-chkp.c: Remove.
2755 * rtl-chkp.h: Remove.
2756 * tree-chkp-opt.c: Remove.
2757 * tree-chkp.c: Remove.
2758 * tree-chkp.h: Remove.
2760 2018-06-07 Carl Love <cel@us.ibm.com>
2762 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2763 vextract_fp_from_shortl): Add BE support.
2765 2018-06-07 Paul Koning <ni1d@arrl.net>
2767 * compare-elim.c (try_merge_compare): Don't merge compare if
2768 address contains a side effect.
2769 (try_eliminate_compare): Likewise.
2771 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2773 * config.gcc: Support "tremont".
2774 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2775 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2777 * config/i386/i386.c (m_TREMONT): Define.
2778 (processor_target_table): Add "tremont".
2779 (PTA_TREMONT): Define.
2780 (ix86_lea_outperforms): Add TARGET_TREMONT.
2781 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2782 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2783 and M_INTEL_GOLDMONT_PLUS.
2784 (fold_builtin_cpu): Add "tremont".
2785 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2786 (ix86_option_override_internal): Add "tremont".
2787 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2788 (processor_type): Add PROCESSOR_TREMONT.
2789 * config/i386/x86-tune.def: Add m_TREMONT.
2790 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2792 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2794 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2795 symbol defined for msp430i* devices to be lower case.
2797 gcc/testsuite/gcc.target/msp430/
2798 * msp430i-device-symbol.c: New test.
2799 * msp430f-device-symbol.c: New test.
2800 * msp430.h: New test header file.
2802 2018-06-07 Richard Biener <rguenther@suse.de>
2804 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2805 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2806 Properly wrap signed arithmetic if overflow wraps.
2808 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2810 PR tree-optimization/69615
2811 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2812 of a cast from a same precision integral SSA_NAME in a bb dominated
2813 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2814 cast to utype if rhs2 has already a compatible type.
2816 2018-06-07 Richard Biener <rguenther@suse.de>
2818 PR tree-optimization/85935
2819 * graphite-scop-detection.c (find_params_in_bb): Analyze
2820 condition operands with respect to the correct loop. Assert
2821 the analysis doesn't fail.
2823 2018-06-04 Carl Love <cel@us.ibm.com>
2825 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2826 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2827 as it is slightly cheaper.
2828 (first_match_or_eos_index_<mode>):
2829 Calculate index using natural element order.
2830 (first_match_index_<mode>):
2831 Calculate index using natural element order.
2832 (first_match_or_eos_index_<mode>):
2833 Calculate index using natural order.
2834 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2835 for BE and LE modes.
2836 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2837 P9V_BUILTIN_VCLZLSBB_V16QI.
2838 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2841 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2843 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2844 indentation and line wrap for many prototypes. Add missing
2845 @smallexample directives around block of prototypes for vec_xl and
2848 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2850 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2851 track if we pass or return IEEE 128-bit floating point.
2852 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2853 C++ mangling that is compatible with GCC 8.1.
2854 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2855 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2856 floating point types.
2857 (rs6000_function_arg_advance_1): Likewise.
2858 (rs6000_mangle_type): Optionally generate mangled names that match
2859 what GCC 8.1 generated for IEEE 128-bit floating point types.
2860 (rs6000_globalize_decl_name): If we have an external function that
2861 passes or returns IEEE 128-bit types, generate a weak reference
2862 from the mangled name used in GCC 8.1 to the current mangled
2864 (rs6000_init_builtins): Make __ibm128 use the long double type if
2865 long double is IBM extended double. Make __float128 use the long
2866 double type if long double is IEEE 128-bit.
2869 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2870 macro for __ibm128 built-in functions.
2871 (PACK_IF): Add __ibm128 pack/unpack functions.
2872 (UNPACK_IF): Likewise.
2873 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2874 enable long double built-in functions if long double is IEEE
2875 128-bit floating point.
2876 (rs6000_invalid_builtin): Update long double built-in function
2878 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2879 functions, adjust the built-in function to use the long double
2880 built-in function if __ibm128 and long double are the same type.
2881 * doc/extend.texi (PowerPC builtins): Update documention for
2882 __builtin_{,un}pack_longdouble. Add documentation for
2883 __builtin_{,un}pack_ibm128.
2885 2018-06-06 Jim Wilson <jimw@sifive.com>
2887 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2888 (struct machine_function): New field interrupt_mode.
2889 (riscv_handle_type_attribute): New function. Add forward declaration.
2890 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2891 (riscv_expand_epilogue): Check interrupt_mode field.
2892 (riscv_set_current_function): Check interrupt attribute args and
2893 set interrupt_mode field.
2894 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2895 (riscv_sret, riscv_uret): New.
2896 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2897 new arguments to interrupt attribute.
2899 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2902 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2903 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2905 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2907 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2908 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2909 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2910 several redundant entries.
2912 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2914 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2915 type from "rtx" to "rtx_insn *".
2916 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2917 for local "call_insn", removing cast.
2918 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2920 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2922 PR tree-optimization/86066
2923 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2924 for BIT_INSERT_EXPR stores.
2926 2018-06-06 Richard Biener <rguenther@suse.de>
2928 PR tree-optimization/86062
2929 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2930 component refs ontop
2931 of to be offsetted base.
2933 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2935 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2936 to be static and remove check on interrupt attribute name.
2938 gcc/testsuite/gcc.target/msp430/
2939 * function-attributes-4.c: New test.
2940 * static-interrupts.c: New test.
2942 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2944 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2945 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2947 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2950 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2952 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2953 Remove second argument, change how error is called.
2954 (aarch64_layout_arg): Remove second argument from
2955 aarch64_err_no_fpadvsimd call.
2956 (aarch64_init_cumulative_args): Ditto.
2957 (aarch64_gimplify_va_arg_expr): Ditto.
2958 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2960 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2962 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2963 (*simple_return_indirect_internal<mode>): Rename from
2964 simple_return_indirect_internal. Use W mode iterator.
2965 (rstorssp): New expander.
2966 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2967 (clrssbsy): New expander.
2968 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2970 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2972 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2974 (cmse_check_pointed_object): Likewise.
2976 2018-06-05 Martin Liska <mliska@suse.cz>
2978 PR gcov-profile/47618
2979 * doc/invoke.texi: Document how -fprofile-dir format
2982 2018-06-05 Richard Biener <rguenther@suse.de>
2984 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2985 removal pretend DOM info isn't available so we do not update
2986 it and only remove edges, not dominated blocks. Actually free
2987 DOM info in case we removed something. Remove unreachable blocks.
2988 (mfb_keep_latches): Work with either DOM info or marked backedges.
2989 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2990 first. Mark backedges if DOM info isn't available.
2991 (Re-)compute DOM info after cleanup_control_flow_pre.
2993 2018-06-05 Richard Biener <rguenther@suse.de>
2995 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2996 (locus_discrim_hasher::hash): Adjust.
2997 (locus_discrim_hasher::equal): Likewise.
2998 (next_discriminator_for_locus): Work on line directly.
2999 (same_line_p): Pass in expanded locus1 as well.
3000 (assign_discriminators): Avoid redundant location expansions.
3002 2018-06-05 Richard Biener <rguenther@suse.de>
3004 PR tree-optimization/86046
3005 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3006 if required after clearing TREE_ADDRESSABLE.
3008 2018-06-05 Richard Biener <rguenther@suse.de>
3010 PR tree-optimization/86047
3011 * tree-ssa-loop.c (for_each_index): Glob handling of all
3012 decls and constants and really handle all of them.
3014 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3017 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3018 qualifier_void_pointer and qualifier_const_void_pointer.
3019 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3020 (arm_init_builtins): Handle the above.
3021 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3022 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3023 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3026 2018-06-05 Martin Liska <mliska@suse.cz>
3028 * auto-profile.c (read_autofdo_file): Do not use
3029 gcov_ctr_summary struct.
3030 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3031 * coverage.c (struct counts_entry): Likewise.
3032 (read_counts_file): Read just single summary entry.
3033 (get_coverage_counts): Use gcov_summary struct.
3034 * coverage.h (get_coverage_counts): Likewise.
3035 * gcov-dump.c (dump_working_sets): Likewise.
3036 (tag_summary): Dump just single summary.
3037 * gcov-io.c (gcov_write_summary): Write just histogram
3039 (gcov_read_summary): Read just single summary.
3040 (compute_working_sets): Use gcov_summary struct.
3041 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3042 of GCOV_COUNTERS_SUMMABLE.
3043 (GCOV_COUNTERS_SUMMABLE): Remove.
3044 (GCOV_FIRST_VALUE_COUNTER): Replace with
3045 GCOV_COUNTER_V_INTERVAL.
3046 (struct gcov_ctr_summary): Remove.
3047 (struct gcov_summary): Directly use fields of former
3049 (compute_working_sets): Use gcov_summary struct.
3050 * gcov.c (read_count_file): Do not use ctrs fields.
3051 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3053 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3055 * profile.c: Likewise.
3056 * profile.h: Likewise.
3058 2018-06-05 Martin Liska <mliska@suse.cz>
3060 PR gcov-profile/84846
3061 * gcov.c (output_lines): Print working directory only
3062 in intermediate format.
3064 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3066 * config/s390/s390-builtin-types.def: Add void function type.
3067 * config/s390/s390-builtins.def: Use the function type for the
3070 2018-06-04 Jim Wilson <jimw@sifive.com>
3072 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3074 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3075 and maybe_eh_return. Change regno to unsigned int. Use new args to
3076 handle EH_RETURN_DATA_REGNO registers properly.
3077 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3078 (riscv_expand_epilogue): Update comment. Change argument name and
3079 type. Update code to use new name and type. Pass new args to
3080 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3082 * config/riscv/riscv.md (NORMAL_RETURN): New.
3083 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3084 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3085 (eh_return): Call gen_eh_return_internal and emit barrier.
3086 (eh_return_internal): Call riscv_expand_epilogue.
3088 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3090 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3091 bit_insertion field and declare can_be_merged_into method.
3092 (merged_store_group::can_be_merged_into): New method.
3093 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3094 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3095 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3097 2018-06-04 Richard Biener <rguenther@suse.de>
3099 PR tree-optimization/85955
3100 * builtins.c (fold_builtin_sincos): Convert pointers to
3101 destination to appropriate type before dereferencing.
3103 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3105 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3107 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3109 * expr.c (expand_expr_real_1): Force the operand into memory if
3110 its TYPE_MODE is BLKmode and if there is no integer mode for
3111 the number of bits being extracted.
3113 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3117 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3118 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3120 2018-06-04 Richard Biener <rguenther@suse.de>
3122 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3123 (cleanup_tree_cfg_noloop): ... single caller. Do
3124 start_recording_case_labels later.
3126 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3128 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3129 to _IMMINTRIN_H_INCLUDED.
3130 * config/i386/pconfigintrin.h: Ditto.
3131 * config/i386/waitpkgintrin.h: Ditto.
3132 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3133 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3134 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3135 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3136 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3137 waitpkgintrin.h and cldemoteintrin.h.
3139 2018-06-04 Richard Biener <rguenther@suse.de>
3141 PR tree-optimization/86038
3142 * tracer.c (find_best_successor): Check probability for
3143 being initialized, bail out if not.
3145 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3148 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3149 of bits to ignore when comparing architectures.
3151 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3153 PR tree-optimization/69615
3154 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
3155 maximum or minimum of the type, try to merge it also as if
3156 range1 is + [-, x - 1] or + [x + 1, -].
3159 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3161 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
3163 PR tree-optimization/86034
3164 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
3165 the unsigned bitfield type in a bit insertion sequence if it does not
3166 have a larger precision than the bitfield size.
3167 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
3169 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
3171 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
3173 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3175 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
3176 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
3177 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
3178 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
3180 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
3182 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
3183 Disable -fdelete-null-pointer-checks for ELF toolchain.
3185 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3186 Kito Cheng <kito.cheng@gmail.com>
3188 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
3189 (nds32le-*-*, nds32be-*-*): Integrate checking process.
3190 (nds32*-*-*): Add glibc and uclibc conditions.
3191 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
3192 (TARGET_EXCEPT_UNWIND_INFO): Define.
3193 * config/nds32/elf.h: New file.
3194 * config/nds32/linux.h: New file.
3195 * config/nds32/nds32-elf.opt: New file.
3196 * config/nds32/nds32-linux.opt: New file.
3197 * config/nds32/nds32-fp-as-gp.c
3198 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
3199 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
3201 (nds32_asm_file_end): Ditto.
3202 (nds32_print_operand): Ditto.
3203 (nds32_insert_attributes): Ditto.
3204 (nds32_init_libfuncs): New function.
3205 (TARGET_HAVE_TLS): Define.
3206 (TARGET_INIT_LIBFUNCS): Define.
3207 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
3209 (TARGET_ELF): Apply different mcmodel setting.
3210 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
3211 been migrated into elf.h and linux.h files.
3212 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
3213 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
3214 (mcmodel): The content has been migrated into nds32-elf.opt and
3215 nds32-linux.opt files.
3216 * config/nds32/t-elf: New file.
3217 * config/nds32/t-linux: New file.
3219 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
3220 Shiva Chen <shiva0217@gmail.com>
3222 * config/nds32/constants.md (unspec_volatile_element): Add
3223 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
3224 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
3226 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
3227 (make_pass_nds32_fp_as_gp): Declare.
3228 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
3230 (nds32_asm_function_end_prologue): Remove unused asm output.
3231 (nds32_asm_function_begin_epilogue): Remove unused asm output.
3232 (nds32_asm_file_start): Output necessary fp_as_gp information.
3233 (nds32_option_override): Adjust register usage.
3234 (nds32_expand_prologue): Consider fp_as_gp situation.
3235 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
3236 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
3239 (simple_return): Ditto.
3240 (omit_fp_begin): Output special directive for fp_as_gp.
3241 (omit_fp_end): Output special directive for fp_as_gp.
3242 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
3243 mforbid-fp-as-gp): New options.
3245 2018-06-01 Mark Wielaard <mark@klomp.org>
3247 * dwarf2out.c (dwarf2out_finish): Remove generation of
3248 DW_AT_loclists_base.
3250 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
3252 * gimple-ssa-store-merging.c: Include gimple-fold.h.
3253 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
3254 (struct merged_store_group): Add bit_insertion field.
3255 (dump_char_array): Use standard hexadecimal format.
3256 (merged_store_group::merged_store_group): Set bit_insertion to false.
3257 (merged_store_group::apply_stores): Use optimal buffer size. Deal
3258 with BIT_INSERT_EXPR stores. Move up code updating the mask and
3259 also print the mask in the dump file.
3260 (pass_store_merging::gate): Minor tweak.
3261 (imm_store_chain_info::coalesce_immediate): Fix wrong association
3262 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
3263 stores with INTEGER_CST stores.
3264 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
3265 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
3266 and use it throughout. Generate bit insertion sequences if need be.
3267 (pass_store_merging::process_store): Remove redundant condition.
3268 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
3270 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3272 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3273 the 128-bit floating point types. Fix function comment.
3275 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3277 * config/aarch64/aarch64-simd.md
3278 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
3280 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
3283 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3285 PR tree-optimization/85989
3286 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3288 (backprop::intersect_uses): Check it when deciding whether this
3289 is a backedge reference.
3290 (backprop::process_block): Add each phi to m_visited_phis
3291 after visiting it, then clear it at the end.
3293 2018-06-01 Richard Biener <rguenther@suse.de>
3295 * tree-vectorizer.h (vect_dr_stmt): New function.
3296 (vect_get_load_cost): Adjust.
3297 (vect_get_store_cost): Likewise.
3298 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
3299 Use vect_dr_stmt instead of DR_SMTT.
3300 (vect_record_base_alignments): Likewise.
3301 (vect_calculate_target_alignment): Likewise.
3302 (vect_compute_data_ref_alignment): Likewise and make static.
3303 (vect_update_misalignment_for_peel): Likewise.
3304 (vect_verify_datarefs_alignment): Likewise.
3305 (vector_alignment_reachable_p): Likewise.
3306 (vect_get_data_access_cost): Likewise. Pass down
3307 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
3308 (vect_get_peeling_costs_all_drs): Likewise.
3309 (vect_peeling_hash_get_lowest_cost): Likewise.
3310 (vect_enhance_data_refs_alignment): Likewise.
3311 (vect_find_same_alignment_drs): Likewise.
3312 (vect_analyze_data_refs_alignment): Likewise.
3313 (vect_analyze_group_access_1): Likewise.
3314 (vect_analyze_group_access): Likewise.
3315 (vect_analyze_data_ref_access): Likewise.
3316 (vect_analyze_data_ref_accesses): Likewise.
3317 (vect_vfa_segment_size): Likewise.
3318 (vect_small_gap_p): Likewise.
3319 (vectorizable_with_step_bound_p): Likewise.
3320 (vect_prune_runtime_alias_test_list): Likewise.
3321 (vect_analyze_data_refs): Likewise.
3322 (vect_supportable_dr_alignment): Likewise.
3323 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3324 (vect_gen_prolog_loop_niters): Likewise.
3325 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3326 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
3328 (vect_recog_mask_conversion_pattern): Likewise.
3329 (vect_try_gather_scatter_pattern): Likewise.
3330 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
3331 to vect_get_store_cost.
3332 (vect_get_store_cost): Get stmt_info instead of DR.
3333 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
3334 (vect_get_load_cost): Get stmt_info instead of DR.
3336 2018-06-01 Richard Biener <rguenther@suse.de>
3339 * gimple-fold.c (var_decl_component_p): Also allow offsetted
3340 vars wrapped in MEM_REFs.
3342 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
3344 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3345 Fix subreg tests so that we only return a choice between
3346 GENERAL_REGS and FP_REGS if the original classes included both.
3348 2018-06-01 Richard Biener <rguenther@suse.de>
3351 * tree-ssa-structalias.c (get_function_part_constraint):
3352 Handle NULL fi->decl.
3353 (find_func_aliases_for_call): Properly handle indirect
3354 fi from direct call.
3355 (find_func_clobbers): Likewise.
3356 (ipa_pta_execute): Likewise.
3357 (create_variable_info_for): For functions that are ifunc_resolver
3358 resolve to a varinfo that contains the result of the resolver call.
3359 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
3362 2018-05-31 Michael Collison <michael.collison@arm.com>
3364 * config/aarch64/aarch64.md:
3365 (*fix_to_zero_extenddfdi2): New pattern.
3366 * gcc.target/aarch64/fix_extend1.c: New testcase.
3368 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
3372 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
3373 and BUILT_IN_STRNCMP_EQ.
3374 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
3375 BUILT_IN_STRNCMP_EQ.
3376 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
3377 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3378 (gimple_fold_builtin): Likewise.
3379 * tree-ssa-strlen.c (compute_string_length): New function.
3380 (determine_min_obsize): New function.
3381 (handle_builtin_string_cmp): New function to handle calls to
3382 string compare functions.
3383 (strlen_optimize_stmt): Add handling to builtin string compare
3385 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3386 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
3387 * tree.c (build_common_builtin_nodes): Add new defines of
3388 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
3390 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3393 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3394 functions with naked attribute.
3396 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3398 * config/i386/sse.md (avx_vec_concat<mode>):
3399 Substitute concat_tg_mode mode attribute with xtg_mode.
3400 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
3401 (concat_tg_mode): Remove mode attribute.
3403 2018-05-31 Martin Sebor <msebor@redhat.com>
3406 * calls.c (alloc_max_size): Correct a logic error/typo.
3407 Treat excessive arguments as infinite. Warn for invalid arguments.
3408 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3410 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3413 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3414 and movx for Haswell.
3416 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
3417 Cesar Philippidis <cesar@codesourcery.com>
3420 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
3421 when emitting error on private/firstprivate reductions.
3422 * omp-low.c (lower_omp_target): Avoid reference-type processing
3423 on pointers for firstprivate clause.
3425 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3427 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
3430 * config/aarch64/aarch64-simd.md
3431 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
3432 (aarch64_ld1_x3_<mode>): Likewise
3433 (aarch64_st1x2<VALLDIF:mode>): Likewise
3434 (aarch64_st1_x2_<mode>): Likewise
3435 (aarch64_st1x3<VALLDIF:mode>): Likewise
3436 (aarch64_st1_x3_<mode>): Likewise
3437 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
3438 (vld1_s8_x3): Likewise.
3439 (vld1_u16_x3): Likewise.
3440 (vld1_s16_x3): Likewise.
3441 (vld1_u32_x3): Likewise.
3442 (vld1_s32_x3): Likewise.
3443 (vld1_u64_x3): Likewise.
3444 (vld1_s64_x3): Likewise.
3445 (vld1_f16_x3): Likewise.
3446 (vld1_f32_x3): Likewise.
3447 (vld1_f64_x3): Likewise.
3448 (vld1_p8_x3): Likewise.
3449 (vld1_p16_x3): Likewise.
3450 (vld1_p64_x3): Likewise.
3451 (vld1q_u8_x3): Likewise.
3452 (vld1q_s8_x3): Likewise.
3453 (vld1q_u16_x3): Likewise.
3454 (vld1q_s16_x3): Likewise.
3455 (vld1q_u32_x3): Likewise.
3456 (vld1q_s32_x3): Likewise.
3457 (vld1q_u64_x3): Likewise.
3458 (vld1q_s64_x3): Likewise.
3459 (vld1q_f16_x3): Likewise.
3460 (vld1q_f32_x3): Likewise.
3461 (vld1q_f64_x3): Likewise.
3462 (vld1q_p8_x3): Likewise.
3463 (vld1q_p16_x3): Likewise.
3464 (vld1q_p64_x3): Likewise.
3465 (vst1_s64_x2): Likewise.
3466 (vst1_u64_x2): Likewise.
3467 (vst1_f64_x2): Likewise.
3468 (vst1_s8_x2): Likewise.
3469 (vst1_p8_x2): Likewise.
3470 (vst1_s16_x2): Likewise.
3471 (vst1_p16_x2): Likewise.
3472 (vst1_s32_x2): Likewise.
3473 (vst1_u8_x2): Likewise.
3474 (vst1_u16_x2): Likewise.
3475 (vst1_u32_x2): Likewise.
3476 (vst1_f16_x2): Likewise.
3477 (vst1_f32_x2): Likewise.
3478 (vst1_p64_x2): Likewise.
3479 (vst1q_s8_x2): Likewise.
3480 (vst1q_p8_x2): Likewise.
3481 (vst1q_s16_x2): Likewise.
3482 (vst1q_p16_x2): Likewise.
3483 (vst1q_s32_x2): Likewise.
3484 (vst1q_s64_x2): Likewise.
3485 (vst1q_u8_x2): Likewise.
3486 (vst1q_u16_x2): Likewise.
3487 (vst1q_u32_x2): Likewise.
3488 (vst1q_u64_x2): Likewise.
3489 (vst1q_f16_x2): Likewise.
3490 (vst1q_f32_x2): Likewise.
3491 (vst1q_f64_x2): Likewise.
3492 (vst1q_p64_x2): Likewise.
3493 (vst1_s64_x3): Likewise.
3494 (vst1_u64_x3): Likewise.
3495 (vst1_f64_x3): Likewise.
3496 (vst1_s8_x3): Likewise.
3497 (vst1_p8_x3): Likewise.
3498 (vst1_s16_x3): Likewise.
3499 (vst1_p16_x3): Likewise.
3500 (vst1_s32_x3): Likewise.
3501 (vst1_u8_x3): Likewise.
3502 (vst1_u16_x3): Likewise.
3503 (vst1_u32_x3): Likewise.
3504 (vst1_f16_x3): Likewise.
3505 (vst1_f32_x3): Likewise.
3506 (vst1_p64_x3): Likewise.
3507 (vst1q_s8_x3): Likewise.
3508 (vst1q_p8_x3): Likewise.
3509 (vst1q_s16_x3): Likewise.
3510 (vst1q_p16_x3): Likewise.
3511 (vst1q_s32_x3): Likewise.
3512 (vst1q_s64_x3): Likewise.
3513 (vst1q_u8_x3): Likewise.
3514 (vst1q_u16_x3): Likewise.
3515 (vst1q_u32_x3): Likewise.
3516 (vst1q_u64_x3): Likewise.
3517 (vst1q_f16_x3): Likewise.
3518 (vst1q_f32_x3): Likewise.
3519 (vst1q_f64_x3): Likewise.
3520 (vst1q_p64_x3): Likewise.
3522 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3524 * config/msp430/msp430.c (msp430_output_labelref): Prepend
3525 user_label_prefix to name.
3527 * tree-core.h: Update comment about the format of NAME string
3528 passed to handler in attribute_spec.
3530 * config/msp430/msp430.md: Remove erroneous subreg expression from
3531 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3532 zero_extend{q,h}isi2.
3534 2018-05-30 Borislav Petkov <bp@suse.de>
3536 * doc/extend.texi: Document some architecture specific
3537 constraints and sort entries.
3539 2018-05-30 Martin Sebor <msebor@redhat.com>
3542 * builtins.c (expand_builtin_stpcpy_1): New function.
3543 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3544 only if the former succeeds.
3546 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3548 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3551 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3553 * doc/invoke.texi (-flinker-output): Document
3555 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3557 * passes.c (ipa_write_summaries): Only modify statements if body
3559 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3560 incrementally linking.
3561 (ipa_passes): Likewise.
3562 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3563 pass down resolution info.
3564 * common.opt (flag_incremental_link): Update info.
3565 * gcc.c (plugin specs): Turn flinker-output=* to
3566 -plugin-opt=-linker-output-known
3567 * toplev.c (compile_file): Also cut compilation when doing incremental
3569 * flag-types. (enum lto_partition_model): Add
3570 LTO_LINKER_OUTPUT_NOLTOREL.
3571 (invoke.texi): Add -flinker-output docs.
3572 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3573 link same way as WPA; do not stream in dead initializers.
3575 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3576 darf2out_register_external_decl): Support incremental link.
3578 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3580 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3582 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3584 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3585 it down to simple_object_copy_lto_debug_sections.
3586 (run_gcc): Determine incremental LTO link time and configure
3587 lto1 into non-wpa mode, disable renaming of debug sections.
3589 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3591 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3592 descriptions of various incorrectly documented functions.
3594 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3597 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3598 address check not strict.
3600 2018-05-30 Richard Biener <rguenther@suse.de>
3602 PR tree-optimization/85964
3603 * tracer.c (better_p): Drop initialized count check, we only
3604 call the function with initialized counts now.
3605 (find_best_successor): Do find a best edge if one
3606 has uninitialized count.
3607 (find_best_predecessor): Likewise. Do BB frequency check only
3608 if count is initialized.
3610 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3612 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3613 (aarch64_ldrstr_offset_compare): New.
3614 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3615 load/store orderings.
3616 (aarch64_gen_adjusted_ldpstp): Likewise.
3618 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3620 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3621 Check for subset of GENERAL_REGS and FP_REGS.
3622 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3625 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3627 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3628 and wi::to_poly_offset. Add the current offset and then check
3629 whether the sum fits, rather than using an unchecked addition of
3630 a checked term. Check for a shwi rather than a uhwi.
3631 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3632 (store_constructor): Use poly_int_tree_p.
3633 (expand_expr_real_1): Likewise.
3634 * function.c (assign_temp): Likewise.
3635 * fold-const.c (const_binop): Use poly_int_tree_p and
3637 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3639 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3640 to_poly_offset for MEM offsets.
3641 * ipa-icf.c (sem_variable::equals): Likewise.
3642 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3643 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3644 wi::to_poly_offset for BIT_FIELD_REF offsets.
3645 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3647 * var-tracking.c (emit_note_insn_var_location): Use
3648 tree_to_poly_uint64.
3650 2018-05-29 Jim Wilson <jimw@sifive.com>
3652 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3654 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3657 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3658 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3660 (sse4_1_round<mode>2): Use nonimmediate_operand
3661 for operand 1 predicate.
3663 2018-05-29 Martin Sebor <msebor@redhat.com>
3664 Richard Biener <rguenther@suse.de>
3667 * calls.c (get_size_range): Call determine_value_range instead
3668 of get_value_range..
3669 * tree-vrp.h (determine_value_range): Declared new function.
3670 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3672 2018-05-29 Richard Biener <rguenther@suse.de>
3674 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3675 sure to use non-pattern stmts for get_earlier_stmt arguments.
3676 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3677 called on pattern stmts.
3678 (get_later_stmt): Likewise.
3680 2018-05-29 Martin Liska <mliska@suse.cz>
3682 PR gcov-profile/85759
3683 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3686 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3688 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3690 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3694 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3695 VEC_PACK_FLOAT_EXPR): New tree codes.
3696 * tree-pretty-print.c (op_code_prio): Handle
3697 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3698 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3699 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3700 * tree-inline.c (estimate_operator_cost): Likewise.
3701 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3702 * fold-const.c (const_binop): Likewise.
3703 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3704 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3705 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3706 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3707 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3708 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3709 * expr.c (expand_expr_real_2): Likewise.
3710 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3711 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3712 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3714 * optabs.c (expand_widen_pattern_expr): For
3715 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3716 sign from result type rather than operand's type.
3717 (expand_binop_directly): For vec_packu_float_optab and
3718 vec_packs_float_optab allow result type to be different from operand's
3720 * optabs-tree.c (optab_for_tree_code): Handle
3721 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3722 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3723 * tree-vect-generic.c (expand_vector_operations_1): Handle
3724 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3725 VEC_PACK_FLOAT_EXPR.
3726 * tree-vect-stmts.c (supportable_widening_operation): Handle
3728 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3729 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3730 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3731 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3732 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3734 (vec_pack<floatprefix>_float_<mode>): New expander.
3735 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3737 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3738 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3739 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3740 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3741 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3743 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3744 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3745 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3746 VEC_PACK_FLOAT_EXPR): Document.
3748 2018-05-29 Richard Biener <rguenther@suse.de>
3750 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3752 (stmt_vec_info_vec): Make pointer.
3753 (init_stmt_vec_info_vec): Remove.
3754 (free_stmt_vec_info_vec): Likewise.
3755 (set_stmt_vec_info_vec): New function.
3756 (free_stmt_vec_infos): Likewise.
3757 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3758 (set_vinfo_for_stmt): Likewise.
3759 (get_earlier_stmt): Likewise.
3760 (get_later_stmt): Likewise.
3761 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3762 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3763 (vec_info::~vec_info): Free stmt_vec_infos.
3764 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3765 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3766 (pass_slp_vectorize::execute): Likewise.
3767 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3768 (free_stmt_vec_info_vec): Likewise.
3769 (set_stmt_vec_info_vec): New function.
3770 (free_stmt_vec_infos): Likewise.
3771 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3772 the global stmt_vec_info_vec.
3773 * tree-parloops.c (gather_scalar_reductions): Use
3774 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3777 2018-05-29 Richard Biener <rguenther@suse.de>
3779 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3781 2018-05-29 Martin Liska <mliska@suse.cz>
3782 David Malcolm <dmalcolm@redhat.com>
3784 * vec.c (test_reverse): New.
3785 (vec_c_tests): Add new test.
3786 * vec.h (vl_ptr>::reverse): New function.
3788 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3790 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3792 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3795 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3797 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3799 2018-05-28 Richard Biener <rguenther@suse.de>
3801 PR tree-optimization/85933
3802 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3803 look at stmts marked as vectorizable.
3805 2018-05-28 Richard Biener <rguenther@suse.de>
3807 PR tree-optimization/85934
3808 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3809 vector boolean check before scalar optimization.
3811 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3813 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3816 2018-05-28 Mark Wielaard <mark@klomp.org>
3818 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3819 if it is an expression containing a minus sign.
3821 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3823 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3825 2018-05-27 Paul Koning <ni1d@arrl.net>
3827 * config/pdp11/pdp11.md (truncsihi2): Remove.
3829 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3830 Chung-Ju Wu <jasonwucj@gmail.com>
3832 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3834 (unaligned_store_dw): Ditto.
3835 * config/nds32/nds32-memory-manipulation.c
3836 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3837 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3838 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3839 (emit_setmem_word_loop): Rename to ...
3840 (emit_setmem_doubleword_loop): ... this.
3841 (nds32_gen_dup_4_byte_to_word_value): New function.
3842 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3843 (nds32_expand_setmem_loop): Refine implementation.
3844 (nds32_expand_setmem_loop_v3m): Ditto.
3845 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3848 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3850 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3852 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3855 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3856 machine->attr_no_prologue_p.
3857 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3858 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3859 (nds32_expand_epilogue): Consider attr_naked_p.
3860 (nds32_expand_epilogue_v3pop): Likewise.
3861 (nds32_can_use_return_insn): Likewise.
3862 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3863 attr_no_prologue_p fields.
3864 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3866 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3869 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3871 * config/i386/sse.md
3872 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3874 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3876 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3878 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3880 (*<floatsuffix>floatv2div2sf2): Rename to ...
3881 (*float<floatunssuffix>v2div2sf2): ... this.
3882 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3883 (float<floatunssuffix>v2div2sf2_mask): ... this.
3884 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3885 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3886 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3888 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3890 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3892 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3894 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3896 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3898 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3899 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3900 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3901 gen_ufix_truncv8dfv8si2.
3902 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3903 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3904 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3905 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3906 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3907 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3908 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3909 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3911 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3915 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3917 2018-05-25 Jim Wilson <jimw@sifive.com>
3919 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3920 * config/riscv/riscv.c (struct machine_function): Add
3921 interrupt_handler_p and attribute_checked_p fields.
3922 (riscv_attribute_table): Add interrupt.
3923 (riscv_interrupt_type_p): New.
3924 (riscv_save_reg_p): Save extra regs for interrupt handler.
3925 (riscv_use_save_libcall): Return false for interrupt handler.
3926 (riscv_first_stack_step): Add forward declaration.
3927 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3928 for interrupt handler with large frame. Use it for saved reg list.
3929 (riscv_expand_prologue): Move flag_stack_usage_info support to
3930 eliminate duplication.
3931 (riscv_expand_epilogue): Generate mret for interrupt handler.
3932 (riscv_epilogue_uses): New.
3933 (riscv_can_use_return_insn): Return false for interrupt handler.
3934 (riscv_function_ok_for_sibcall): Likewise.
3935 (riscv_set_current_function): Add interrupt handler support.
3936 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3937 * config/riscv/riscv.md (UNSPECV_MRET): New.
3939 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3941 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3943 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3945 PR tree-optimization/85712
3946 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3947 this candidate has already been replaced in-situ by a copy.
3949 2018-05-25 Jason Merrill <jason@redhat.com>
3951 PR c++/80485 - inline function non-zero address.
3952 * symtab.c (nonzero_address): Check DECL_COMDAT.
3954 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3957 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3958 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3959 extension of SImode operation. Use const123_operand predicate.
3960 (*saddsi_1): Remove.
3961 (*saddl_se_1): Ditto.
3963 (*ssubl_se_1): Ditto.
3964 * config/alpha/predicates.md (const123_operand): New predicate.
3965 * config/alpha/constraints.md (P): Use IN_RANGE.
3967 2018-05-25 Richard Biener <rguenther@suse.de>
3969 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3971 (ref_maybe_used_by_stmt_p): Likewise.
3972 (stmt_may_clobber_ref_p): Likewise.
3973 (stmt_may_clobber_ref_p_1): Likewise.
3974 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3976 (ref_maybe_used_by_stmt_p): Likewise.
3977 (stmt_may_clobber_ref_p): Likewise.
3978 (stmt_may_clobber_ref_p_1): Likewise.
3979 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3980 the alias oracle to disambiguate DRs with stmts DR analysis
3982 (vect_analyze_data_refs): Do not give up on not analyzable
3983 DRs for BB vectorization. Remove code truncating the dataref
3986 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3989 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3990 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3991 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3993 2018-05-25 Richard Biener <rguenther@suse.de>
3995 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3996 function, combining stmt data ref gathering and fatal analysis
3998 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3999 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4000 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4001 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4002 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4004 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4006 PR tree-optimization/85720
4007 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4008 SCC if all partitions are builtins.
4009 (version_loop_by_alias_check): New parameter. Generate cancelable
4010 runtime alias check if all partitions are builtins.
4011 (distribute_loop): Update call to above function.
4013 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4015 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4016 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4017 (parm_default_def_partition_arg): Ditto.
4018 (set_parm_default_def_partition): Ditto.
4019 (get_parm_default_def_partitions): Ditto and make it static.
4020 (get_undefined_value_partitions): Ditto and make it static.
4021 (remove_ssa_form): Refactor call to init_var_map here.
4022 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4023 computation for loop region.
4024 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4025 (register_default_def): Delete.
4026 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4027 (parm_default_def_partition_arg): Ditto.
4028 (set_parm_default_def_partition): Ditto.
4029 (get_parm_default_def_partitions): Ditto and make it static.
4030 (get_undefined_value_partitions): Ditto and make it static.
4031 (coalesce_with_default, coalesce_with_default): Update comment.
4032 (create_coalesce_list_for_region): New func factored out from
4033 create_outofssa_var_map.
4034 (populate_coalesce_list_for_outofssa): New func factored out from
4035 create_outofssa_var_map and coalesce_ssa_name.
4036 (create_outofssa_var_map): Delete.
4037 (coalesce_ssa_name): Refactor to support live range computation.
4038 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4039 (get_parm_default_def_partitions): Delete.
4040 (get_undefined_value_partitions): Ditto.
4041 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4042 computation for loop region.
4043 (new_tree_live_info, loe_visit_block): Ditto.
4044 (live_worklist, set_var_live_on_entry): Ditto.
4045 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4046 * tree-ssa-live.h (struct _var_map): New fields.
4047 (init_var_map): Change decl.
4048 (region_contains_p): New.
4050 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4052 * tree-ssa-live.h (live_merge_and_clear): Delete.
4054 2018-05-25 Richard Biener <rguenther@suse.de>
4057 * tree-dump.c (dequeue_and_dump): Remove access to removed
4058 operand 2 of a SWITCH_EXPR.
4060 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4062 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4063 multiplication and division.
4064 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4065 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4066 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4067 (cond_udiv_optab, cond_umod_optab): New optabs.
4068 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4069 (IFN_COND_RDIV): New internal functions.
4070 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4071 TRUNC_MOD_EXPR and RDIV_EXPR.
4072 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4073 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4075 (SVE_INT_BINARY): Include mult.
4076 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4077 (optab, sve_int_op): Handle mult.
4078 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4080 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4081 for SVE_INT_BINARY_SD.
4083 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4085 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4086 (optab, sve_int_op): Handle div and udiv.
4087 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4088 for SVE_INT_BINARY_SD.
4089 (*<optab><mode>3): New insn for the same.
4091 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4093 * tree-vect-patterns.c: Include predict.h.
4094 (vect_recog_divmod_pattern): Restrict check for division support
4095 to when optimizing for size.
4097 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4099 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4100 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4101 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4102 (gimple_match_op::set_op): Likewise.
4103 (gimple_resimplify4): Declare.
4104 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4105 (expr::gen_transform): Likewise.
4106 (decision_tree::gen): Generate a simplification routine for 4 operands.
4107 * gimple-match-head.c (gimple_simplify): Add an overload for
4108 4 operands. In the top-level function, handle up to 4 call
4109 arguments and call gimple_resimplify4.
4110 (gimple_resimplify4): New function.
4111 (build_call_internal): Pass a fourth operand.
4112 (maybe_push_to_seq): Likewise.
4113 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4114 Fold VEC_COND_EXPRs of an operation and a default value into
4115 an IFN_COND_* function if possible.
4116 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4118 (SVE_COND_FP_BINARY): Include them.
4119 (optab, sve_fp_op): Handle them.
4120 (SVE_INT_BINARY_REV): New code iterator.
4121 (SVE_COND_FP_BINARY_REV): New int iterator.
4122 (commutative): New int attribute.
4123 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4125 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4127 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4128 (*cond_<optab><mode>): New patterns for reversed operands.
4130 2018-05-25 Richard Biener <rguenther@suse.de>
4132 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4133 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4134 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4135 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4136 * tree-vect-data-refs.c (everywhere): Adjust users.
4137 * tree-vect-loop.c (everywhere): Likewise.
4138 * tree-vect-slp.c (everywhere): Likewise.
4139 * tree-vect-stmts.c (everywhere): Likewise.
4140 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4142 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4144 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4146 (gcc_cv_as_section_exclude): ... this.
4147 Try Solaris as #exclude syntax.
4148 * configure: Regenerate.
4149 * config.in: Regenerate.
4150 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
4152 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
4153 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
4155 * varasm.c (default_elf_asm_named_section): Don't check if
4156 HAVE_GAS_SECTION_EXCLUDE is defined.
4158 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4160 * doc/md.texi: Update the documentation of the cond_* optabs
4161 to mention the new final operand. Fix GET_MODE_NUNITS call.
4162 Describe the scalar case too.
4163 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
4164 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
4166 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
4167 (get_conditional_internal_fn): Update comment.
4168 * tree-vect-loop.c (vectorizable_reduction): Pass the original
4169 accumulator value as a final argument to conditional functions.
4170 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
4171 a define_expand and add an "else" operand. Assert for now that
4172 the else operand is equal to operand 2. Use SVE_INT_BINARY and
4173 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
4174 (*cond_<optab><mode>): New patterns.
4175 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
4176 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4177 (UNSPEC_COND_EOR): Delete.
4178 (optab): Remove associated mappings.
4179 (SVE_INT_BINARY): New code iterator.
4180 (sve_int_op): Remove int attribute and add "minus" to the code
4182 (SVE_COND_INT_OP): Delete.
4183 (SVE_COND_FP_OP): Rename to...
4184 (SVE_COND_FP_BINARY): ...this.
4186 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4188 * optabs.c (can_reuse_operands_p): New function.
4189 (maybe_legitimize_operands): Try to reuse the results for
4192 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4194 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
4195 Add {q} suffix to insn mnemonic.
4197 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4199 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
4200 (msp430_warn_func_return): New.
4202 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
4204 * fold-const.c (tree_nonzero_bits): New function.
4205 * fold-const.h (tree_nonzero_bits): Likewise.
4206 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
4207 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
4209 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4213 * varasm.c (assemble_alias): Check ifunc_resolver only on
4216 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
4219 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
4220 when memory input operand is handled.
4222 2018-05-24 Luis Machado <luis.machado@linaro.org>
4224 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
4226 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
4228 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4230 * match.pd: Delay FMA folds until after vectorization.
4232 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
4235 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4236 address check not strict.
4238 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
4240 * gimple-match.h (gimple_match_op): New class.
4241 (mprts_hook): Replace parameters with a gimple_match_op *.
4242 (maybe_build_generic_op): Likewise.
4243 (gimple_simplified_result_is_gimple_val): Replace parameters with
4244 a const gimple_match_op *.
4245 (gimple_simplify): Replace code_helper * and tree * parameters with
4246 a gimple_match_op * parameter.
4247 (gimple_resimplify1): Replace code_helper *, tree and tree *
4248 parameters with a gimple_match_op * parameter.
4249 (gimple_resimplify2): Likewise.
4250 (gimple_resimplify3): Likewise.
4251 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
4252 parameters with a gimple_match_op * parameter.
4253 * gimple-match-head.c (gimple_simplify): Change prototypes of
4254 auto-generated functions to take a gimple_match_op * instead of
4255 separate code_helper * and tree * parameters. Make the same
4256 change in the top-level overload and update calls to the
4257 gimple_resimplify routines. Update calls to the auto-generated
4258 functions and to maybe_push_res_to_seq in the publicly-facing
4259 operation-specific gimple_simplify overloads.
4260 (gimple_match_op::MAX_NUM_OPS): Define.
4261 (gimple_resimplify1): Replace rcode and ops with a single res_op
4262 parameter. Update call to gimple_simplify.
4263 (gimple_resimplify2): Likewise.
4264 (gimple_resimplify3): Likewise.
4265 (mprts_hook): Replace parameters with a gimple_match_op *.
4266 (maybe_build_generic_op): Likewise.
4267 (build_call_internal): Replace type, nargs and ops with
4268 a gimple_match_op *.
4269 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
4270 with a single gimple_match_op *. Update calls to mprts_hook,
4271 build_call_internal and gimple_simplified_result_is_gimple_val.
4272 Factor out code that is common to the tree_code and combined_fn cases.
4273 * genmatch.c (expr::gen_transform): Replace tem_code and
4274 tem_ops with a gimple_match_op called tem_op. Update calls
4275 to the gimple_resimplify functions and maybe_push_res_to_seq.
4276 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
4277 res_ops. Update call to the gimple_resimplify functions.
4278 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
4279 (decision_tree::gen): Make the functions take a gimple_match_op *
4280 called res_op instead of separate res_code and res_ops parameters.
4281 Update call accordingly.
4282 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
4283 and ops with a single res_op parameter. Update calls to
4284 maybe_build_generic_op and maybe_push_res_to_seq.
4285 (fold_stmt_1): Update calls to gimple_simplify and
4286 replace_stmt_with_simplification.
4287 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
4288 and gimple_simplified_result_is_gimple_val.
4289 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
4291 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
4292 with a gimple_match_op *.
4293 (vn_nary_build_or_lookup): Likewise. Update call to
4294 vn_nary_build_or_lookup_1.
4295 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
4296 gimple_match_op *. Update calls to the gimple_resimplify routines
4297 and to gimple_simplified_result_is_gimple_val.
4298 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
4299 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
4300 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
4301 (visit_nary_op): Likewise.
4302 (visit_reference_op_load): Likewise.
4304 2018-05-23 Luis Machado <luis.machado@linaro.org>
4306 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
4307 modifier for printing the step amount.
4309 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
4312 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
4315 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
4317 * doc/sourcebuild.texi (Endianness): New subsubsection.
4319 2018-05-23 Luis Machado <luis.machado@linaro.org>
4321 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4322 <prefetch_dynamic_strides>: New const bool field.
4323 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4324 prefetch_dynamic_strides.
4325 (exynosm1_prefetch_tune): Likewise.
4326 (thunderxt88_prefetch_tune): Likewise.
4327 (thunderx_prefetch_tune): Likewise.
4328 (thunderx2t99_prefetch_tune): Likewise.
4329 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
4331 (aarch64_override_options_internal): Update to set
4332 PARAM_PREFETCH_DYNAMIC_STRIDES.
4333 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4334 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4335 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4336 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4337 prefetch-dynamic-strides setting.
4339 2018-05-23 Luis Machado <luis.machado@linaro.org>
4341 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4342 <minimum_stride>: New const int field.
4343 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4344 minimum_stride field defaulting to -1.
4345 (exynosm1_prefetch_tune): Likewise.
4346 (thunderxt88_prefetch_tune): Likewise.
4347 (thunderx_prefetch_tune): Likewise.
4348 (thunderx2t99_prefetch_tune): Likewise.
4349 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
4350 <default_opt_level>: Set to 3.
4351 (aarch64_override_options_internal): Update to set
4352 PARAM_PREFETCH_MINIMUM_STRIDE.
4353 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4354 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4355 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4356 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4357 stride is constant and is below the minimum stride threshold.
4359 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4361 * config/arm/arm-cpus.in (mode26): Delete.
4362 (armv4): Delete mode26 reference.
4363 * config/arm/arm.c (arm_configure_build_target): Delete use of
4366 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4368 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
4370 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
4371 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
4373 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
4374 (floatunsdidf2): Ditto.
4376 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
4378 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
4379 (fixuns_trunc<mode>si2_avx512f): Ditto.
4380 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
4381 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
4382 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
4384 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4386 PR rtl-optimization/79985
4387 * df-scan.c (df_insn_refs_collect): Remove special case for
4388 global registers and asm statements.
4390 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
4392 * extend.texi (Global Register Variables): Rewrite the bullet list.
4393 Note that the register is available for allocation. Note that access
4394 via inline asm must use constraints. Add note about async-signal
4395 handlers. Remove paragraph about automagic register selection.
4397 2018-05-23 Richard Biener <rguenther@suse.de>
4399 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
4400 of fixed offset from memset VN.
4402 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
4404 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
4406 (alloc_cand_and_find_basis): Initialize first_interp field.
4407 (slsr_process_mul): Modify first_interp field.
4408 (slsr_process_add): Likewise.
4409 (slsr_process_cast): Modify first_interp field for each new
4411 (slsr_process_copy): Likewise.
4412 (dump_candidate): Dump first_interp field.
4413 (replace_mult_candidate): Process all interpretations, not just
4415 (replace_rhs_if_not_dup): Likewise.
4416 (replace_one_candidate): Likewise.
4418 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
4420 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
4422 (aarch64_needs_frame_chain): New function.
4423 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
4425 2018-05-23 Sudakshina Das <sudi.das@arm.com>
4428 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
4429 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
4430 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
4431 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
4432 as true for strict-align.
4433 (aarch64_can_inline_p): Perform checks even when callee has no
4434 attributes to check for strict alignment.
4435 * doc/extend.texi (AArch64 Function Attributes): Document
4437 * doc/invoke.texi: (AArch64 Options): Likewise.
4439 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
4441 PR tree-optimization/85853
4442 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
4443 the handling of the root of the node to...
4444 (vect_slp_analyze_node_operations_1): ...this new function,
4445 and run the whole thing with the child nodes' def types
4446 set according to their SLP node's def type.
4448 2018-05-23 Richard Biener <rguenther@suse.de>
4451 * tree-data-ref.c (create_runtime_alias_checks): Defer
4452 and ignore overflow warnings.
4454 2018-05-23 Yury Gribov <tetra2005@gmail.com>
4456 PR tree-optimization/85822
4457 * tree-vrp.c (is_masked_range_test): Fix handling of negative
4460 2018-05-23 Richard Biener <rguenther@suse.de>
4462 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4463 memset constants via native_interpret_expr.
4465 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
4468 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
4470 (cgraph_node::create_alias): Likewise.
4471 (cgraph_node::get_availability): Check ifunc_resolver instead
4472 of looking up ifunc attribute.
4473 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4474 * varasm.c (do_assemble_alias): Likewise.
4475 (assemble_alias): Likewise.
4476 (default_binds_local_p_3): Likewise.
4477 * cgraph.h (cgraph_node): Add ifunc_resolver.
4478 (cgraph_node::only_called_directly_or_aliased_p): Return false
4480 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
4482 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
4483 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
4484 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
4485 instead of looking up ifunc attribute.
4487 2018-05-22 Luis Machado <luis.machado@linaro.org>
4489 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
4491 2018-05-22 Martin Sebor <msebor@redhat.com>
4494 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
4495 only when expasion succeeds.
4496 (expand_builtin_strcmp): Same.
4497 (expand_builtin_strncmp): Same.
4499 2018-05-22 Martin Sebor <msebor@redhat.com>
4501 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
4503 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4504 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4506 * config/aarch64/aarch64-ldpstp.md: Replace uses of
4507 aarch64_mem_pair_operand with memory_operand and delete operand swapping
4509 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4510 Add check for legitimate_address.
4511 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
4512 (aarch64_swap_ldrstr_operands): New.
4513 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
4516 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
4517 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4519 * config/aarch64/aarch64.md: New patterns to generate stp
4521 (store_pair_sw, store_pair_dw): New patterns to generate stp for
4522 single words and double words.
4523 (load_pair_sw, load_pair_dw): Likewise.
4524 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4526 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4528 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4529 for different mode ldpstp and add peephole for merged zero stores.
4531 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4533 (aarch64_gen_store_pair): Rename calls to match new patterns.
4534 (aarch64_gen_load_pair): Rename calls to match new patterns.
4535 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4536 (load_pair<DREG:mode><DREG2:mode>): ... This.
4537 (store_pair<mode>): Rename to...
4538 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4539 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4541 (V_INT_EQUIV): Handle SImode.
4542 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4545 2018-05-22 Martin Sebor <msebor@redhat.com>
4548 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4549 or ajust the presumed bound on an operation to avoid unnecessary
4552 2018-05-22 Martin Sebor <msebor@redhat.com>
4554 PR tree-optimization/85826
4555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4556 assuming that a DECL necesarily has a constant size.
4558 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4561 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4563 2018-05-22 Richard Biener <rguenther@suse.de>
4565 PR tree-optimization/85834
4566 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4567 non-constant and non-zero memset arguments.
4569 2018-05-22 Martin Liska <mliska@suse.cz>
4572 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4574 2018-05-22 Richard Biener <rguenther@suse.de>
4576 PR tree-optimization/85863
4577 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4578 comparisons when vectype is specified.
4579 (vectorizable_condition): Do not specify vectype for
4580 vect_is_simple_cond when SLP vectorizing.
4582 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4585 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4586 define __ibm128 as long double.
4587 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4588 as a distinct type when IEEE 128-bit support is enabled.
4589 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4590 128-bit types to use the correct functions.
4591 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4592 convert between 128-bit floating point types that have different
4593 modes but the same representation, instead of using gen_lowpart to
4595 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4597 (IFKF_reg): New attributes to give the register constraints for
4599 (extend<mode>tf2_internal): New insns to mark an explicit
4600 conversion between 128-bit floating point types that have a
4601 different mode but share the same representation.
4603 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4605 PR tree-optimization/85814
4606 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4607 a null return from get_strinfo when unsharing the next
4608 strinfo in the chain.
4610 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4613 * varasm.c (weak_finish): Clean up weak_decls.
4615 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4617 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4618 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4619 UNSPEC_UADALP values.
4620 * config/aarch64/iterators.md (ABAL): New int iterator.
4623 (sur): Add mappings for the above.
4624 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4626 (aarch64_<sur>abal<mode>_4): Likewise.
4627 (aarch64_<sur>adalp<mode>_3): Likewise.
4628 (<sur>sadv16qi): New define_expand.
4630 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4632 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4633 (*movdf_internal): Ditto.
4634 (*rcpsf2_sse): Ditto.
4635 (*rsqrtsf2_sse): Ditto.
4636 (*sqrt<mode>2_sse): Ditto.
4638 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4640 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4642 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4643 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4644 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4645 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4647 * config/aarch64/arm_neon.h: Likewise.
4648 * config/aarch64/iterators.md (VQ_I): New.
4650 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4652 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4654 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4657 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4659 * config/nds32/constraints.md (S): New constraint.
4660 * config/nds32/nds32.md (call_internal): Use constraint S.
4661 (call_value_internal): Likewise.
4662 (sibcall_internal): Likewise.
4663 (sibcall_value_internal): Likewise.
4665 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4666 Chung-Ju Wu <jasonwucj@gmail.com>
4668 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4671 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4672 Chung-Ju Wu <jasonwucj@gmail.com>
4674 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4675 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4676 (nds32_rtx_costs_impl): Simplify.
4677 (nds32_address_cost_impl): Simplify.
4678 (nds32_init_rtx_costs): New function.
4679 (nds32_rtx_costs_speed_prefer): Likewise.
4680 (nds32_rtx_costs_size_prefer): Likewise.
4681 (nds32_address_cost_speed_prefer): Likewise.
4682 (nds32_address_cost_speed_fwprop): Likewise.
4683 (nds32_address_cost_size_prefer): Likewise.
4684 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4685 * config/nds32/nds32.c (nds32_option_override): Use
4686 nds32_init_rtx_costs function.
4688 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4690 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4691 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4692 (TARGET_PIPELINE_N8): Likewise.
4693 (TARGET_PIPELINE_N10): Likewise.
4694 (TARGET_PIPELINE_N13): Likewise.
4695 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4697 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4699 * config/nds32/nds32-fpu.md: Update copyright year.
4701 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4703 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4705 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4707 * config/nds32/nds32.c
4708 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4709 * config/nds32/nds32.opt (minline-asm-r15): New option.
4711 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4713 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4715 * config/nds32/nds32.md (abssi2): New pattern.
4717 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4719 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4720 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4721 Merge insn pattern from sse_cvtsi2ss<round_name> and
4722 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4723 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4724 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4725 using SWI48 mode iterator.
4726 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4727 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4728 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4729 pattern from sse_cvttss2si<round_saeonly_name>
4730 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4731 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4732 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4733 using SWI48 mode iterator.
4734 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4735 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4736 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4737 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4738 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4739 using SWI48 mode iterator.
4740 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4741 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4742 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4743 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4744 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4745 SWI48 mode iterator.
4746 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4747 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4748 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4749 pattern from sse_cvttsd2si<round_saeonly_name>
4750 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4752 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4754 * config/nds32/nds32-md-auxiliary.c
4755 (nds32_valid_smw_lwm_base_p): Refine.
4756 (nds32_output_smw_single_word): Refine.
4757 (nds32_output_smw_double_word): New.
4758 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4760 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4762 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4763 (nds32_output_stack_pop): Refine.
4764 (nds32_expand_unaligned_load): Refine.
4765 (nds32_expand_unaligned_store): Refine.
4767 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4768 Chung-Ju Wu <jasonwucj@gmail.com>
4770 * config/nds32/constants.md: Add TP_REGNUM constant.
4771 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4772 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4774 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4775 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4776 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4777 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4778 and PIC code generation.
4779 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4781 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4783 * config/nds32/nds32.md: Support TLS and PIC.
4784 * config/nds32/nds32.c: Support TLS and PIC.
4785 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4786 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4789 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4791 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4792 mode with E_ prefix.
4794 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4795 Chung-Ju Wu <jasonwucj@gmail.com>
4797 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4798 * config/nds32/nds32-md-auxiliary.c
4799 (symbolic_reference_mentioned_p): New.
4800 (nds32_legitimize_ict_address): New.
4801 (nds32_expand_ict_move): New.
4802 (nds32_indirect_call_referenced_p): New.
4803 (nds32_symbol_binds_local_p): Delete.
4804 (nds32_long_call_p): Modify.
4805 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4806 * config/nds32/nds32-protos.h
4807 (symbolic_reference_mentioned_p): Declare.
4808 (nds32_legitimize_ict_address): Declare.
4809 (nds32_expand_ict_move): Declare.
4810 (nds32_indirect_call_referenced_p): Declare.
4811 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4812 (nds32_relax_group): Use nds32_ict_const_p as condition.
4813 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4814 (nds32_asm_file_start): Output ict_model directive in asm code.
4815 (nds32_legitimate_address_p): Consider indirect call.
4816 (nds32_print_operand): Consider indirect call.
4817 (nds32_print_operand_address): Consider indirect call.
4818 (nds32_insert_attributes): Handle "indirect_call" attribute.
4819 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4820 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4821 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4822 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4823 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4824 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4825 (TARGET_ICT_MODEL_SMALL): Define.
4826 (TARGET_ICT_MODEL_LARGE): Define.
4827 * config/nds32/nds32.md (movsi): Consider ict model.
4828 (call, call_value): Consider ict model.
4829 (sibcall, sibcall_value): Consider ict model.
4830 * config/nds32/nds32.opt (mict-model): New option.
4831 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4834 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4835 Monk Chiang <sh.chiang04@gmail.com>
4836 Jim Wilson <jimw@sifive.com>
4838 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4839 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4840 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4841 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4842 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4843 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4844 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4845 compute save_libcall_adjustment properly.
4846 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4847 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4848 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4849 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4850 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4851 (ABI_SPEC): Handle mabi=ilp32e.
4852 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4853 (RVE): Add RVE mask.
4854 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4855 <-march>: Add rv32e as an example.
4857 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4860 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4861 (intra_create_variable_infos): Handle C++ constructors.
4863 2018-05-18 Martin Liska <mliska@suse.cz>
4865 * passes.def: Remove a redundant pass.
4867 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4870 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4872 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4874 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4876 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4878 (armv2, armv2a, armv3, armv3m): Delete architectures.
4879 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4880 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4881 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4883 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4884 (*mulsidi3adddi): Likewise.
4885 (mulsidi3): Likewise.
4886 (*mulsidi3_nov6): Likewise.
4887 (umulsidi3): Likewise.
4888 (umulsidi3_nov6): Likewise.
4889 (umaddsidi4): Likewise.
4890 (*umulsidi3adddi): Likewise.
4891 (smulsi3_highpart): Likewise.
4892 (*smulsi3_highpart_nov6): Likewise.
4893 (umulsi3_highpart): Likewise.
4894 (*umulsi3_highpart_nov6): Likewise.
4895 * config/arm/arm.h (arm_arch3m): Delete.
4896 * config/arm/arm.c (arm_arch3m): Delete.
4897 (arm_option_override_internal): Update armv3-related comment.
4898 (arm_configure_build_target): Delete use of isa_bit_mode32.
4899 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4900 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4901 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4903 (mulusa3): Likewise.
4904 * config/arm/arm-protos.h (arm_arch3m): Delete.
4905 * config/arm/arm-tables.opt: Regenerate.
4906 * config/arm/arm-tune.md: Likewise.
4907 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4908 deleted architectures.
4910 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4912 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4913 (armv5t, armv5te): New features.
4914 (ARMv5, ARMv5e): Delete fgroups.
4915 (ARMv5t, ARMv5te): Adjust for above changes.
4917 (armv5, armv5e): Delete arches.
4918 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4920 (*call_reg_arm): Likewise.
4921 (*call_value_reg_armv5): Likewise.
4922 (*call_value_reg_arm): Likewise.
4923 (*call_symbol): Likewise.
4924 (*call_value_symbol): Likewise.
4925 (*sibcall_insn): Likewise.
4926 (*sibcall_value_insn): Likewise.
4928 (prefetch): Likewise.
4929 (define_split and define_peephole2 dependent on arm_arch5):
4931 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4933 (TARGET_ARM_QBIT): Likewise.
4934 (TARGET_DSP_MULTIPLY): Likewise.
4935 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4936 (arm_arch5, arm_arch5e): Delete.
4937 (arm_arch5t, arm_arch5te): Declare.
4938 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4939 (arm_arch5t): Declare.
4940 (arm_option_reconfigure_globals): Update for the above.
4941 (arm_options_perform_arch_sanity_checks): Update comment, replace
4942 use of arm_arch5 with arm_arch5t.
4943 (use_return_insn): Likewise.
4944 (arm_emit_call_insn): Likewise.
4945 (output_return_instruction): Likewise.
4946 (arm_final_prescan_insn): Likewise.
4947 (arm_coproc_builtin_available): Likewise.
4948 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4949 arm_arch5e with arm_arch5t and arm_arch5te.
4950 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4951 (arm_arch5t, arm_arch5te): Declare.
4952 * config/arm/arm-tables.opt: Regenerate.
4953 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4954 * config/arm/t-multilib: Likewise.
4955 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4956 instead of arm_arch5.
4957 (*call_reg_thumb1): Likewise.
4958 (*call_value_reg_thumb1_v5): Likewise.
4959 (*call_value_reg_thumb1): Likewise.
4960 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4962 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4964 2018-05-18 Martin Liska <mliska@suse.cz>
4966 PR gcov-profile/84846
4967 * doc/gcov.texi: Document -t option of gcov tool.
4969 2018-05-18 Martin Liska <mliska@suse.cz>
4971 PR gcov-profile/84846
4972 * gcov.c (print_usage): Add new -t option.
4973 (process_args): Handle the option.
4974 (generate_results): Use stdout as output when requested by
4977 2018-05-18 Martin Liska <mliska@suse.cz>
4979 PR gcov-profile/84846
4980 * coverage.c (coverage_init): Write PWD to .gcno file.
4981 * doc/gcov.texi: Document how working directory is printed.
4982 * gcov-dump.c (dump_gcov_file): Print PWD.
4983 * gcov.c (output_intermediate_file): Likewise.
4984 (read_graph_file): Read PWD string.
4985 (output_lines): Print PWD.
4987 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4990 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4991 for retval and return false if all args to phi are zero.
4993 2018-05-18 Richard Biener <rguenther@suse.de>
4995 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4997 (evrp_dom_walker::before_dom_children): Call it.
4999 2018-05-18 Richard Biener <rguenther@suse.de>
5001 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5002 results when processing array refs with variable index.
5004 2018-05-18 Toon Moene <toon@moene.org>
5006 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5007 directly after that of -floop-interchange. Indicate that both
5008 options are enabled by default when specifying -O3.
5010 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5012 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5013 iterator. Delete separate integer-mode vec_set<mode> expander.
5014 (aarch64_simd_vec_setv2di): Delete.
5015 (vec_setv2di): Delete.
5016 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5017 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5018 the "w, r" alternative.
5020 2018-05-18 Martin Liska <mliska@suse.cz>
5022 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5023 * tree-pass.h (make_pass_lower_switch_O0): New function.
5024 * tree-switch-conversion.c (node_has_low_bound): Remove.
5025 (node_has_high_bound): Likewise.
5026 (node_is_bounded): Likewise.
5027 (class pass_lower_switch): Make it a template type and create
5029 (pass_lower_switch::execute): Add template argument.
5030 (make_pass_lower_switch): New function.
5031 (make_pass_lower_switch_O0): New function.
5032 (do_jump_if_equal): Remove.
5033 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5034 all the hard work to tree optimization passes.
5036 2018-05-18 Martin Liska <mliska@suse.cz>
5038 * dbgcnt.c (limit_low): Renamed from limit.
5039 (limit_high): New variable.
5040 (dbg_cnt_is_enabled): Check for upper limit.
5041 (dbg_cnt): Adjust dumping.
5042 (dbg_cnt_set_limit_by_index): Add new argument for high
5044 (dbg_cnt_set_limit_by_name): Likewise.
5045 (dbg_cnt_process_single_pair): Parse new format.
5046 (dbg_cnt_process_opt): Use strtok.
5047 (dbg_cnt_list_all_counters): Remove 'value' and add
5049 * doc/invoke.texi: Document changes.
5051 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5053 * doc/sourcebuild.texi (scalar_all_fma): Document.
5054 * tree.def (FMA_EXPR): Delete.
5055 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5056 * internal-fn.c (ternary_direct): New macro.
5057 (expand_ternary_optab_fn): Likewise.
5058 (direct_ternary_optab_supported_p): Likewise.
5059 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5060 * builtins.c (fold_builtin_fma): Delete.
5061 (fold_builtin_3): Don't call it.
5062 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5063 * expr.c (expand_expr_real_2): Likewise.
5064 * fold-const.c (operand_equal_p): Likewise.
5065 (fold_ternary_loc): Likewise.
5066 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5067 * gimple.c (DEFTREECODE): Likewise.
5068 * gimplify.c (gimplify_expr): Likewise.
5069 * optabs-tree.c (optab_for_tree_code): Likewise.
5070 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5071 * tree-eh.c (operation_could_trap_p): Likewise.
5072 (stmt_could_throw_1_p): Likewise.
5073 * tree-inline.c (estimate_operator_cost): Likewise.
5074 * tree-pretty-print.c (dump_generic_node): Likewise.
5075 (op_code_prio): Likewise.
5076 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5077 * tree-ssa-operands.c (get_expr_operands): Likewise.
5078 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5079 * fold-const-call.h (fold_fma): Delete.
5080 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5081 CFN_FNMA and CFN_FNMS.
5083 * genmatch.c (combined_fn): New enum.
5084 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5085 (commutative_op): New function.
5086 (commutate): Use it. Handle more than 2 operands.
5087 (dt_operand::gen_gimple_expr): Use commutative_op.
5088 (parser::parse_expr): Allow :c to be used with non-binary
5089 operators if the commutative operand is known.
5090 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5091 CFN_FMS, CFN_FNMA and CFN_FNMS.
5092 (backprop::process_assign_use): Remove FMA_EXPR handling.
5093 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5094 (gen_hsa_fma): New function.
5095 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5096 IFN_FNMA and IFN_FNMS.
5097 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5098 * gimple-fold.h (follow_all_ssa_edges): Declare.
5099 * gimple-fold.c (follow_all_ssa_edges): New function.
5100 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5101 gimple_build interface and use follow_all_ssa_edges to fold the result.
5102 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5103 instead of checking for optabs directly.
5104 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5105 rather than FMA_EXPRs.
5106 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5107 call to IFN_FMA instead of an FMA_EXPR.
5109 2018-05-17 Jim Wilson <jimw@sifive.com>
5111 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5112 sign extended promoted subreg, and the range does not have the sign bit
5113 set, then do a sign extend.
5115 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5116 test, check for sign extended subreg and/or constant operands, and
5117 do a sign extend in that case.
5119 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5121 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5122 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5124 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5125 Change logics_shift_reg to logics_shift_imm.
5126 (thunderx2t99_fp_loadpair_basic): Delete.
5127 (thunderx2t99_fp_storepair_basic): Delete.
5128 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5129 (thunderx2t99_asimd_polynomial): Delete.
5130 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5131 and neon_fp_mul_d_scalar_q.
5132 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5133 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5134 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5135 (thunderx2t99_asimd_lut): Add missing tbl types.
5136 (thunderx2t99_asimd_ext): Delete.
5137 (thunderx2t99_asimd_load1_1_mult): Delete.
5138 (thunderx2t99_asimd_load1_2_mult): Delete.
5139 (thunderx2t99_asimd_load1_ldp): New.
5140 (thunderx2t99_asimd_load1): New.
5141 (thunderx2t99_asimd_load2): Add missing *load2* types.
5142 (thunderx2t99_asimd_load3): New.
5143 (thunderx2t99_asimd_load4): New.
5144 (thunderx2t99_asimd_store1_1_mult): Delete.
5145 (thunderx2t99_asimd_store1_2_mult): Delete.
5146 (thunderx2t99_asimd_store2_mult): Delete.
5147 (thunderx2t99_asimd_store2_onelane): Delete.
5148 (thunderx2t99_asimd_store_stp): New.
5149 (thunderx2t99_asimd_store1): New.
5150 (thunderx2t99_asimd_store2): New.
5151 (thunderx2t99_asimd_store3): New.
5152 (thunderx2t99_asimd_store4): New.
5154 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
5156 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
5157 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
5159 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
5160 Segher Boessenkool <segher@kernel.crashing.org>
5163 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
5166 2018-05-17 Richard Biener <rguenther@suse.de>
5168 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
5169 for pruning loop and prune defs feeding only already visited PHIs.
5171 2018-05-17 Richard Biener <rguenther@suse.de>
5173 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
5175 2018-05-17 Bin Cheng <bin.cheng@arm.com>
5176 Richard Biener <rguenther@suse.de>
5178 PR tree-optimization/85793
5179 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
5180 for VMAT_ELEMENTWISE.
5182 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5184 * internal-fn.h (lookup_internal_fn): Declare
5185 * internal-fn.c (lookup_internal_fn): New function.
5186 * gimple.c (gimple_build_call_from_tree): Handle calls to
5188 * gimple-pretty-print.c (dump_gimple_call): Print "." before
5189 internal function names.
5190 * tree-pretty-print.c (dump_generic_node): Likewise.
5191 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
5193 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
5195 * gimple-fold.h (gimple_build): Make the function forms take
5196 combined_fn rather than built_in_function.
5197 (gimple_simplify): Likewise.
5198 * gimple-match-head.c (gimple_simplify): Likewise.
5199 * gimple-fold.c (gimple_build): Likewise.
5200 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
5201 rather than gimple_build_call_internal.
5202 (get_initial_defs_for_reduction): Likewise.
5203 (vect_create_epilog_for_reduction): Likewise.
5204 (vectorizable_live_operation): Likewise.
5206 2018-05-17 Martin Liska <mliska@suse.cz>
5208 * gimple-ssa-sprintf.c (format_directive): Do not use
5209 space in between 'G_' and '('.
5211 2018-05-17 Jakub Jelinek <jakub@redhat.com>
5214 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5215 even if the mask is not all ones.
5218 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
5220 (ix86_gimple_fold_builtin): Likewise.
5223 * config/i386/i386.c: Include tree-vector-builder.h.
5224 (ix86_vector_shift_count): New function.
5225 (ix86_fold_builtin): Fold shift builtins by scalar count.
5226 (ix86_gimple_fold_builtin): Likewise.
5228 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
5229 _mm512_setzero): New intrinsics.
5231 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
5232 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5234 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5235 code generation for cases where splatting a value is not useful.
5236 * simplify-rtx.c (simplify_ternary_operation): Simplify
5237 vec_merge across a vec_duplicate and a paradoxical subreg forming
5238 a vector mode to a vec_concat.
5240 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
5242 * config.gcc: Support "goldmont-plus".
5243 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5245 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5246 PROCESSOR_GOLDMONT_PLUS.
5247 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
5248 (processor_target_table): Add "goldmont-plus".
5249 (PTA_GOLDMONT_PLUS): Define.
5250 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
5251 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
5252 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
5253 (fold_builtin_cpu): Add "goldmont-plus".
5254 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
5255 (ix86_option_override_internal): Add "goldmont-plus".
5256 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
5257 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
5258 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
5259 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
5261 2018-05-17 Richard Biener <rguenther@suse.de>
5263 PR tree-optimization/85757
5264 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
5265 remove defs that only feed that PHI from further processing.
5267 2018-05-16 Jim Wilson <jimw@sifive.com>
5269 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5271 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
5273 2018-05-16 Mark Wielaard <mark@klomp.org>
5275 * dwarf2out.c (count_index_strings): New function.
5276 (output_indirect_strings): Call count_index_strings and generate
5277 header for dwarf_version >= 5.
5279 2018-05-16 Mark Wielaard <mark@klomp.org>
5281 * dwarf2out.c (dwarf_FORM): New function.
5282 (set_indirect_string): Use dwarf_FORM.
5283 (reset_indirect_string): Likewise.
5284 (size_of_die): Likewise.
5285 (value_format): Likewise.
5286 (output_die): Likewise.
5287 (add_skeleton_AT_string): Likewise.
5288 (output_macinfo_op): Likewise.
5289 (index_string): Likewise.
5290 (output_index_string_offset): Likewise.
5291 (output_index_string): Likewise.
5292 (count_index_strings): Likewise.
5294 2018-05-16 Carl Love <cel@us.ibm.com>
5296 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5297 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
5299 2018-05-16 Martin Jambor <mjambor@suse.cz>
5301 * ipa-prop.c (ipa_free_all_edge_args): Remove.
5302 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
5304 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
5306 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
5307 (fnma<mode>4): Likewise.
5308 (fms<mode>4): Likewise.
5309 (fnms<mode>4): Likewise.
5310 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
5311 (aarch64_fnma<mode>4): Likewise.
5312 (aarch64_fms<mode>4): Likewise.
5313 (aarch64_fnms<mode>4): Likewise.
5314 (aarch64_fnmadd<mode>4): Likewise.
5316 2018-05-16 Jason Merrill <jason@redhat.com>
5318 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
5320 2018-05-16 Richard Biener <rguenther@suse.de>
5322 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
5323 (dump_stmt_cost): Declare.
5324 (add_stmt_cost): Dump cost we add.
5325 (add_stmt_costs): New function.
5326 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
5328 (vect_analyze_stmt): Adjust prototype.
5329 (vectorizable_condition): Likewise.
5330 (vectorizable_live_operation): Likewise.
5331 (vectorizable_reduction): Likewise.
5332 (vectorizable_induction): Likewise.
5333 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
5334 cost vector to pass to vectorizable_ and record afterwards.
5335 (vect_model_reduction_cost): Take cost vector argument and adjust.
5336 (vect_model_induction_cost): Likewise.
5337 (vectorizable_reduction): Likewise.
5338 (vectorizable_induction): Likewise.
5339 (vectorizable_live_operation): Likewise.
5340 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
5341 SLP_TREE_NUMBER_OF_VEC_STMTS.
5342 (vect_analyze_slp_cost_1): Remove.
5343 (vect_analyze_slp_cost): Likewise.
5344 (vect_slp_analyze_node_operations): Take visited args and
5345 a target cost vector. Avoid processing already visited stmt sets.
5346 (vect_slp_analyze_operations): Use a local cost vector to gather
5347 costs and register those of non-discarded instances.
5348 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
5349 (vect_schedule_slp_instance): Remove copying of
5350 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
5352 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
5353 adding cost. Record cost entry location.
5354 (vect_prologue_cost_for_slp_op): Function to compute cost of
5355 a constant or invariant generated for SLP vect in the prologue,
5356 split out from vect_analyze_slp_cost_1.
5357 (vect_model_simple_cost): Make static. Adjust for SLP costing.
5358 (vect_model_promotion_demotion_cost): Likewise.
5359 (vect_model_store_cost): Likewise, make static.
5360 (vect_model_load_cost): Likewise.
5361 (vectorizable_bswap): Add cost vector arg and adjust.
5362 (vectorizable_call): Likewise.
5363 (vectorizable_simd_clone_call): Likewise.
5364 (vectorizable_conversion): Likewise.
5365 (vectorizable_assignment): Likewise.
5366 (vectorizable_shift): Likewise.
5367 (vectorizable_operation): Likewise.
5368 (vectorizable_store): Likewise.
5369 (vectorizable_load): Likewise.
5370 (vectorizable_condition): Likewise.
5371 (vectorizable_comparison): Likewise.
5372 (can_vectorize_live_stmts): Likewise.
5373 (vect_analyze_stmt): Likewise.
5374 (vect_transform_stmt): Adjust calls to vectorizable_*.
5375 * tree-vectorizer.c: Include gimple-pretty-print.h.
5376 (dump_stmt_cost): New function.
5378 2018-05-16 Richard Biener <rguenther@suse.de>
5380 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
5381 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
5382 * tree-ssa-dse.c: Include tree-ssa-loop.h.
5383 (check_name): New callback.
5384 (dse_classify_store): Track cycles via a visited bitmap of PHI
5385 defs and simplify handling of in-loop and across loop dead stores
5386 and properly fail for loop-variant refs. Handle byte-tracking with
5387 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
5390 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
5392 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
5393 (vect_get_mask_type_for_stmt): Likewise.
5394 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
5396 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
5397 to determine the statement's vector type and the vector type that
5398 should be used for calculating nunits. Deal with cases in which
5399 the type has to be deferred.
5400 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
5401 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
5402 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
5403 (vect_determine_vf_for_stmt): New functions, split out from...
5404 (vect_determine_vectorization_factor): ...here.
5405 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
5406 (vect_get_mask_type_for_stmt): New functions, split out from
5407 vect_determine_vectorization_factor.
5409 2018-05-16 Richard Biener <rguenther@suse.de>
5411 * tree-cfg.c (verify_gimple_assign_ternary): Properly
5412 verify the [VEC_]COND_EXPR embedded comparison.
5414 2018-05-15 Martin Sebor <msebor@redhat.com>
5416 PR tree-optimization/85753
5417 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
5418 RECORD_TYPE in addition to ARRAY_TYPE.
5420 2018-05-15 Martin Sebor <msebor@redhat.com>
5423 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
5425 2018-05-15 Richard Biener <rguenther@suse.de>
5427 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
5428 add by_clobber_p one. Change algorithm to collect all defs
5429 representing uses we need to walk and try reducing them to
5430 a single one before failing.
5431 (dse_dom_walker::dse_optimize_stmt): Adjust.
5433 2018-05-13 Mark Wielaard <mark@klomp.org>
5435 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
5436 (size_of_loc_descr): Likewise.
5437 (output_loc_operands): Likewise.
5438 (output_loc_operands_raw): Likewise.
5439 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
5440 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
5441 (hash_loc_operands): Likewise.
5442 (compare_loc_operands): Likewise.
5444 2018-05-14 Mark Wielaard <mark@klomp.org>
5446 * dwarf2out.c (count_index_addrs): New function.
5447 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
5449 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5451 PR tree-optimization/83648
5452 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
5453 return value as malloc candidate.
5455 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5458 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
5459 param as true in call to suggest_attribute.
5461 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
5463 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
5464 -mreadonly-in-sdata.
5466 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5468 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
5470 (aarch64_crypto_aesd_fused): Likewise.
5472 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
5474 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
5475 (movsi_aarch64): Likewise.
5476 (load_pairsi): Likewise.
5477 (load_pairdi): Likewise.
5478 (store_pairsi): Likewise.
5479 (store_pairdi): Likewise.
5480 (load_pairsf): Likewise.
5481 (load_pairdf): Likewise.
5482 (store_pairsf): Likewise.
5483 (store_pairdf): Likewise.
5484 (zero_extend): Likewise.
5485 (trunc): Swap alternatives.
5486 (fcvt_target): Add '?' to prefer w over r.
5488 2018-05-14 Jakub Jelinek <jakub@redhat.com>
5491 * config/i386/i386.md: Disallow non-commutative arithmetics in
5492 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
5493 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
5494 in the peephole2 before it.
5496 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
5498 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
5499 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
5500 (ix86_handle_option): Handle -mcldemote.
5501 * config.gcc: New header.
5502 * config/i386/cldemoteintrin.h: New file.
5503 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
5504 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
5506 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5507 OPTION_MASK_ISA_CLDEMOTE.
5508 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
5509 (ix86_valid_target_attribute_inner_p): Ditto.
5510 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
5511 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
5512 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
5513 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
5514 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
5516 * config/i386/i386.opt: Add -mcldemote.
5517 * config/i386/x86intrin.h: New header.
5518 * doc/invoke.texi: Add -mcldemote.
5520 2018-05-14 Richard Biener <rguenther@suse.de>
5522 * doc/match-and-simplify.texi: Adjust :s documentation.
5524 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5526 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5527 intended memcpy size.
5528 (REORDER_45): Likewise.
5530 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5532 * sort.cc: New file.
5533 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5534 * vec.c (qsort_chk): Use gcc_qsort.
5535 * Makefile.in (OBJS-libcommon): Add sort.o.
5536 (build/sort.o): New target. Use it...
5537 (BUILD_RTL): ... here, and...
5538 (build/gencfn-macros): ... here, and...
5539 (build/genmatch): ... here.
5541 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5542 Chung-Ju Wu <jasonwucj@gmail.com>
5544 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5545 * config/nds32/nds32-graywolf.md: New file.
5546 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5547 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5549 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5550 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5551 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5552 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5553 * config/nds32/pipelines.md: Include n15 settings.
5555 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5556 Chung-Ju Wu <jasonwucj@gmail.com>
5558 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5559 * config/nds32/nds32-n13.md: New file.
5560 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5561 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5563 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5564 * config/nds32/nds32.md (pipeline_model): Add n13.
5565 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5566 * config/nds32/pipelines.md: Include n13 settings.
5568 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5569 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5572 * config/nds32/nds32-n10.md: New file.
5573 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5574 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5576 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5577 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5578 * config/nds32/nds32.md (pipeline_model): Add n10.
5579 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5580 * config/nds32/pipelines.md: Include n10 settings.
5582 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5583 Kito Cheng <kito.cheng@gmail.com>
5584 Chung-Ju Wu <jasonwucj@gmail.com>
5586 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5587 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5588 Add enum values for DSP extension instructions.
5589 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5591 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5592 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5594 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5595 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5596 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5597 * config/nds32/nds32-intrinsic.md: Likewise.
5598 * config/nds32/nds32_intrinsic.h: Likewise.
5599 * config/nds32/nds32-md-auxiliary.c: Likewise.
5600 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5601 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5602 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5603 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5604 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5605 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5606 TYPE_DMAC in switch statement.
5607 * config/nds32/nds32.c: New checking and implementation for DSP
5608 extension instructions.
5609 * config/nds32/nds32.h: Likewise.
5610 * config/nds32/nds32.md: Likewise.
5611 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5612 * config/nds32/predicates.md: Implement new predicates for DSP
5615 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5617 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5618 Reformat alternatives and attributes so it is easier to identify
5619 which constraints/attributes go with which instruction.
5620 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5621 (mov<mode>_softfloat32, FMOVE64): Likewise.
5622 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5623 (mov<mode>_softfloat64, FMOVE64): Likewise.
5625 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5627 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5629 (Basic PowerPC Built-in Functions): The new name of the
5630 subsection previously known as "PowerPC Built-in Functions".
5631 (Basic PowerPC Built-in Functions Available on all Configurations):
5633 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5634 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5635 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5636 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5638 2018-05-11 Martin Jambor <mjambor@suse.cz>
5641 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5644 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5647 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5649 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5651 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5652 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5653 (ix86_handle_option): Handle -mwaitpkg.
5654 * config.gcc: New header.
5655 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5656 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5657 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5659 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5660 OPTION_MASK_ISA_WAITPKG.
5661 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5662 (ix86_option_override_internal): Add PTA_WAITPKG.
5663 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5664 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5665 IX86_BUILTIN_TPAUSE.
5666 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5667 __builtin_ia32_umwait and __builtin_ia32_tpause.
5668 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5669 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5670 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5671 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5672 UNSPECV_TPAUSE): New.
5673 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5674 * config/i386/i386.opt: Add -mwaitpkg.
5675 * config/i386/waitpkgintrin.h: New file.
5676 * config/i386/x86intrin.h: New header.
5677 * doc/invoke.texi: Add -mwaitpkg.
5679 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5682 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5684 (cortex-m0): Use armv6s-m isa.
5685 (cortex-m0plus): Likewise.
5686 (cortex-m1): Likewise.
5687 (cortex-m0.small-multiply): Likewise.
5688 (cortex-m0plus.small-multiply): Likewise.
5689 (cortex-m1.small-multiply): Likewise.
5691 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5692 Jakub Jelinek <jakub@redhat.com>
5694 PR tree-optimization/85692
5695 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5696 source permute as well.
5698 2018-05-11 Martin Liska <mliska@suse.cz>
5701 * doc/extend.texi: Document LLVM style format for no_sanitize
5704 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5706 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5707 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5708 (mode_supports_vsx_dform_quad): Likewise.
5709 (mode_supports_vmx_dform): Move these functions to be next to the
5710 other mode_supports functions.
5711 (mode_supports_dq_form): Likewise.
5712 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5713 mode_supports_dq_form.
5714 (reg_offset_addressing_ok_p): Likewise.
5715 (offsettable_ok_by_alignment): Likewise.
5716 (rs6000_legitimate_offset_address_p): Likewise.
5717 (legitimate_lo_sum_address_p): Likewise.
5718 (rs6000_legitimize_address): Likewise.
5719 (rs6000_legitimize_reload_address): Likewise.
5720 (rs6000_secondary_reload_inner): Likewise.
5721 (rs6000_preferred_reload_class): Likewise.
5722 (rs6000_output_move_128bit): Likewise.
5724 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5726 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5727 Generate SImode target register for null target.
5728 <case IX86_BUILTIN_XGETBV>: Ditto.
5729 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5730 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5732 2018-05-10 Carl Love <cel@us.ibm.com>
5734 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5735 dcbtt and dcbtstt if operands[2] is 0.
5737 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5740 * config/i386/sse.md (usadv64qi): New expander.
5742 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5744 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5745 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5746 -maltivec=be support.
5747 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5748 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5749 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5750 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5751 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5752 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5753 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5754 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5755 altivec_vsumsws): Adjust.
5756 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5757 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5758 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5760 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5761 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5762 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5763 (altivec_lve<VI_char>x): Delete expand.
5764 (*altivec_lve<VI_char>x_internal): Rename to...
5765 (altivec_lve<VI_char>x): ... this.
5766 (altivec_lvxl_<mode>): Delete expand.
5767 (*altivec_lvxl_<mode>_internal): Rename to ...
5768 (altivec_lvxl_<mode>): ... this.
5769 (altivec_stvxl_<mode>): Delete expand.
5770 (*altivec_stvxl_<mode>_internal): Rename to ...
5771 (altivec_stvxl_<mode>): ... this.
5772 (altivec_stve<VI_char>x): Delete expand.
5773 (*altivec_stve<VI_char>x_internal): Rename to ...
5774 (altivec_stve<VI_char>x): ... this.
5775 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5776 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5777 reduc_plus_scal_<mode>): Adjust.
5778 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5780 (rs6000_cpu_cpp_builtins): Adjust.
5781 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5782 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5783 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5784 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5785 -maltivec=be support.
5786 (rs6000_split_vec_extract_var): Adjust.
5787 (rs6000_split_v4si_init): Adjust.
5788 (swap_selector_for_mode): Delete.
5789 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5790 altivec_expand_stvex_be): Delete.
5791 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5792 -maltivec=be support.
5793 (rs6000_gimple_fold_builtin): Ditto.
5794 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5796 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5797 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5798 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5799 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5800 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5801 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5802 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5803 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5804 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5805 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5806 anonymous split): Adjust.
5807 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5808 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5810 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5812 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5813 when --with-gxx-include-dir is also specified.
5814 * configure: Regenerate.
5816 2018-05-09 Jim Wilson <jimw@sifive.com>
5819 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5820 * config/riscv/t-withmultilib: New.
5821 * config/riscv/withmultilib.h: New.
5822 * doc/install.texi: Document RISC-V --with-multilib-list support.
5824 2018-05-09 Richard Biener <rguenther@suse.de>
5826 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5828 (vect_bb_vectorization_profitable_p): Adjust. Compute
5829 actual scalar cost using the cost vector and the add_stmt_cost
5832 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5834 PR rtl-optimization/85645
5835 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5836 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5838 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5840 PR rtl-optimization/85645
5841 * shrink-wrap.c (spread_components): Return a boolean saying if
5842 anything was changed.
5843 (try_shrink_wrapping_separate): Iterate spread_components until
5844 nothing changes anymore.
5846 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5848 PR rtl-optimization/85645
5849 * regrename.c (build_def_use): Also kill the chains that include the
5850 destination of a REG_CFA_REGISTER note.
5852 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5854 PR rtl-optimization/85645
5855 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5856 insn that has a REG_CFA_REGISTER note.
5858 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5860 * cfgexpand.c (expand_clobber): New function.
5861 (expand_gimple_stmt_1): Use it.
5862 * tree-vect-stmts.c (vect_clobber_variable): New function,
5864 (vectorizable_simd_clone_call): ...here.
5865 (vectorizable_store): Emit a clobber either side of an
5866 IFN_STORE_LANES sequence.
5867 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5869 2018-05-09 Tom de Vries <tom@codesourcery.com>
5872 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5873 (define_insn "trap_if_false"): Add exit after trap.
5875 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5877 PR rtl-optimization/85638
5878 * bb-reorder.c: Include common/common-target.h.
5879 (create_forwarder_block): New function extracted from...
5880 (fix_up_crossing_landing_pad): ...here. Rename into...
5881 (dw2_fix_up_crossing_landing_pad): ...this.
5882 (sjlj_fix_up_crossing_landing_pad): New function.
5883 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5884 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5885 from both partitions and exit the loop after one iteration.
5887 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5890 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5892 (Basic PowerPC Built-in Functions): The new name of the
5893 subsection previously known as "PowerPC Built-in Functions".
5894 (Basic PowerPC Built-in Functions Available on all Configurations):
5896 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5898 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5900 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5902 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5905 2018-05-08 Jim Wilson <jimw@sifive.com>
5907 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5908 (LD_EMUL_SUFFIX): New.
5909 (LINK_SPEC): Use it.
5911 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5913 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5915 (Basic PowerPC Built-in Functions): The new name of the
5916 subsection previously known as "PowerPC Built-in Functions".
5917 (Basic PowerPC Built-in Functions Available on all Configurations):
5919 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5921 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5923 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5925 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5928 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5931 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5932 after cmpelim optimization.
5934 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5936 * config.gcc: Support "goldmont".
5937 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5938 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5940 * config/i386/i386.c (m_GOLDMONT): Define.
5941 (processor_target_table): Add "goldmont".
5942 (PTA_GOLDMONT): Define.
5943 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5944 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5945 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5946 (fold_builtin_cpu): Add "goldmont".
5947 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5948 (ix86_option_override_internal): Add "goldmont".
5949 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5950 (processor_type): Add PROCESSOR_GOLDMONT.
5951 * config/i386/i386.md: Add CPU "glm".
5952 * config/i386/glm.md: New file.
5953 * config/i386/x86-tune.def: Add m_GOLDMONT.
5954 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5956 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5959 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5961 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5962 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5963 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5966 * config/i386/i386.c (ix86_fold_builtin): Handle
5967 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5970 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5971 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5973 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5976 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5977 (check_arch): Likewise.
5978 (check_fpu): Return the result rather than printing it.
5979 (end arch): Fix operator precedence.
5980 (end cpu): Likewise.
5981 (END): Print the result from check_fpu.
5983 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5984 Alan Hayward <alan.hayward@arm.com>
5985 David Sherwood <david.sherwood@arm.com>
5987 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5988 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5989 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5990 (*fcmuo<mode>_and): New patterns.
5992 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5994 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5995 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5996 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5997 (cmp_op, sve_imm_con): New code attributes.
5998 (SVE_COND_INT_CMP, imm_con): Delete.
5999 (cmp_op): Remove above unspecs from int attribute.
6000 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6002 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6003 comparison-specific unspecs.
6004 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6005 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6006 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6007 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6008 (*vec_fcm<cmp_op><mode>): Rename to...
6009 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6010 (*vec_fcmuo<mode>): Rename to...
6011 (*fcmuo<mode>): ...this and adjust likewise.
6012 (*pred_fcm<cmp_op><mode>): New pattern.
6013 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6014 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6016 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6018 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6019 (aarch64_emit_sve_predicated_cond): New function.
6020 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6021 (aarch64_emit_unspec_cond_or): Replace with...
6022 (aarch64_emit_sve_or_conds): ...this new function. Use
6023 aarch64_emit_sve_ptrue_op for the individual comparisons and
6024 aarch64_emit_binop to OR them together.
6025 (aarch64_emit_inverted_unspec_cond): Replace with...
6026 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6027 aarch64_emit_sve_ptrue_op for the comparison and
6028 aarch64_emit_unop to invert the result.
6029 (aarch64_expand_sve_vec_cmp_float): Update after the above
6030 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6032 2018-05-07 Nathan Sidwell <nathan@acm.org>
6034 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6035 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6036 (Backwards Compatibility): Likewise.
6038 2018-05-07 Luis Machado <luis.machado@linaro.org>
6042 2018-05-07 Luis Machado <luis.machado@linaro.org>
6044 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6045 <prefetch_dynamic_strides>: New const bool field.
6046 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6047 prefetch_dynamic_strides.
6048 (exynosm1_prefetch_tune): Likewise.
6049 (thunderxt88_prefetch_tune): Likewise.
6050 (thunderx_prefetch_tune): Likewise.
6051 (thunderx2t99_prefetch_tune): Likewise.
6052 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6054 (aarch64_override_options_internal): Update to set
6055 PARAM_PREFETCH_DYNAMIC_STRIDES.
6056 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6057 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6058 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6059 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6060 prefetch-dynamic-strides setting.
6062 2018-05-07 Luis Machado <luis.machado@linaro.org>
6064 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6065 <minimum_stride>: New const int field.
6066 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6067 minimum_stride field.
6068 (exynosm1_prefetch_tune): Likewise.
6069 (thunderxt88_prefetch_tune): Likewise.
6070 (thunderx_prefetch_tune): Likewise.
6071 (thunderx2t99_prefetch_tune): Likewise.
6072 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6073 (aarch64_override_options_internal): Update to set
6074 PARAM_PREFETCH_MINIMUM_STRIDE.
6075 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6076 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6077 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6078 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6079 stride is constant and is below the minimum stride threshold.
6081 2018-05-07 Luis Machado <luis.machado@linaro.org>
6083 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6086 2018-05-07 Luis Machado <luis.machado@linaro.org>
6088 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6089 <prefetch_dynamic_strides>: New const bool field.
6090 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6091 prefetch_dynamic_strides.
6092 (exynosm1_prefetch_tune): Likewise.
6093 (thunderxt88_prefetch_tune): Likewise.
6094 (thunderx_prefetch_tune): Likewise.
6095 (thunderx2t99_prefetch_tune): Likewise.
6096 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6098 (aarch64_override_options_internal): Update to set
6099 PARAM_PREFETCH_DYNAMIC_STRIDES.
6100 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6101 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6102 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6103 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6104 prefetch-dynamic-strides setting.
6106 2018-05-07 Luis Machado <luis.machado@linaro.org>
6108 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6109 <minimum_stride>: New const int field.
6110 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6111 minimum_stride field.
6112 (exynosm1_prefetch_tune): Likewise.
6113 (thunderxt88_prefetch_tune): Likewise.
6114 (thunderx_prefetch_tune): Likewise.
6115 (thunderx2t99_prefetch_tune): Likewise.
6116 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6117 (aarch64_override_options_internal): Update to set
6118 PARAM_PREFETCH_MINIMUM_STRIDE.
6119 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6120 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6121 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6122 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6123 stride is constant and is below the minimum stride threshold.
6125 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6128 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6129 the type is addressable. Don't force op into register if it has
6132 2018-05-05 Roland McGrath <mcgrathr@google.com>
6135 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6136 any section for which we don't know a specific type it should have,
6137 regardless of name. Previously this was done only for the exact
6138 names ".init_array", ".fini_array", and ".preinit_array".
6139 (default_elf_asm_named_section): Add comment about
6140 relationship with default_section_type_flags and SECTION_NOTYPE.
6141 (get_section): Don't consider it a type conflict if one side has
6142 SECTION_NOTYPE and the other doesn't, as long as neither has the
6143 SECTION_BSS et al used in the default_section_type_flags logic.
6145 2018-05-05 Tom de Vries <tom@codesourcery.com>
6148 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6149 (workaround_barsyncs): New function.
6150 (nvptx_reorg): Use workaround_barsyncs.
6151 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6152 (define_expand "nvptx_membar_cta"): New define_expand.
6153 (define_insn "*nvptx_membar_cta"): New insn.
6155 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
6157 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
6158 To improve optimization opportunities.
6159 * builtin-types.def: The new needed builtin types for the above.
6161 2018-05-04 Richard Biener <rguenther@suse.de>
6163 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
6164 * gimple-ssa-store-merging.c
6165 (imm_store_chain_info::output_merged_store): Remove redundant create,
6166 release split_store vector contents on failure.
6167 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
6168 scalar stmt vector on cache hit.
6170 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
6172 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
6174 * config.gcc (powerpc-xilinx-eabi*): Remove.
6175 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
6177 (fusion_addis_mem_combo_load): Ditto.
6178 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
6180 (rs6000_cpu_cpp_builtins): Ditto.
6181 * config/rs6000/rs6000-linux.c
6182 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
6183 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
6184 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
6186 (rs6000_setup_reg_addr_masks): Ditto.
6187 (rs6000_init_hard_regno_mode_ok): Ditto.
6188 (rs6000_option_override_internal): Ditto.
6189 (legitimate_lo_sum_address_p): Ditto.
6190 (rs6000_legitimize_address): Ditto.
6191 (rs6000_legitimize_reload_address): Ditto.
6192 (rs6000_legitimate_address_p): Ditto.
6193 (abi_v4_pass_in_fpr): Ditto.
6194 (setup_incoming_varargs): Ditto.
6195 (rs6000_gimplify_va_arg): Ditto.
6196 (rs6000_split_multireg_move): Ditto.
6197 (rs6000_savres_strategy): Ditto.
6198 (rs6000_emit_prologue_components): Ditto.
6199 (rs6000_emit_epilogue_components): Ditto.
6200 (rs6000_emit_prologue): Ditto.
6201 (rs6000_emit_epilogue): Ditto.
6202 (rs6000_elf_file_end): Ditto.
6203 (rs6000_function_value): Ditto.
6204 (rs6000_libcall_value): Ditto.
6205 * config/rs6000/rs6000.h: Ditto.
6206 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
6207 (TARGET_MINMAX): ... this. New.
6208 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
6209 * config/rs6000/rs6000.md: Remove Xilinx FP support.
6210 (*movsi_internal1_single): Delete.
6211 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
6212 mfpu=, mxilinx-fpu): Delete.
6213 * config/rs6000/singlefp.h: Delete.
6214 * config/rs6000/sysv4.h: Remove Xilinx FP support.
6215 * config/rs6000/t-rs6000: Ditto.
6216 * config/rs6000/t-xilinx: Delete.
6217 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
6218 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
6219 (VStype_simple): Delete.
6220 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
6221 * config/rs6000/xfpu.h: Delete.
6222 * config/rs6000/xfpu.md: Delete.
6223 * config/rs6000/xilinx.h: Delete.
6224 * config/rs6000/xilinx.opt: Delete.
6225 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
6226 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
6228 2018-05-04 Tom de Vries <tom@codesourcery.com>
6231 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
6234 2018-05-04 Richard Biener <rguenther@suse.de>
6237 * tree-complex.c (update_complex_assignment): We are always in SSA form.
6238 (expand_complex_div_wide): Likewise.
6239 (expand_complex_operations_1): Likewise.
6240 (expand_complex_libcall): Preserve EH info of the original stmt.
6241 (tree_lower_complex): Handle removed blocks.
6242 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
6243 on complex multiplication and division libcall builtins.
6245 2018-05-04 Richard Biener <rguenther@suse.de>
6248 * fold-const.c (negate_expr_p): Restrict negation of operand
6249 zero of a division to when we know that can happen without
6251 (fold_negate_expr_1): Likewise.
6253 2018-05-04 Jakub Jelinek <jakub@redhat.com>
6256 * real.h (real_nextafter): Declare.
6257 * real.c (real_nextafter): New function.
6258 * fold-const-call.c (fold_const_nextafter): New function.
6259 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
6260 CASE_CFN_NEXTTOWARD.
6261 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
6262 even when arg1_mode is different from arg0_mode.
6264 2018-05-03 Nathan Sidwell <nathan@acm.org>
6266 * doc/extend.texi (Deprecated Features): Remove
6268 (Backwards Compatibility): Likewise.
6269 * doc/invoke.texi (C++ Language Options): Likewise.
6270 (C++ Dialect Options): Likewise.
6272 2018-05-03 Jakub Jelinek <jakub@redhat.com>
6275 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
6276 _mm512_mask_mullox_epi64): New intrinsics.
6278 2018-05-03 Tom de Vries <tom@codesourcery.com>
6281 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6282 dump files): Add offload-tree.
6284 2018-05-03 Richard Biener <rguenther@suse.de>
6286 PR tree-optimization/85615
6287 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
6288 to loops not nested in BBs loop father to avoid creating multi-entry
6291 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6293 PR tree-optimization/70291
6294 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
6295 arguments. Change return type to tree. Emit libcall as a new
6296 statement rather than replacing existing one when inplace_p is true.
6297 (expand_complex_multiplication_components): New function.
6298 (expand_complex_multiplication): Expand floating-point complex
6299 multiplication using the above.
6300 (expand_complex_division): Rename inner_type parameter to type.
6301 Update expand_complex_libcall call-site.
6302 (expand_complex_operations_1): Update expand_complex_multiplication
6303 and expand_complex_division call-sites.
6305 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6308 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6309 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6310 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
6311 the highest significant bit of the shift count mask is clear. In
6312 check whether and[sq]i3 is needed verify that all significant bits
6313 of the shift count other than the highest are set.
6315 2018-05-02 Tom de Vries <tom@codesourcery.com>
6318 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
6319 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
6320 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
6321 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
6322 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
6323 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
6324 __builtin_goacc_parlevel_size.
6326 2018-05-02 Richard Biener <rguenther@suse.de>
6328 PR tree-optimization/85597
6329 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
6330 do not use split vect_get_vec_defs call but call vect_get_slp_defs
6333 2018-05-02 Tom de Vries <tom@codesourcery.com>
6336 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6337 dump files): Add ltrans-tree.
6339 2018-05-02 Tom de Vries <tom@codesourcery.com>
6342 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
6343 dump files): Add wpa-ipa.
6345 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
6347 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
6348 powerpc*-*-linux*paired* target.
6349 * config/rs6000/750cl.h: Delete.
6350 * config/rs6000/paired.h: Delete.
6351 * config/rs6000/paired.md: Delete.
6352 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
6354 * config/rs6000/rs6000-builtin.def: Remove paired float support.
6355 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
6356 comment. Remove paired float support.
6357 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
6358 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6360 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
6361 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
6363 * config/rs6000/rs6000.c: Remove paired float support.
6364 (paired_expand_vector_init, paired_expand_vector_move,
6365 paired_emit_vector_compare, paired_emit_vector_cond_expr,
6366 (paired_expand_lv_builtin, paired_expand_stv_builtin,
6367 paired_expand_builtin, paired_expand_predicate_builtin,
6368 paired_init_builtins): Delete.
6369 * config/rs6000/rs6000.h: Remove paired float support.
6370 * config/rs6000/rs6000.md: Remove paired float support.
6371 (move_from_CR_ov_bit): Delete.
6372 * config/rs6000/rs6000.opt (mpaired): Delete.
6373 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
6374 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
6376 2018-05-02 Richard Biener <rguenther@suse.de>
6379 * gimplify.c (gimplify_save_expr): When in SSA form allow
6380 SAVE_EXPRs to compute to SSA vars.
6382 2018-05-02 Jakub Jelinek <jakub@redhat.com>
6385 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
6386 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
6387 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
6388 clobber operands[2], instead use a new pseudo. Formatting fixes.
6390 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
6392 PR tree-optimization/85586
6393 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
6394 exit early for statements in the same group if the accesses are
6397 2018-05-02 Tom de Vries <tom@codesourcery.com>
6400 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
6403 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
6405 PR tree-optimization/85143
6406 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
6408 2018-05-01 Tom de Vries <tom@codesourcery.com>
6411 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
6412 not found" error message.
6414 2018-05-01 Tom de Vries <tom@codesourcery.com>
6417 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
6418 * vec.c (test_ordered_remove_if): New function.
6419 (vec_c_tests): Call test_ordered_remove_if.
6420 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
6421 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
6422 * tree-vect-patterns.c (vect_pattern_recog_1): Use
6423 VEC_ORDERED_REMOVE_IF.
6425 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6427 PR tree-optimization/82665
6428 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
6429 pointer subtraction where arguments come from a memchr call.
6431 2018-05-01 Jakub Jelinek <jakub@redhat.com>
6433 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
6434 --push-state --as-needed and --pop-state instead of --as-needed and
6435 --no-as-needed if ld supports it.
6436 * configure: Regenerated.
6439 * doc/install.texi2html: Replace _002d with - and _002a with * in
6440 generated html files using sed.
6442 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6445 * gcc-rich-location.c (blank_line_before_p): New function.
6446 (use_new_line): New function.
6447 (gcc_rich_location::add_fixit_insert_formatted): New function.
6448 * gcc-rich-location.h
6449 (gcc_rich_location::add_fixit_insert_formatted): New function.
6451 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6453 * selftest.c (assert_streq): Rename "expected" and "actual" to
6454 "val1" and "val2". Extend NULL-handling to cover both inputs
6455 symmetrically, while still requiring both to be non-NULL for a pass.
6456 * selftest.h (assert_streq): Rename "expected" and "actual" to
6458 (ASSERT_EQ): Likewise.
6459 (ASSERT_EQ_AT): Likewise.
6460 (ASSERT_KNOWN_EQ): Likewise.
6461 (ASSERT_KNOWN_EQ_AT): Likewise.
6462 (ASSERT_NE): Likewise.
6463 (ASSERT_MAYBE_NE): Likewise.
6464 (ASSERT_MAYBE_NE_AT): Likewise.
6465 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
6466 the assertion to pass.
6467 (ASSERT_STREQ_AT): Likewise.
6469 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6471 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
6472 interaction with -pie.
6474 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6476 * selftest.h: Fix alphabetization of per-source-file selftest
6479 2018-04-30 Jason Merrill <jason@redhat.com>
6481 PR c++/61982 - dead stores to destroyed objects.
6482 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
6485 2018-04-30 Jason Merrill <jason@redhat.com>
6487 * tree.c (build_clobber): New.
6488 * tree.h: Declare it.
6489 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
6491 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6493 * diagnostic-show-locus.c (layout::layout): Update for
6494 location_get_source_line returning a char_span.
6495 (struct char_span): Move to input.h.
6496 (struct correction): Update for fields in char_span becoming
6498 (struct source_line): Update for location_get_source_line
6499 returning a char_span.
6500 (layout::print_line): Likewise.
6501 * edit-context.c (edited_file::print_content): Likewise.
6502 (edited_file::print_diff_hunk): Likewise.
6503 (edited_file::print_run_of_changed_lines): Likewise.
6504 (edited_file::get_num_lines): Likewise.
6505 (edited_line::edited_line): Likewise.
6506 * final.c (asm_show_source): Likewise.
6507 * input.c (location_get_source_line): Convert return type
6508 from const char * to char_span, losing the final "line_len"
6510 (dump_location_info): Update for the above.
6511 (get_substring_ranges_for_loc): Likewise. Use a char_span
6512 when handling the literal within the line.
6513 (test_reading_source_line): Update for location_get_source_line
6514 returning a char_span.
6515 * input.h (class char_span): Move here from
6516 diagnostic-show-locus.c, converting from a struct to a class.
6517 Make data members private.
6518 (char_span::operator bool): New.
6519 (char_span::length): New.
6520 (char_span::get_buffer): New.
6521 (char_span::operator[]): New.
6522 (char_span::subspan): Make const.
6523 (char_span::xstrdup): New.
6524 (location_get_source_line): Convert return type from const char *
6525 to char_span, losing the final "line_size" param.
6527 2018-04-30 Jan Hubicka <jh@suse.cz>
6529 * lto-wrapper.c (ltrans_priorities): New static var.
6530 (cmp_priority): New.
6531 (run_gcc): Read priorities and if doing parallel build order
6532 the Makefile by them.
6534 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6536 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6538 2018-04-30 Richard Biener <rguenther@suse.de>
6540 * tree-cfg.c (verify_address): Remove base argument, add
6541 flag whether to check TREE_ADDRESSABLE and do that.
6542 (verify_expr): Remove.
6543 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6544 (verify_gimple_assign_single): Likewise.
6545 (verify_gimple_switch): Likewise.
6546 (verify_expr_location_1): Dereference tp once. Add (disabled)
6547 piece from verify_expr.
6548 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6550 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6552 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6554 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6556 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6557 (small_data_pattern): Likewise.
6558 (arc_rewrite_small_data): Likewise.
6559 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6560 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6561 (get_symbol_alignment): New function.
6562 (legitimate_small_data_address_p): Likewise.
6563 (legitimate_scaled_address): Update, call
6564 legitimate_small_data_address_p.
6565 (output_sdata): New static variable.
6566 (arc_print_operand): Update how we handle small data operands.
6567 (arc_print_operand_address): Likewise.
6568 (arc_legitimate_address_p): Update, use
6569 legitimate_small_data_address_p.
6570 (arc_rewrite_small_data_p): Remove.
6571 (arc_rewrite_small_data_1): Likewise.
6572 (arc_rewrite_small_data): Likewise.
6573 (small_data_pattern): Likewise.
6574 (compact_sda_memory_operand): Update to use
6575 legitimate_small_data_address_p and get_symbol_alignment.
6576 (prepare_move_operands): Don't rewite sdata pattern.
6577 (prepare_extend_operands): Remove.
6578 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6580 (zero_extendqisi2): Likewise.
6581 (zero_extendhisi2): Likewise.
6582 (extendqihi2): Likewise.
6583 (extendqisi2): Likewise.
6584 (extendhisi2): Likewise.
6588 * config/arc/constraints.md (Usd): Change it to memory constraint.
6590 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6592 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6593 as source of std instructions.
6594 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6595 allow 6-bit constants as source for store instructions.
6596 (movdi_insn): Update instruction pattern to allow 6-bit constants
6597 as source for store instructions.
6599 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6601 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6603 2018-04-30 Nathan Sidwell <nathan@acm.org>
6604 Sandra Loosemore <sandra@codesourcery.com>
6606 * dumpfile.c (dump_open): Allow '-' for stdout.
6607 * doc/invoke.texi (Developer Options): Document dump filename
6608 determination early. Document stdin/stdout selection.
6610 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6612 Microblaze Target: PIC data text relative
6614 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6615 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6617 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6618 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6619 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6620 New addressing mode for data-text relative position indepenedent code.
6621 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6622 'ADDRESS_SYMBOLIC_TXT_REL'.
6623 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6624 (microblaze_legitimate_pic_operand): Exclude function calls from
6625 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6626 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6628 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6629 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6630 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6631 for 'address + offset'.
6632 (microblaze_expand_prologue): Add new function prologue call for
6634 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6635 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6636 table in case of TARGET_PIC_DATA_TEXT_REL.
6637 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6638 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6639 Add new macros 'UNSPEC_TEXT',
6640 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6641 + exclude function calls from 'UNSPEC_PLT' in case of data text
6643 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6644 new target hook for generating address diff vector tables in case of
6646 * doc/tm.texi : Regenerate.
6647 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6648 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6649 of addr diff vector generation.
6650 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6651 target hook definition.
6652 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6653 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6654 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6655 option for data text relative.
6657 2018-04-30 Richard Biener <rguenther@suse.de>
6659 * tree-chrec.h (evolution_function_is_constant_p): Remove
6661 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6663 2018-04-30 Richard Biener <rguenther@suse.de>
6666 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6668 2018-04-30 Richard Biener <rguenther@suse.de>
6670 PR tree-optimization/28364
6671 PR tree-optimization/85275
6672 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6673 copying first exit test.
6675 2018-04-28 Mark Wielaard <mark@klomp.org>
6677 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6679 (dwarf_AT): Handle DW_AT_addr_base.
6680 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6682 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6685 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6686 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6687 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6688 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6690 2018-04-28 Richard Biener <rguenther@suse.de>
6692 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6693 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6694 to reflect use. Only add interesting stmts.
6696 2018-04-27 Martin Jambor <mjambor@suse.cz>
6699 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6700 the jump function allows for passing through aggregate values.
6702 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6704 * input.h (in_system_header_at): Convert from macro to inline
6706 (from_macro_expansion_at): Likewise.
6707 (from_macro_definition_at): Likewise.
6709 2018-04-27 Jeff Law <law@redhat.com>
6711 * config.gcc: Mark tile* targets as deprecated/obsolete.
6713 2018-04-27 Richard Biener <rguenther@suse.de>
6715 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6718 2018-04-27 Richard Biener <rguenther@suse.de>
6720 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6722 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6724 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6725 with Yd constraint. Set "preferred_for_speed" attribute from
6726 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6728 (*movdi_internal): Ditto.
6729 (movti_interunit splitters): Remove
6730 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6731 (movdi_interunit splitters): Ditto.
6732 * config/i386/constraints.md (Ye): Remove.
6733 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6735 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6738 * config/aarch64/constraints.md (Usg): Limit to 31.
6741 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6743 PR tree-optimization/85529
6744 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6745 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6746 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6747 zero extension or masking of the MSB bit.
6748 (optimize_range_tests): Add FIRST_BB argument, pass it through
6749 to optimize_range_tests_var_bound.
6750 (maybe_optimize_range_tests, reassociate_bb): Adjust
6751 optimize_range_tests callers.
6753 2018-04-26 Richard Biener <rguenther@suse.de>
6754 Jakub Jelinek <jakub@redhat.com>
6756 * cgraph.h (symbol_table): Just declare debug method here.
6757 * symtab.c (symbol_table::debug): Define.
6759 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6761 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6763 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6765 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6766 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6767 (*movdi_internal): Substitute Yi and Yj constraint with x
6768 and Ym and Yn constraint with y constraint. Update "isa"
6769 attribute and set "preferred_for_speed" attribute from
6770 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6771 (*movsi_internal): Ditto.
6772 (*movdf_internal): Ditto.
6773 (*movsf_internal): Ditto.
6774 (*zero_extendsidi2): Ditto.
6775 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6776 (sse2_loadld): Ditto.
6777 (*vec_extract<ssevecmodelower>_0): Ditto.
6778 (*vec_extractv4si_0_zext_sse4): Ditto.
6779 (vec_concatv2di): Ditto.
6780 (*vec_dup<mode>): Ditto.
6781 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6782 * config/i386/constraints.md (Yi): Remove.
6787 2018-04-26 Nathan Sidwell <nathan@acm.org>
6789 * dumpfile.c (dump_open): New.
6790 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6791 (dump_finish): Detect stdio/stderr by value not name.
6793 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6795 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6797 2018-04-26 Tom de Vries <tom@codesourcery.com>
6800 * config/nvptx/nvptx.c (verify_neutering_jumps)
6801 (verify_neutering_labels): New function
6802 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6804 2018-04-26 Tom de Vries <tom@codesourcery.com>
6807 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6808 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6811 2018-04-26 Richard Biener <rguenther@suse.de>
6812 Tom de Vries <tom@codesourcery.com>
6815 * lto-streamer-out.c (output_function): Fixup loops if required to match
6816 discovery done in the reader.
6818 2018-04-26 Richard Biener <rguenther@suse.de>
6820 PR tree-optimization/85116
6821 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6822 have a loop exit from the single latch predecessor. Remove
6823 case of header with just condition.
6824 (ch_base::copy_headers): Exclude infinite loops from any
6826 (pass_ch::execute): Record exits.
6828 2018-04-26 Richard Biener <rguenther@suse.de>
6830 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6831 prologue cost vector and pass it to vect_get_load_cost.
6832 (vect_get_peeling_costs_all_drs): Likewise.
6833 (vect_peeling_hash_get_lowest_cost): Likewise.
6834 (vect_enhance_data_refs_alignment): Likewise.
6836 2018-04-26 Richard Biener <rguenther@suse.de>
6839 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6840 checking of integer<->pointer conversions.
6841 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6842 sign-/zero-extending pointer types.
6843 (expand_omp_for_static_chunk): Likewise.
6845 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6846 Jean Lee <xiaoyur347@gmail.com>
6848 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6849 (TARGET_ASAN_SHADOW_OFFSET): Define.
6850 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6851 true for -fsanitize=address.
6853 2018-04-25 Mark Wielaard <mark@klomp.org>
6855 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6858 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6860 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6861 than "alu", remove explicit "memory" and "imm_disp" attributes.
6862 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6865 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6866 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6867 gen_lowpart_no_emit.
6869 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6872 * config/i386/i386.c (ix86_expand_builtin): Change memory
6873 operand to XI, extend p0 to Pmode.
6874 * config/i386/i386.md: Change unspec volatile and operand
6875 1 mode to XI, change operand 0 mode to P.
6877 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6879 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6880 GET_MODE_MASK before any checking.
6881 (nds32_can_use_bset_p): Likewise.
6882 (nds32_can_use_btgl_p): Likewise.
6884 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6886 * config/nds32/nds32-doubleword.md: New define_split pattern for
6887 illegal register number.
6889 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6891 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6893 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6895 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6897 2018-04-25 Richard Biener <rguenther@suse.de>
6899 * lto-streamer.h (LTO_major_version): Bump to 8.
6901 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6903 * BASE-VER: Set to 9.0.0.
6905 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6907 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6908 in __abskf2 and __powikf2.
6910 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6913 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6914 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6915 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6916 Use the above on operand 2. Reindent.
6917 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6919 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6922 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6924 * config/i386/i386.opt (mcet): Removed.
6925 * doc/install.texi: Remove -mcet documentation.
6926 * doc/invoke.texi: Likewise.
6928 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6931 * doc/install.texi: Remove -mcet from bootstrap-cet.
6933 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6936 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6937 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6941 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6942 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6943 containing a CONST_VECTOR.
6945 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6947 * doc/install.texi: Update newlib dependency for nvptx.
6949 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6952 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6953 instead of INTVAL when shifting x left.
6955 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6957 PR tree-optimization/85478
6958 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6959 vect_grouped_store_supported for single element vectors.
6961 2018-04-24 Richard Biener <rguenther@suse.de>
6964 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6965 load cost increase to the case of non-constant step.
6967 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6970 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6971 destination if any_malformed_asm.
6973 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6976 * expr.c (store_field): In the bitfield case, if the value comes from
6977 a function call and is returned in registers by means of a PARALLEL,
6978 do not change the mode of the temporary unless BLKmode and VOIDmode.
6980 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6982 PR rtl-optimization/85423
6983 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6984 dependencies to debug insns when the previous insn is non-debug.
6986 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6988 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6989 enums into a single definition.
6990 (fls): Fix predicates and printing.
6993 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6995 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6996 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6997 and short u6 immediate.
6998 (check_if_valid_sleep_operand): Remove.
6999 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7001 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7003 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7004 flag_always_save_lp condition.
7005 * config/nds32/nds32.opt (malways-save-lp): New option.
7007 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7009 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7010 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7011 * config/nds32/nds32.h
7012 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7013 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7015 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7017 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7018 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7020 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7021 Chung-Ju Wu <jasonwucj@gmail.com>
7023 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7025 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7026 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7028 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7030 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7032 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7034 * config/nds32/nds32-protos.h (nds32_data_alignment,
7035 nds32_local_alignment): Declare.
7036 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7037 nds32_local_alignment): New functions.
7038 (TARGET_CONSTANT_ALIGNMENT): Define.
7039 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7041 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7043 * config/nds32/nds32.c
7044 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7045 (TARGET_MODES_TIEABLE_P): Likewise.
7047 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7049 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7052 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7053 Chung-Ju Wu <jasonwucj@gmail.com>
7055 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7056 for unaligned access.
7057 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7059 * config/nds32/nds32-intrinsic.md: Likewise.
7060 * config/nds32/nds32_intrinsic.h: Likewise.
7061 * config/nds32/nds32.h (nds32_builtins): Likewise.
7062 * config/nds32/nds32.opt (munaligned-access): New option.
7063 * config/nds32/nds32.c (nds32_asm_file_start): Display
7064 flag_unaligned_access status.
7066 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7068 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7069 -mno-relax is present.
7070 * config/riscv/linux.h (LINK_SPEC): Ditto.
7072 2018-04-20 Martin Sebor <msebor@redhat.com>
7075 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7077 (gimple_fold_builtin_stxcpy_chk): Same.
7078 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7080 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7083 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7084 __powikf2 when long double is IEEE 128-bit.
7086 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7088 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7089 step to make sure stack always aligned.
7091 2018-04-20 Carl Love <cel@us.ibm.com>
7094 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7095 size check for arg0.
7097 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7098 Tom de Vries <tom@codesourcery.com>
7101 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7102 Emit insns for calls too.
7103 (nvptx_find_par): Always look for worker-level predecessor insn.
7104 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7106 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7107 (nvptx_process_pars): Propagate frames for calls.
7109 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7112 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7114 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7115 (ix86_handle_option): Don't handle OPT_mibt.
7116 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7118 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7120 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7121 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7122 (ix86_target_macros): Define __CET__ with flag_cf_protection
7123 for -fcf-protection.
7124 * config/i386/i386.c (isa2_opts): Remove -mibt.
7125 * config/i386/i386.h (TARGET_IBT): Removed.
7126 (TARGET_IBT_P): Likewise.
7127 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7128 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7129 * config/i386/i386.opt (mcet): Update help message.
7132 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7133 -mcet as an alias for -mshstk.
7135 2018-04-20 Richard Biener <rguenther@suse.de>
7138 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7139 complexity by forcing a single use of the multiply operand.
7141 2018-04-20 Martin Jambor <mjambor@suse.cz>
7144 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7145 recursion dependency to only apply to non-clones.
7147 2018-04-20 Martin Jambor <mjambor@suse.cz>
7150 * ipa-cp.c (create_specialized_node): Check that clones of
7151 self-recursive edges exist during IPA-CP.
7153 2018-04-19 Toon Moene <toon@moene.org>
7155 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
7158 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7160 PR tree-optimization/85467
7161 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
7162 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
7163 VECTOR_CST element to type.
7165 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7168 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
7169 * config/i386/i386.md (builtin_setjmp_setup): Removed.
7170 (builtin_longjmp): Likewise.
7171 (save_stack_nonlocal): New pattern.
7172 (restore_stack_nonlocal): Likewise.
7174 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7177 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7178 Replace ASM_OUTPUT_LABEL with fprintf.
7180 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
7183 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
7184 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
7185 * config/i386/i386-c.c (ix86_target_macros_internal): Also
7186 define __IBT__ and __SHSTK__ for -fcf-protection.
7187 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
7189 (ix86_trampoline_init): Likewise.
7190 (x86_output_mi_thunk): Likewise.
7191 (ix86_notrack_prefixed_insn_p): Likewise.
7192 (ix86_option_override_internal): Don't disallow -fcf-protection.
7193 * config/i386/i386.md (rdssp<mode>): Also enable for
7195 (incssp<mode>): Likewise.
7196 (nop_endbr): Likewise.
7197 * config/i386/i386.opt (mcet): Change help message to built-in
7201 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
7202 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
7203 enable CET built-in functions.
7205 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
7207 * common/config/i386/i386-common.c
7208 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
7209 OPTION_MASK_ISA_MOVDIRI_UNSET,
7210 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
7211 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
7212 * config.gcc (movdirintrin.h): New header.
7213 * config/i386/cpuid.h (bit_MOVDIRI,
7214 bit_MOVDIR64B): New bits.
7215 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
7217 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
7218 (VOID, PVOID, PCVOID)): New function types.
7219 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
7220 __builtin_ia32_directstoreu_u64,
7221 __builtin_ia32_movdir64b): New builtins.
7222 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
7223 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
7225 (ix86_valid_target_attribute_inner_p): Ditto.
7226 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
7227 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
7228 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
7229 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
7230 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
7231 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
7232 (movdiri<mode>, movdir64b_<mode>): New.
7233 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
7234 * config/i386/immintrin.h: Include movdirintrin.h.
7235 * config/i386/movdirintrin.h: New file.
7236 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
7238 2018-04-19 Richard Biener <rguenther@suse.de>
7241 * cfg.c (clear_bb_flags): When loop state says we have
7242 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
7244 2018-04-19 Richard Biener <rguenther@suse.de>
7246 PR tree-optimization/84737
7247 * tree-vect-data-refs.c (vect_copy_ref_info): New function
7248 copying restrict info.
7249 (vect_setup_realignment): Use it.
7250 * tree-vectorizer.h (vect_copy_ref_info): Declare.
7251 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
7252 the first DR to all generated stores.
7253 (vectorizable_load): Likewise for loads.
7255 2018-04-19 Jakub Jelinek <jakub@redhat.com>
7257 PR tree-optimization/85446
7258 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
7259 the integral and pointer types to have the same precision.
7261 * doc/install.texi: Document --disable-cet being the default and
7264 2018-04-18 Martin Liska <mliska@suse.cz>
7266 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
7269 2018-04-18 Martin Liska <mliska@suse.cz>
7272 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7275 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7276 arguments if they are comparable.
7278 2018-04-18 Martin Liska <mliska@suse.cz>
7281 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7284 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7287 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
7290 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
7291 ENDBR after calling __morestack.
7293 2018-04-18 David Malcolm <dmalcolm@redhat.com>
7296 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
7297 by using gcc_base_ver to generate a gcc_driver_version, and use
7298 it when generating GCC_DRIVER_NAME.
7299 * configure: Regenerate.
7301 2018-04-18 Jakub Jelinek <jakub@redhat.com>
7304 * config.gcc: Obsolete powerpc*-*-*spe*.
7306 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7309 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
7310 (stabstr_D): Change type of unum from unsigned int to
7311 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
7314 2018-04-17 Jim Wilson <jimw@sifive.com>
7317 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
7318 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
7319 Set arg_pointer_offset after using pretend_args_size.
7321 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7323 PR rtl-optimization/85431
7324 * dse.c (record_store): Ignore zero width stores.
7327 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
7328 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
7329 __builtin_stack_restore rather than after it.
7330 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
7331 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
7332 argument instead of virtual_dynamic_stack_rtx.
7334 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
7336 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
7338 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7339 Add note to error message to explain internal mapping of overloaded
7340 built-in function name to non-overloaded built-in function name.
7341 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
7344 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7347 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
7348 where the inputs overlap with the output.
7350 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7352 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
7353 (=v, v) alternative and explicit "memory" attribute.
7354 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
7355 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7357 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
7358 "sselog1" type instead of "sselog".
7359 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
7360 "sselog". Remove explicit "memory" attribute.
7361 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
7362 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7364 (vec_extract_hi_v32hi): Merge all alternatives into one, use
7365 "sselog1" type instead of "sselog". Remove explicit "memory"
7367 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
7368 use "sselog1" type instead of "sselog". Remove explicit "memory"
7370 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
7371 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
7373 (vec_extract_hi_v64qi): Merge all alternatives into one, use
7374 "sselog1" type instead of "sselog". Remove explicit "memory"
7376 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
7377 use "sselog1" type instead of "sselog". Remove explicit "memory"
7381 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
7384 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
7387 2018-04-17 Martin Jambor <mjambor@suse.cz>
7390 * ipa-cp.c (create_specialized_node): Call
7391 expand_all_artificial_thunks if necessary.
7393 2018-04-17 Martin Liska <mliska@suse.cz>
7396 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
7397 in message, remote space in between '_G' and '('.
7399 2018-04-17 Jakub Jelinek <jakub@redhat.com>
7402 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
7403 avx512f_vmcmp<mode>3<round_saeonly_name>,
7404 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
7405 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
7406 avx512f_rndscale<mode><round_saeonly_name>,
7407 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
7408 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7409 Use %<iptr>2 instead of %2 for -masm=intel.
7410 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
7411 avx512f_vcvttss2usi<round_saeonly_name>,
7412 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
7414 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
7415 avx512f_vcvttsd2usi<round_saeonly_name>,
7416 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
7417 Use %q1 instead of %1 for -masm=intel.
7418 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
7419 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
7420 of %3 for -masm=intel.
7421 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
7423 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
7425 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
7427 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
7428 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
7430 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
7432 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
7433 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
7434 %g1 and one with %0 and %1.
7435 (avx512er_vmrcp28<mode><round_saeonly_name>,
7436 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
7438 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
7439 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
7440 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
7441 of %0 and %{%4%} for -masm=intel.
7442 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
7443 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
7444 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
7445 order of %0 and %{%5%}%{z%} for -masm=intel.
7447 2018-04-17 Jan Hubicka <jh@suse.cz>
7450 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
7452 2018-04-17 Martin Liska <mliska@suse.cz>
7455 * multiple_target.c (create_dispatcher_calls): Set apostrophes
7456 for target_clone error message. Make default implementation
7457 clone to be a local declaration.
7458 (separate_attrs): Add new argument and check for an empty
7460 (expand_target_clones): Handle it.
7461 (ipa_target_clone): Make redirection just for target_clones
7464 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
7465 Tom de Vries <tom@codesourcery.com>
7468 * omp-expand.c (expand_oacc_for): Add dummy false branch for
7469 tiled basic blocks without omp continue statements.
7471 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
7474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
7475 vec_extract expression as having side effects to make sure it gets
7478 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
7481 * config/i386/i386.c (get_builtin_code_for_version): Check
7484 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
7487 * gcc/config.gcc: Support "skylake".
7488 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
7490 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
7491 (processor_target_table): Add "skylake".
7492 (ix86_option_override_internal): Add "skylake".
7493 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
7494 PROCESSOR_CANNONLAKE.
7495 (get_builtin_code_for_version): Fix priority for
7496 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
7497 PROCESSOR_SKYLAKE-AVX512.
7498 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
7499 (processor_type): Add PROCESSOR_SKYLAKE.
7501 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
7502 Jason Merrill <jason@redhat.com>
7505 * convert.c (convert_to_integer_1): Use direct recursion for
7506 enumeral types and types with a precision less than the number
7507 of bits in their mode.
7509 2018-04-16 Julia Koval <julia.koval@intel.com>
7512 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
7513 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
7515 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
7518 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
7519 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
7520 and -mno-direct-move.
7522 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
7525 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7526 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7527 Ensure negative shifts result in {0}.
7529 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7531 PR rtl-optimization/79916
7532 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7533 regs (if any) to define how to gnerate SD moves when LRA is in
7536 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7538 PR rtl-optimization/85393
7539 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7540 * except.c (expand_dw2_landing_pad_for_region): Make static.
7541 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7542 a label and unconditional jump to old_bb, rather than
7543 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7546 PR rtl-optimization/85376
7547 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7548 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7549 instead of a specific value.
7551 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7552 Bin Cheng <bin.cheng@arm.com>
7554 PR tree-optimization/82965
7555 PR tree-optimization/83991
7556 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7557 by_profile_only parameter.
7558 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7559 information if the loop was predicted to iterate too many times.
7560 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7562 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7565 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7568 2018-04-13 Martin Liska <mliska@suse.cz>
7569 Jakub Jelinek <jakub@redhat.com>
7572 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7573 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7574 * builtins.c (expand_builtin_memory_copy_args): Use
7575 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7576 handle dest_addr == pc_rtx.
7578 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7581 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7582 asked to not generate direct moves.
7583 (fix_trunc<mode>si2_stfiwx): Similar.
7584 (fix_trunc<mode>si2_internal): Similar.
7586 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7589 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7590 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7591 lookup if dest in some wider mode is known to be const0_rtx and
7592 if so, record permanent equivalence for it to be ZERO_EXTEND of
7593 the narrower mode destination.
7595 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7597 * lto-streamer-out.c (output_function): Revert 259346.
7598 * omp-expand.c (expand_oacc_for): Likewise.
7600 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7602 PR rtl-optimization/85354
7603 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7604 * sel-sched.c (sel_global_init): ... here.
7606 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7609 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7610 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7611 mode for PE-COFF targets.
7612 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7613 (i386_pe_asm_lto_end): Likewise.
7614 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7615 (TARGET_ASM_LTO_END): Likewise.
7616 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7617 (i386_pe_asm_lto_start): New function.
7618 (i386_pe_asm_lto_end): Likewise.
7620 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7621 Richard Biener <rguenther@suse.de>
7624 * lto-streamer-out.c (output_function): Fix CFG loop state before
7626 * omp-expand.c (expand_oacc_for): Handle calls to internal
7627 functions like regular functions.
7629 2018-04-12 Richard Biener <rguenther@suse.de>
7632 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7633 for the early LTO debug to properly generate references to it
7634 during DIE emission. Do not re-use that for the skeleton for
7636 (dwarf2out_early_finish): Likewise.
7638 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7641 * config/i386/sse.md
7642 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7643 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7644 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7645 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7646 and output is a reg, avoid creating invalid lowpart subreg, but
7647 instead split into a 512-bit move. Don't split if not AVX512VL,
7648 input is xmm16+ reg and output is a mem.
7649 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7650 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7651 xmm16+ reg and output is a mem.
7653 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7655 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7656 also for flag_dwarf2_cfi_asm.
7658 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7660 PR rtl-optimization/85342
7661 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7662 a bool scalar var inside of the loop instead. Don't try to update
7663 recog_data.operand after failed apply_change_group.
7665 2018-04-12 Tom de Vries <tom@codesourcery.com>
7668 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7669 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7670 array with flexible array member as array without given dimension.
7671 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7672 argument for undefined param to true.
7674 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7677 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7678 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7679 from PowerPC section.
7680 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7681 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7682 help text that is too long.
7683 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7684 help text that is too long.
7685 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7686 help text that is too long.
7688 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7690 * config/alpha/alpha.md (stack_probe_internal): Rename
7691 from "probe_stack". Update all callers.
7693 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7695 PR rtl-optimization/84566
7696 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7697 sched_macro_fuse_insns.
7699 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7702 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7703 (compute_block_dependences): ... from here.
7705 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7707 PR tree-optimization/85331
7708 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7709 from int to HOST_WIDE_INT.
7711 2018-04-11 Martin Jambor <mjambor@suse.cz>
7714 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7715 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7716 not the same as the source val.
7717 (cgraph_edge_brings_value_p): New parameter.
7718 (gather_edges_for_value): Pass destination value to
7719 cgraph_edge_brings_value_p.
7720 (perhaps_add_new_callers): Likewise.
7721 (get_info_about_necessary_edges): Likewise and exclude values brought
7722 only by self-recursive edges.
7723 (create_specialized_node): Redirect only clones of self-calling edges.
7724 (+self_recursive_pass_through_p): New function.
7725 (find_more_scalar_values_for_callers_subset): Use it.
7726 (find_aggregate_values_for_callers_subset): Likewise.
7727 (known_aggs_to_agg_replacement_list): Removed.
7728 (decide_whether_version_node): Re-calculate known constants for all
7729 remaining context clones.
7731 2018-04-11 Richard Biener <rguenther@suse.de>
7734 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7735 from early DWARF output.
7736 (dwarf2out_early_finish): Output line info unconditionally into
7737 early DWARF and add reference to it.
7739 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7742 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7743 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7744 other than V2DFmode using iptr mode attribute.
7745 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7747 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7749 PR rtl-optimization/84659
7750 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7752 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7755 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7757 (output_loc_list): Pass address of a dummy size variable even in the
7758 locview handling loop.
7759 (index_location_lists): Add comment on why skip_loc_list_entry can't
7762 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7765 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7768 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7771 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7772 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7773 and -mstring-compare-inline-limit.
7775 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7778 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7779 for stack clash protection in a register whenever we need it to be in
7782 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7784 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7785 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7787 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7790 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7792 (mlong-double-): Ditto.
7793 * config/rs6000/sysv4.opt (msdata=): Ditto.
7794 (mtls-size=): Ditto.
7796 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7798 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7799 erroneous entries for
7800 "vector int vec_ldl (int, long int *)", and
7801 "vector unsigned int vec_ldl (int, unsigned long int *)".
7802 Add comments and entries for
7803 "vector bool char vec_ldl (int, bool char *)",
7804 "vector bool short vec_ldl (int, bool short *)",
7805 "vector bool int vec_ldl (int, bool int *)",
7806 "vector bool long long vec_ldl (int, bool long long *)",
7807 "vector pixel vec_ldl (int, pixel *)",
7808 "vector long long vec_ldl (int, long long *)",
7809 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7810 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7811 type tree bool_long_long_type_node and correct definition of
7812 bool_V2DI_type_node to make reference to this new type tree.
7813 (rs6000_mangle_type): Replace erroneous reference to
7814 bool_long_type_node with bool_long_long_type_node.
7815 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7816 comments to emphasize sign distinctions for char and int types and
7817 replace RS6000_BTI_bool_long constant with
7818 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7819 use of RS6000_BTI_pixel.
7820 (bool_long_type_node): Remove this macro definition.
7821 (bool_long_long_type_node): New macro definition
7823 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7825 PR rtl-optimization/85300
7826 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7827 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7828 simplify_unary_operation fails.
7830 2018-04-10 Martin Liska <mliska@suse.cz>
7832 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7833 cgraph_edge and ipa_ref.
7835 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7839 * config/i386/sse.md
7840 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7841 computation of the VEC_MERGE selector from mask.
7842 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7843 Fix decoding of the VEC_MERGE selector into mask.
7845 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7847 PR tree-optimization/85286
7848 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7850 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7852 * final.c (final_1): Set insn_last_address as well as
7853 insn_current_address.
7855 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7858 * explow.c (emit_stack_probe): Call validize_mem on memory location
7859 before passing it to gen_probe_stack. Create address operand and
7860 legitimize it for the probe_stack_address case.
7862 2018-04-09 Jan Hubicka <jh@suse.cz>
7865 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7866 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7867 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7869 (free_lang_data): Rebuild type inheritance graph.
7871 2018-04-09 Martin Sebor <msebor@redhat.com>
7873 * invoke.texi (-finline-small-functions): Mention other optimization
7875 (-findirect-inlining, -fpartial-inlining): Same.
7876 (-finline-functions-called-once): Same.
7877 (-freorder-blocks-and-partition): Same.
7879 2018-04-09 Jan Hubicka <jh@suse.cz>
7882 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7883 jumps; choose last target that matches the criteria (i.e.
7884 no partition changes for non-crossing jumps).
7885 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7886 support for redirecting crossing jumps to non-crossing.
7888 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7890 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7891 also for naked functions.
7893 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7895 * config/arc/arc.md (add_shift): New pattern.
7896 (add_shift2): Likewise.
7897 (sub_shift): Likewise.
7898 (sub_shift_cmp0_noout): Likewise.
7899 (compare_si_ashiftsi): Likewise.
7900 (xbfu_cmp0_noout): New combine pattern.
7901 (xbfu_cmp0"): Likewise.
7902 (movsi_set_cc_insn): Place the predicable variant first.
7903 (commutative_binary_cmp0_noout): Remove clobber.
7904 (commutative_binary_cmp0): New pattern.
7905 (noncommutative_binary_cmp0): Likewise.
7906 (noncommutative_binary_cmp0_noout): Likewise.
7907 (noncommutative_binary_comparison_result_used): Removed.
7908 (rsub_cmp0): New pattern.
7909 (rsub_cmp0_noout): Likewise.
7910 (extzvsi): Changed, keep only meaningful variants.
7911 (SQH, SEZ): New iterators.
7912 (SQH_postfix): New mode attribute.
7913 (SEZ_prefix): New code attribute.
7914 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7915 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7916 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7918 (noncommutative_operator): Check the availability of barrel
7921 2018-04-09 Richard Biener <rguenther@suse.de>
7923 PR tree-optimization/85284
7924 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7925 Only use the niter constraining form of simple_iv when the exit
7928 2018-04-09 Tom de Vries <tom@codesourcery.com>
7931 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7932 (define_expand "*memory_barrier"): New define_expand.
7933 (define_insn "memory_barrier"): New insn.
7935 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7937 PR rtl-optimization/80463
7938 PR rtl-optimization/83972
7939 PR rtl-optimization/83480
7941 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7942 correct producer for the insn.
7943 (tidy_control_flow): Fixup seqnos in case of debug insns.
7945 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7947 PR rtl-optimization/83913
7949 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7950 different sched-times when merging exprs.
7952 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7954 PR rtl-optimization/83962
7956 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7957 tidy_fallthru_edge and tidy_control_flow.
7959 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7961 PR rtl-optimization/83530
7963 * sel-sched.c (force_next_insn): New global variable.
7964 (remove_insn_for_debug): When force_next_insn is true, also leave only
7965 next insn in the ready list.
7966 (sel_sched_region): When the region wasn't scheduled, make another pass
7967 over it with force_next_insn set to 1.
7969 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7971 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7973 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7974 for interrupt control.
7975 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7976 functions for interrupt control.
7977 * config/nds32/nds32-intrinsic.md: Likewise.
7978 * config/nds32/nds32_intrinsic.h: Likewise.
7979 * config/nds32/nds32.h (nds32_builtins): Likewise.
7981 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7983 * config/nds32/nds32.c (nds32_init_machine_status,
7984 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7985 strict_aligned_p field.
7986 (nds32_expand_to_rtl_hook): New function.
7987 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7988 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7990 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7991 Chung-Ju Wu <jasonwucj@gmail.com>
7993 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7994 * config/nds32/nds32-n7.md: New file.
7995 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7996 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7998 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7999 * config/nds32/nds32.md (pipeline_model): Add n7.
8000 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8001 * config/nds32/pipelines.md: Include n7 settings.
8003 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8004 Chung-Ju Wu <jasonwucj@gmail.com>
8006 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8007 * config/nds32/nds32-e8.md: New file.
8008 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8009 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8011 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8012 * config/nds32/nds32.md (pipeline_model): Add e8.
8013 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8014 * config/nds32/pipelines.md: Include e8 settings.
8016 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8017 Chung-Ju Wu <jasonwucj@gmail.com>
8019 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8020 * config/nds32/nds32-n8.md: New file.
8021 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8022 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8024 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8025 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8026 * config/nds32/nds32.md (pipeline_model): Add n8.
8027 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8028 * config/nds32/pipelines.md: Include n8 settings.
8030 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8031 Chung-Ju Wu <jasonwucj@gmail.com>
8033 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8034 * config/nds32/nds32-n9-2r1w.md: New file.
8035 * config/nds32/nds32-n9-3r2w.md: New file.
8036 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8037 nds32_register_ports): New or modify for cpu n9.
8038 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8040 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8041 * config/nds32/nds32-utils.c: New file.
8042 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8043 TARGET_MUL_SLOW): Define.
8044 * config/nds32/nds32.md (pipeline_model): New attribute.
8045 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8046 New options that support cpu n9.
8047 * config/nds32/pipelines.md: Include n9 settings.
8048 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8050 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8052 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8053 information if necessary.
8054 (output_cond_branch_compare_zero): Likewise.
8055 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8056 (nds32_target_alignment): Refine for alignment.
8057 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8058 (FUNCTION_BOUNDARY): Modify.
8059 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8061 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8063 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8065 * config/nds32/constants.md (unspec_volatile_element): Add values for
8066 TLB operation and data prefetch.
8067 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8068 functions for TLB operation and data prefetch.
8069 * config/nds32/nds32-intrinsic.md: Likewise.
8070 * config/nds32/nds32_intrinsic.h: Likewise.
8071 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8072 (nds32_print_operand): Likewise.
8073 * config/nds32/nds32.h (nds32_builtins): Likewise.
8075 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8076 Andrew Pinski <pinsika@gcc.gnu.org>
8079 * match.pd: Use constant_boolean_node of correct type instead of
8080 boolean_true_node or boolean_false_node for simplifying
8081 pointer comparisons to zero.
8083 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8085 PR tree-optimization/80021
8086 * tree.c (verify_type_variant): Make error call in verify_variant_match
8087 translatable and remove final full stop.
8089 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8091 * config/nds32/constants.md (unspec_volatile_element): Add
8092 UNSPEC_VOLATILE_EH_RETURN.
8093 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8094 nds32_output_stack_pop): Support dwarf exception handling process.
8095 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8096 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8097 exception handling process.
8098 (nds32_compute_stack_frame): Likewise.
8099 (nds32_return_addr_rtx): Likewise.
8100 (nds32_initial_elimination_offset): Likewise.
8101 (nds32_expand_prologue): Likewise.
8102 (nds32_expand_epilogue): Likewise.
8103 (nds32_dynamic_chain_address): New function.
8104 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8106 (DYNAMIC_CHAIN_ADDRESS): Define.
8107 (EH_RETURN_DATA_REGNO): Define.
8108 (EH_RETURN_STACKADJ_RTX): Define.
8109 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8110 patterns for dwarf exception handling.
8112 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8114 * config/nds32/nds32.h: Clean up obsolete macros.
8116 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8118 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8119 Add enum values for particular instructions.
8120 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8121 particular intrinsic functions.
8122 * config/nds32/nds32-intrinsic.md: Likewise.
8123 * config/nds32/nds32_intrinsic.h: Likewise.
8124 * config/nds32/nds32.h (nds32_builtins): Likewise.
8125 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8126 (btst, ave): New patterns for particular instructions.
8128 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8130 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8131 Add enum values for atomic load/store and memory sync.
8132 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8134 * config/nds32/nds32-intrinsic.md: Likewise.
8135 * config/nds32/nds32_intrinsic.h: Likewise.
8136 * config/nds32/nds32.h (nds32_builtins): Likewise.
8138 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8140 PR tree-optimization/85257
8141 * fold-const.c (native_encode_vector): If not all elts could fit
8142 and off is -1, return 0 rather than offset.
8143 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8144 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8145 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8146 adjust buffer in native_interpret_expr call.
8148 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8150 * config/nds32/constants.md (unspec_volatile_element): Add cache
8151 control enum values.
8152 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
8153 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
8154 * config/nds32/nds32.c (nds32_cctl_names): New.
8155 (nds32_print_operand): Handle cache control register names.
8156 * config/nds32/nds32.h (nds32_builtins): New enum values.
8157 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
8159 * config/nds32/nds32.md (type): Add mmu.
8160 * config/nds32/pipelines.md (simple_insn): Add mmu.
8162 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8164 * config/nds32/nds32.md (type): Remove call.
8165 * config/nds32/pipelines.md (simple_insn): Likewise.
8167 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8169 * config/nds32/constants.md (unspec_volatile_element): Add
8170 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
8171 UNSPEC_VOLATILE_FMFCFG.
8172 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
8173 description for fmfcfg and fmfcsr.
8174 (bdesc_1arg): Add fmtcsr.
8175 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
8176 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
8177 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
8178 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
8179 unspec_fmfcfg): New patterns.
8180 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
8181 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
8182 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
8183 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
8184 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
8185 __nds32__fmfcfg): Define.
8187 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8189 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
8190 intrinsic register names.
8191 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
8192 intrinsic register enum values and macros.
8194 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8196 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
8197 for load/store addressing form.
8198 (nds32_print_operand_address): Likewise.
8200 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
8203 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
8204 based on LABEL_REF. Remove useless assertion.
8205 (pic_address_needs_scratch): Fix formatting.
8206 (sparc_legitimize_pic_address): Minor tweaks.
8207 (sparc_delegitimize_address): Adjust assertion accordingly.
8208 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
8209 into symbolic_operand.
8210 (movsi_high_pic_label_ref): Likewise.
8211 (movsi_lo_sum_pic_label_ref): Likewise.
8212 (movdi_pic_label_ref): Likewise.
8213 (movdi_high_pic_label_ref): Likewise.
8214 (movdi_lo_sum_pic_label_ref): Likewise.
8216 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
8218 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
8219 custom LIB_SPEC setup.
8221 2018-04-06 Ruslan Bukin <br@bsdpad.com>
8222 Kito Cheng <kito.cheng@gmail.com>
8224 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
8225 * config/riscv/freebsd.h: New.
8227 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8229 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
8230 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
8233 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8234 Kito Cheng <kito.cheng@gmail.com>
8236 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
8237 nds32_output_call, nds32_symbol_binds_local_p): New functions.
8238 * config/nds32/nds32-protos.h (nds32_output_call,
8239 nds32_output_return): Declare.
8240 * config/nds32/nds32.md: Refine all the call and return patterns.
8242 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8245 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
8246 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
8248 PR rtl-optimization/84872
8249 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
8250 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
8253 2018-04-06 Tamar Christina <tamar.christina@arm.com>
8255 * expr.c (copy_blkmode_to_reg): Revert 254862.
8256 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
8258 2018-04-06 Richard Biener <rguenther@suse.de>
8261 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
8262 after seeing a component reference with an adjacent field. Treat
8263 refs to arrays at struct end of external decls similar to
8264 refs to unconstrained commons.
8266 2018-04-06 Jakub Jelinek <jakub@redhat.com>
8269 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
8270 look through SAVE_EXPRs with non-side-effects argument. Adjust
8272 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
8275 2018-04-06 Richard Biener <rguenther@suse.de>
8278 * alias.c (find_base_term): New wrapper around find_base_term
8279 unwinding CSELIB_VAL_PTR changes.
8280 (find_base_term): Do not restore CSELIB_VAL_PTR during the
8283 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8285 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
8287 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
8288 constant definitions.
8289 ("nop"): lr 0,0 -> nopr r0
8290 ("nop_lr0", "nop_lr1"): New insn definitions.
8292 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
8294 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
8295 NDS32_V3PUSH_AVAILABLE_P macro.
8297 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
8298 Chung-Ju Wu <jasonwucj@gmail.com>
8300 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
8301 (nds32*-*-*): Add float and fpu_config into supported_defaults.
8302 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
8303 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
8304 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
8305 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
8306 * config/nds32/constraints.md: New constraints and checking for hard
8307 float configuration.
8308 * config/nds32/iterators.md: New mode iterator and attribute for hard
8309 float configuration.
8310 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
8312 * config/nds32/nds32-fpu.md: New file.
8313 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
8314 deal with hard float code generation.
8315 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
8317 (abi_type, float_reg_number): New enum type.
8318 * config/nds32/nds32-predicates.c: New predicates for hard float.
8319 * config/nds32/nds32-protos.h: Declare functions for hard float.
8320 * config/nds32/nds32.c: Implementation for hard float configuration.
8321 * config/nds32/nds32.h: Definitions for hard float configuration.
8322 * config/nds32/nds32.md: Include hard float machine description and
8323 modify patterns for hard float configuration.
8324 * config/nds32/nds32.opt: New options for hard float configuration.
8325 * config/nds32/predicates.md: New predicates for hard float
8328 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8330 * common/config/nds32/nds32-common.c
8331 (nds32_option_optimization_table): Enable -mreleax-hint by default.
8333 2018-04-05 Jakub Jelinek <jakub@redhat.com>
8336 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
8337 CONSTRUCTOR_ELT (ctor, ...)->value.
8339 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
8342 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
8344 2018-04-05 Tom de Vries <tom@codesourcery.com>
8347 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
8350 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8351 Kito Cheng <kito.cheng@gmail.com>
8353 * config/nds32/constraints.md (U33): Fine-tune checking condition.
8354 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
8355 * config/nds32/nds32.h (nds32_16bit_address_type): Add
8356 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
8358 2018-04-05 Shiva Chen <shiva0217@gmail.com>
8359 Kito Cheng <kito.cheng@gmail.com>
8361 * config/nds32/constraints.md (Ufe): New memory constraint.
8362 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
8363 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
8364 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
8366 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
8367 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
8369 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8371 * config/nds32/nds32.md: Use optimize_size in the condition for
8372 alu-shift instructions.
8374 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8376 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
8378 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8380 * config/nds32/nds32.md (negsi2): Refine pattern.
8382 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8383 Chung-Ju Wu <jasonwucj@gmail.com>
8385 * config/nds32/iterators.md (shift_rotate): New code iterator.
8386 (shift): New code attribute.
8387 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
8388 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
8389 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
8390 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
8391 bit-wise operations.
8392 (andsi3, *andsi3): Ditto.
8393 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
8394 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
8395 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
8396 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
8397 nds32_ior_operand, nds32_xor_operand): New predicates.
8399 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8401 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
8402 (addsi3, subsi3): ... this.
8404 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8406 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
8408 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8410 * config/nds32/nds32.md: Adjust indention.
8412 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
8414 * config/nds32/nds32.md (feature): New attribute.
8416 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
8418 * config/nds32/nds32.md (subtype): New attribute.
8420 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
8423 * config/arm/arm-builtins.c (arm_expand_builtin): Change
8424 expansion to perform a bitwise AND of the argument followed by a
8425 boolean negation of the result.
8427 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
8429 PR rtl-optimization/84878
8430 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
8431 the basic block. Assert the use reference is not artificial and that
8432 it has an associated insn.
8434 2018-04-04 Michael Matz <matz@suse.de>
8436 * builtins.c (compute_objsize): Pass correct operand
8437 to array_at_struct_end_p.
8439 2018-04-04 Richard Biener <rguenther@suse.de>
8442 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
8443 from contexts for DINFO_LEVEL_TERSE and below.
8445 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8447 * config/nds32/nds32-doubleword.md (move_<mode>): Require
8448 resiter_operand condition.
8449 * config/nds32/nds32.md (*move<mode>): Ditto.
8451 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8452 Monk Chiang <sh.chiang04@gmail.com>
8454 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
8456 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8458 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
8460 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8461 Kito Cheng <kito.cheng@gmail.com>
8463 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
8464 nds32_cond_code_str, output_cond_branch,
8465 output_cond_branch_compare_zero, nds32_expand_cbranch,
8466 nds32_expand_cstore, nds32_expand_movcc,
8467 nds32_output_cbranchsi4_equality_zero,
8468 nds32_output_cbranchsi4_equality_reg,
8469 nds32_output_cbranchsi4_equality_reg_or_const_int,
8470 nds32_output_cbranchsi4_greater_less_zero: New functions.
8471 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
8472 nds32_expand_cstore, nds32_expand_movcc,
8473 nds32_output_cbranchsi4_equality_zero,
8474 nds32_output_cbranchsi4_equality_reg,
8475 nds32_output_cbranchsi4_equality_reg_or_const_int,
8476 nds32_output_cbranchsi4_greater_less_zero): Declare.
8477 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
8478 nds32_rimm11s_operand): New predicates.
8479 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
8480 * config/nds32/nds32.md: Rewrite all the branch and conditional move
8483 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
8485 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
8486 * config/nds32/nds32.md: Ditto.
8487 * config/nds32/pipelines.md: Ditto.
8489 2018-04-04 Richard Biener <rguenther@suse.de>
8491 PR tree-optimization/85168
8492 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
8493 propagating abnormals.
8495 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8497 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
8499 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
8500 Kito Cheng <kito.cheng@gmail.com>
8502 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
8503 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
8504 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
8505 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
8506 * config/nds32/nds32.md (sibcall_internal): New.
8507 (sibcall_register): Remove.
8508 (sibcall_immediate): Remove.
8509 (sibcall_value_internal): New.
8510 (sibcall_value_register): Remove.
8511 (sibcall_value_immediate): Remove.
8512 * config/nds32/predicates.md (nds32_general_register_operand): New.
8513 (nds32_call_address_operand): New.
8515 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8517 PR rtl-optimization/85167
8518 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
8519 bb_defs if *split_p, instead preinitialize it to NULL.
8521 PR tree-optimization/85156
8522 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
8523 evaluating the argument multiple times.
8525 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8527 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8529 (_mm_cvtpd_ps): Likewise.
8530 (_mm_cvttpd_epi32): Likewise.
8531 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8532 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8533 vector, pixel, and bool following altivec.h include.
8535 2018-04-03 Martin Sebor <msebor@redhat.com>
8537 * doc/extend.texi (Common Function Attributes): Clarify.
8538 (const attribute): Likewise.
8539 (pure attribute): Likewise.
8541 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8544 * config/i386/i386.c (ix86_expand_vector_set): Use
8545 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8547 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8549 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8550 instructions when changing rounding bits to preserve precision bits
8551 in the x87 control word.
8553 2018-04-03 Martin Liska <mliska@suse.cz>
8555 PR tree-optimization/82491
8556 * rtl.h (strip_offset_and_add): Replace += suboffset with
8557 poly_uint64 () + suboffset.
8559 2018-03-29 Martin Liska <mliska@suse.cz>
8560 Martin Jambor <mjambor@suse.cz>
8563 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8564 param_type is not an integral or pointer type.
8566 2018-04-03 Richard Biener <rguenther@suse.de>
8568 * sese.h (recompute_all_dominators): Remove.
8570 2018-04-02 Martin Sebor <msebor@redhat.com>
8572 * doc/invoke.texi (-Wrestrict): Fix typos.
8574 2018-04-02 Jim Wilson <jimw@sifive.com>
8576 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8577 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8578 (<optab>di3, <optab>si3_extend): Likewise.
8579 (<optab>si3_mask, <optab>si3_mask_1): New.
8580 (<optab>di3_mask, <optab>di3_mask_1): New.
8581 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8582 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8583 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8585 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8587 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8590 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8592 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8593 (nds32_canonicalize_comparison): New function.
8595 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8596 Kito Cheng <kito.cheng@gmail.com>
8597 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8599 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8600 * config/nds32/constants.md (unspec_volatile_element): Add
8601 UNSPEC_VOLATILE_RELAX_GROUP.
8602 * config/nds32/nds32-relax-opt.c: New file.
8603 * config/nds32/nds32-predicates.c
8604 (nds32_symbol_load_store_p): New function.
8605 * config/nds32/nds32-protos.h
8606 (nds32_symbol_load_store_p): Declare function.
8607 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8608 * config/nds32/nds32.c
8609 (nds32_register_pass): New function to register pass.
8610 (nds32_register_passes): New function to register passes.
8611 * config/nds32/nds32.md (relax_group): New pattern.
8612 * config/nds32/nds32.opt (mrelax-hint): New option.
8613 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8615 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8617 * config/nds32/t-nds32: Modify files dependency.
8619 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8621 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8622 (PROFILE_HOOK): Define its implementation.
8624 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8626 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8627 type and 32-bit size.
8629 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8632 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8633 (V_128_256): New mode iterator.
8634 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8635 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8636 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8638 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8639 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8641 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8644 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8645 NaN inputs correctly.
8647 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8650 * config/rs6000/vsx.md (??r): New mode attribute.
8651 (*vsx_mov<mode>_64bit): Use it.
8652 (*vsx_mov<mode>_32bit): Likewise.
8654 2018-03-30 Martin Sebor <msebor@redhat.com>
8656 PR tree-optimization/84818
8657 * builtins.c (check_access): Use warning_n.
8659 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8662 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8664 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8667 2018-03-30 Julia Koval <julia.koval@intel.com>
8670 * x86-tune.def (movx, partial_reg_dependency): Enable for
8673 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8676 * lra-constraints.c (process_alt_operands): Move setting
8677 this_alternative_matches below.
8679 2018-03-29 Martin Liska <mliska@suse.cz>
8682 * doc/invoke.texi: Document how LTO works with debug info.
8683 Describe auto-load support of binutils. Mention 'x86-64'
8684 as valid option value of -march option.
8686 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8688 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8691 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8692 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8693 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8696 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8699 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8700 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8701 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8702 for RS6000_BTM_POWERPC64.
8703 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8704 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8705 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8710 2018-03-28 Carl Love <cel@us.ibm.com>
8713 2017-09-27 Carl Love <cel@us.ibm.com>
8715 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8716 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8717 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8720 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8722 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8723 instead of __vector bool.
8724 (_mm_max_pu8): Likewise.
8725 (_mm_min_pi16): Likewise.
8727 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8730 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8731 (DIVWEUO): Likewise.
8733 (DIVDEUO): Likewise.
8734 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8735 DIVWEUO and DIVDEUO.
8736 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8737 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8738 (div_extend): Likewise.
8739 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8741 (__builtin_divweuo): Likewise.
8742 (__builtin_divdeo): Likewise.
8743 (__builtin_divdeuo): Likewise.
8745 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8748 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8749 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8751 PR tree-optimization/82004
8752 * gimple-match-head.c (optimize_pow_to_exp): New function.
8753 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8754 Don't fold to exp if optimize_pow_to_exp is false.
8756 2018-03-28 Martin Liska <mliska@suse.cz>
8759 * calls.c (initialize_argument_information): Fix trailing space.
8760 * common.opt: Fix typo and provide better explanation for
8761 -fsanitize-coverage option.
8762 * config/i386/i386.opt: Fix typo.
8764 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8765 Martin Liska <mliska@suse.cz>
8768 * gimplify.c (asan_poison_variable): Don't do the check for
8769 gimplify_omp_ctxp here.
8770 (gimplify_decl_expr): Do it here.
8771 (gimplify_target_expr): Likewise.
8773 2018-03-28 Martin Liska <mliska@suse.cz>
8776 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8777 chkp_type_bounds_count if MPX is not enabled.
8779 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8781 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8783 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8786 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8787 function to create the function decl for complex long double
8788 multiply and divide for -mabi=ieeelongdouble.
8789 (init_float128_ieee): Call it.
8791 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8794 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8795 -fcf-protection=branch -mibt.
8796 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8798 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8801 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8802 arm_word_relocations.
8804 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8807 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8808 extern array declarations.
8810 2018-03-27 Richard Biener <rguenther@suse.de>
8813 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8814 explicit single_use checks.
8816 2018-03-27 Richard Biener <rguenther@suse.de>
8818 PR tree-optimization/85082
8819 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8822 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8824 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8825 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8826 Turn on fasynchronous-unwind-tables and funwind-tables.
8828 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8831 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8832 (*bmi_blsr_<mode>_ccz): Ditto.
8834 2018-03-26 Tom de Vries <tom@codesourcery.com>
8836 PR tree-optimization/85063
8837 * omp-general.c (offloading_function_p): New function. Factor out
8839 * omp-offload.c (pass_omp_target_link::gate): ... here.
8840 * omp-general.h (offloading_function_p): Declare.
8841 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8842 with attribute omp declare target for offloading functions.
8844 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8846 PR tree-optimization/84005
8847 * tree-data-ref.h (get_base_for_alignment): Declare.
8848 * tree-data-ref.c (get_base_for_alignment_1): New function.
8849 (get_base_for_alignment): Likewise.
8850 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8851 get_base_for_alignment to find a suitable base object, instead
8852 of always using drb->base_address.
8854 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8857 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8858 known size by default.
8860 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8863 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8864 and non BLKmode operands.
8866 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8869 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8870 Clean up attributes.
8872 2018-03-23 Richard Biener <rguenther@suse.de>
8875 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8876 we are going to emit early debug for LTO.
8878 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8881 * function.c (match_asm_constraints_1): Don't optimize if input
8882 doesn't satisfy general_operand predicate for output's mode.
8885 * alias.c (write_dependence_p): Don't require for x_canonicalized
8886 non-VOIDmode if x has VOIDmode.
8889 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8890 just don't try to optimize it rather than assert it never happens.
8892 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8894 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8895 macro expansions for definition of ST_INTERNAL_<mode> and
8896 LD_INTERNAL_<mode> builtins.
8897 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8899 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8901 (altivec_expand_st_builtin): Likewise.
8902 (altivec_expand_builtin): Remove calls to deleted functions.
8903 (rs6000_address_for_altivec): Delete this function.
8904 * config/rs6000/vector.md: Remove expands for
8905 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8907 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8910 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8911 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8912 re-computing once computed.
8913 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8914 (arm_init_machine_status): Initialize
8915 machine->static_chain_stack_bytes.
8917 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8920 * doc/extend.texi: Add four new prototypes for vec_ld.
8921 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8922 definitions for more logical presentation.
8923 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8924 entries for V1TI variants of __builtin_altivec_ld builtin.
8925 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8926 handling of V1TI variant of LVX icode pattern.
8927 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8928 (rs6000_gimple_fold_builtin): Likewise.
8929 (altivec_init_builtins): Add code to define
8930 __builtin_altivec_lvx_v1ti function.
8932 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8935 * function.c (match_asm_constraints_1): Don't do the optimization
8936 if input isn't a REG, SUBREG, MEM or constant.
8938 2018-03-22 Tom de Vries <tom@codesourcery.com>
8940 PR tree-optimization/84956
8941 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8942 bb_has_abnormal_pred.
8944 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8947 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8948 DECL_INITIAL (decl) to decl at the end.
8949 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8952 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8954 * doc/extend.texi (__builtin_tgmath): Document when complex
8955 integer types are treated as _Complex _Float64.
8957 2018-03-21 Tom de Vries <tom@codesourcery.com>
8959 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8961 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8963 PR tree-optimization/84960
8964 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8965 if it is ENTRY block, move them into single succ of ENTRY in that case.
8967 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8969 PR tree-optimization/84811
8970 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8971 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8972 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8973 (known_subrange_p): Update accordingly. Cast each value involved
8974 in the size comparison, rather than casting the result of the
8977 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8979 PR tree-optimization/84982
8980 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8981 by flipping the least significant bit rather than all bits from
8982 bitpos to bitpos + bitsize - 1.
8984 2018-03-21 Nathan Sidwell <nathan@acm.org>
8986 * doc/extend.texi (Deprecated Features): Remove mention of
8987 long-deleted deprecations.
8989 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8992 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8993 * configure: Regenerate.
8995 2018-03-21 Tom de Vries <tom@codesourcery.com>
8997 PR tree-optimization/83126
8998 * tree-parloops.c (num_phis): New function.
8999 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9001 2018-03-21 Nathan Sidwell <nathan@acm.org>
9003 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9004 mention anon-struct/union members and trailing attributes.
9006 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9008 PR tree-optimization/84969
9009 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9010 builtin memset partitions if they set different rhs values.
9012 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9014 PR rtl-optimization/84989
9015 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9016 VEC_DUPLICATE with scalar result mode.
9018 2018-03-21 Martin Liska <mliska@suse.cz>
9021 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9022 not intended return statement.
9024 2018-03-21 Martin Liska <mliska@suse.cz>
9027 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9028 (chkp_find_bound_slots_1): Limit number of iterations.
9030 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9033 * Minor grammar fixes for x86 options.
9035 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9038 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9039 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9041 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9044 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9045 (altivec_lvx_<mode>_1op): Likewise.
9046 (altivec_stvx_<mode>_2op): Likewise.
9047 (altivec_stvx_<mode>_1op): Likewise.
9048 (altivec_lvx_<VM2:mode>): New define_expand.
9049 (altivec_stvx_<VM2:mode>): Likewise.
9050 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9051 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9052 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9053 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9054 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9055 (rs6000_gen_lvx): Likewise.
9056 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9057 (altivec_expand_stv_builtin): Likewise.
9058 (altivec_expand_builtin): Likewise.
9059 * config/rs6000/vector.md: Likewise.
9061 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9064 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9067 2018-03-20 Richard Biener <rguenther@suse.de>
9070 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9071 sign-conversions as zero, fall back to standard scalar_stmt
9074 2018-03-20 Martin Liska <mliska@suse.cz>
9077 * predict.c (rebuild_frequencies): Handle case when we have
9078 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9080 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9083 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9084 flag_section_anchors.
9085 * varasm.c (use_blocks_for_decl_p): Remove hack for
9086 dw2_force_const_mem.
9089 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9091 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9092 be created, use lowpart_subreg of operands[0] rather than operands[0]
9094 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9095 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9096 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9097 and n constraint instead of aarch64_shift_imm_di and Usd.
9098 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9099 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9101 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9104 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9105 to favor GPR over NEON registers.
9106 (<shift>di3_neon): Likewise.
9108 2018-03-20 Tom de Vries <tom@codesourcery.com>
9111 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9112 (nvptx_process_pars): Emit bar.sync asap and alap.
9114 2018-03-20 Tom de Vries <tom@codesourcery.com>
9117 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9118 seen_label if seen_label is already set.
9120 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9123 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9124 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9125 Use 1U instead of 1. Formatting fixes.
9128 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9129 instead of TREE_TYPE (s1) for the return value.
9131 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9133 PR tree-optimization/84946
9134 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9135 bitsize + bitsize in poly_uint64 rather than poly_int64.
9138 * dwarf2asm.c: Include fold-const.c.
9139 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9140 of decl rather than decl itself.
9142 PR rtl-optimization/84643
9143 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9145 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9148 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9149 calling assemble_variable.
9151 2018-03-19 Sudakshina Das <sudi.das@arm.com>
9154 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
9155 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
9157 2018-03-19 Jim Wilson <jimw@sifive.com>
9160 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
9161 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
9162 (riscv_first_stack_step): Likewise.
9163 (riscv_option_override): Use STACK_BOUNDARY instead of
9165 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
9167 (BIGGEST_ALIGNMENT): Set to 128.
9168 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
9169 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
9172 2018-03-19 Richard Biener <rguenther@suse.de>
9174 PR tree-optimization/84933
9175 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
9176 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
9178 2018-03-19 Richard Biener <rguenther@suse.de>
9180 PR tree-optimization/84859
9181 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
9182 (cond_if_else_store_replacement): Perform sinking operation on
9183 single-store BBs regardless of MAX_STORES_TO_SINK setting.
9184 Generalize what a BB with a single eligible store is.
9186 2018-03-19 Richard Biener <rguenther@suse.de>
9188 PR tree-optimization/84929
9189 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
9190 chrec_is_positive against non-chrec arg.
9192 2018-03-19 Tamar Christina <tamar.christina@arm.com>
9195 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
9197 2018-03-18 Martin Liska <mliska@suse.cz>
9199 PR rtl-optimization/84635
9200 * regrename.c (build_def_use): Use matches_mode only when
9203 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
9205 PR tree-optimization/84913
9206 * tree-vect-loop.c (vectorizable_reduction): Don't try to
9207 vectorize chains of COND_EXPRs.
9209 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9211 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
9213 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9215 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
9217 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
9219 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
9221 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9222 Kito Cheng <kito.cheng@gmail.com>
9224 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
9225 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
9226 (nds32_adjust_reg_alloc_order): New function.
9227 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
9229 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
9231 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
9232 nds32_print_operand, nds32_print_operand_address): Use
9233 HOST_WIDE_INT_PRINT_DEC instead.
9235 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
9237 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
9239 2018-03-17 Jakub Jelinek <jakub@redhat.com>
9242 * config/i386/i386.c (initial_ix86_tune_features,
9243 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
9245 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
9246 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
9247 rather than 1u << ix86_tune. Formatting fix.
9248 (ix86_option_override_internal): Change ix86_arch_mask from
9249 unsigned int to unsigned HOST_WIDE_INT, initialize to
9250 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
9251 (ix86_function_specific_restore): Likewise.
9253 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9256 * postreload.c (reload_combine_recognize_pattern): Perform
9257 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
9258 truncate_int_for_mode the result for the destination's mode.
9261 * hsa-gen.c (mem_type_for_type): Fix comment typo.
9262 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
9264 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9267 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
9270 * lra-assigns.c (lra_split_hard_reg_for): Don't use
9271 regno_allocno_class_array and sorted_pseudos.
9272 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
9273 insns where regno is used.
9275 2018-03-16 Martin Liska <mliska@suse.cz>
9278 * multiple_target.c (create_dispatcher_calls): Redirect
9279 reference in the symbol table.
9281 2018-03-16 Martin Liska <mliska@suse.cz>
9284 * multiple_target.c (create_dispatcher_calls): Redirect also
9287 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9291 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
9292 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
9293 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
9295 2018-03-16 Julia Koval <julia.koval@intel.com>
9297 * doc/invoke.texi (Skylake Server): Add CLWB.
9298 Cannonlake): Remove CLWB.
9300 2018-03-16 Jakub Jelinek <jakub@redhat.com>
9302 PR tree-optimization/84841
9303 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
9305 (FLOAT_ONE_CONST_TYPE): Define.
9306 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
9307 (sort_by_operand_rank): Put entries with higher constant_type last
9308 rather than first to match comments.
9310 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
9312 * config/nios2/nios2.md (movsi_internal): Fix thinko in
9315 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9318 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
9319 check and use address of target always.
9321 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
9324 * config/i386/i386.c (indirect_thunk_needed): Update comments.
9325 (indirect_thunk_bnd_needed): Likewise.
9326 (indirect_thunks_used): Likewise.
9327 (indirect_thunks_bnd_used): Likewise.
9328 (indirect_return_needed): New.
9329 (indirect_return_bnd_needed): Likewise.
9330 (output_indirect_thunk_function): Add a bool argument for
9332 (output_indirect_thunk_function): Don't generate alias for
9333 function return thunk.
9334 (ix86_code_end): Call output_indirect_thunk_function to generate
9335 function return thunks.
9336 (ix86_output_function_return): Set indirect_return_bnd_needed
9337 and indirect_return_needed instead of indirect_thunk_bnd_needed
9338 and indirect_thunk_needed.
9340 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
9342 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
9343 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
9344 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
9346 2018-03-15 David Malcolm <dmalcolm@redhat.com>
9347 Paul Hua <paul.hua.gm@gmail.com>
9350 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
9352 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
9354 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
9355 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
9358 2018-03-15 Tamar Christina <tamar.christina@arm.com>
9361 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9362 instead of GET_MODE_SIZE when comparing Units.
9364 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
9367 * varasm.c (hash_section): Return an unchangeble hash value
9368 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
9369 Return !aarch64_can_use_per_function_literal_pools_p ().
9371 2018-03-15 Jakub Jelinek <jakub@redhat.com>
9374 * optabs.c (emit_conditional_move): Pass address of cmode's copy
9375 rather than address of cmode as last argument to prepare_cmp_insn.
9377 2018-03-15 Julia Koval <julia.koval@intel.com>
9379 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
9380 F_AVX512VNNI, F_AVX512BITALG): New.
9382 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
9385 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
9386 insn for floating-point loads and stores.
9388 2018-03-14 Carl Love <cel@us.ibm.com>
9390 * config/rs6000/rs6000-c.c: Add macro definitions for
9391 ALTIVEC_BUILTIN_VEC_PERMXOR.
9392 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
9393 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
9394 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
9395 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
9397 * config/doc/extend.texi: Add prototypes for vec_permxor.
9399 2018-03-14 David Malcolm <dmalcolm@redhat.com>
9402 * diagnostic-show-locus.c (class layout_point): Convert m_line
9403 from int to linenum_type.
9404 (line_span::comparator): Use linenum "compare" function when
9405 comparing line numbers.
9406 (test_line_span): New function.
9407 (layout_range::contains_point): Convert param "row" from int to
9409 (layout_range::intersects_line_p): Likewise.
9410 (layout::will_show_line_p): Likewise.
9411 (layout::print_source_line): Likewise.
9412 (layout::should_print_annotation_line_p): Likewise.
9413 (layout::print_annotation_line): Likewise.
9414 (layout::print_leading_fixits): Likewise.
9415 (layout::annotation_line_showed_range_p): Likewise.
9416 (struct line_corrections): Likewise for field m_row.
9417 (line_corrections::line_corrections): Likewise for param "row".
9418 (layout::print_trailing_fixits): Likewise.
9419 (layout::get_state_at_point): Likewise.
9420 (layout::get_x_bound_for_row): Likewise.
9421 (layout::print_line): Likewise.
9422 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
9423 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
9424 * input.c (selftest::test_linenum_comparisons): New function.
9425 (selftest::input_c_tests): Call it.
9426 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
9427 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
9428 * selftest.h (ASSERT_GT): New macro.
9429 (ASSERT_GT_AT): New macro.
9430 (ASSERT_LT): New macro.
9431 (ASSERT_LT_AT): New macro.
9433 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
9435 PR rtl-optimization/84780
9436 * combine.c (distribute_links): Don't make a link based on pc_rtx.
9438 2018-03-14 Martin Liska <mliska@suse.cz>
9440 * tree.c (record_node_allocation_statistics): Use
9441 get_stats_node_kind.
9442 (get_stats_node_kind): New function extracted from
9443 record_node_allocation_statistics.
9444 (free_node): Use get_stats_node_kind.
9446 2018-03-14 Richard Biener <rguenther@suse.de>
9448 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
9449 that the value-set of ANTIC_IN doesn't grow.
9452 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9454 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9455 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9457 2018-03-14 Julia Koval <julia.koval@intel.com>
9459 * config.gcc (icelake-client, icelake-server): New.
9461 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
9462 (initial_ix86_arch_features): Ditto.
9463 (PTA_SKYLAKE): Add SGX.
9464 (PTA_ICELAKE): Remove.
9465 (PTA_ICELAKE_CLIENT): New.
9466 (PTA_ICELAKE_SERVER): New.
9467 (ix86_option_override_internal): Split up icelake on icelake client and
9469 (get_builtin_code_for_version): Ditto.
9470 (fold_builtin_cpu): Ditto.
9471 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
9472 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
9473 * config/i386/i386.h (processor_type): Ditto.
9474 * doc/invoke.texi: Ditto.
9476 2018-03-14 Jakub Jelinek <jakub@redhat.com>
9479 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
9480 INTEGER_CST offset, add it together with bitpos / 8 and
9481 sign extend based on POINTER_SIZE.
9485 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
9488 * config/i386/constraints.md (Yc): New register constraint.
9489 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
9490 Use Yc constraint for alternative 2 of operand 0. Remove
9491 preferred_for_speed attribute.
9493 2018-03-14 Richard Biener <rguenther@suse.de>
9495 PR tree-optimization/84830
9496 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
9497 with the old one to avoid oscillations.
9499 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
9502 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
9504 (assign_by_spills): Return a flag of reload assignment failure.
9505 Do not process the reload assignment failures. Do not spill other
9506 reload pseudos if they has the same reg class. Update n if
9508 (lra_assign): Add a return arg. Set up from the result of
9509 assign_by_spills call.
9510 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9511 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9512 usage_insns if it is not NULL.
9513 (spill_hard_reg_in_range): New function.
9514 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9515 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9516 function prototypes.
9517 (lra_assign): Change prototype.
9518 * lra.c (lra): Add code to deal with fails by splitting hard reg
9521 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
9523 * config/riscv/riscv.opt (mrelax): New option.
9524 * config/riscv/riscv.c (riscv_file_start): Emit ".option
9525 "norelax" when riscv_mrelax is disabled.
9526 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9528 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9531 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9532 reassociation for int modes.
9534 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9536 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9537 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9539 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9540 * config/aarch64/aarch64-sve.md
9541 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9542 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9543 (*extend<mode><Vwide>2): Rename to...
9544 (aarch64_sve_extend<mode><Vwide>2): ...this.
9545 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9546 renaming the old pattern to...
9547 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9549 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9550 define_expand, renaming the old pattern to...
9551 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9552 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9553 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9554 account when deciding which SVE instruction the optab should use.
9555 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9557 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9559 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9560 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9561 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9562 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9563 (tlsdesc_small_<mode>): Turn a define_expand and use
9564 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9565 (tlsdesc_small_advsimd_<mode>): ...this.
9566 (tlsdesc_small_sve_<mode>): New pattern.
9568 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9570 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9571 (UNSPEC_UMUL_HIGHPART): New constants.
9572 (MUL_HIGHPART): New int iteraor.
9573 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9574 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9576 (*<su>mul<mode>3_highpart): New define_insn.
9578 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9581 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9584 2018-03-13 Martin Liska <mliska@suse.cz>
9587 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9589 (sem_item_optimizer::~sem_item_optimizer): Release it.
9590 (sem_item_optimizer::merge_classes): Register variable aliases.
9591 (sem_item_optimizer::fixup_pt_set): New function.
9592 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9593 * ipa-icf.h: Declare new variables and functions.
9595 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9598 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9599 integer_pow2p@2 and test integer_pow2p in condition.
9600 (A < 0 ? C : 0): Similarly for @1.
9603 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9604 characters starting at p contain '\0' character, don't look beyond
9608 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9609 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9612 * reg-stack.c (change_stack): Change update_end var from int to
9613 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9614 also call set_block_for_insn on the newly added insns and rescan.
9617 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9618 on the last operand.
9621 * tree.c (stabilize_reference_1): Return save_expr (e) for
9622 STATEMENT_LIST even if it doesn't have side-effects.
9624 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9626 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9628 2018-03-12 Renlin Li <renlin.li@arm.com>
9630 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9631 aarch64_output_scalar_simd_mov_immediate.
9633 2018-03-12 Martin Sebor <msebor@redhat.com>
9635 PR tree-optimization/83456
9636 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9637 for perfectly overlapping calls to memcpy.
9638 (gimple_fold_builtin_memory_chk): Same.
9639 (gimple_fold_builtin_strcpy): Handle no-warning.
9640 (gimple_fold_builtin_stxcpy_chk): Same.
9641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9643 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9645 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9646 parameter. Use it for SFmode.
9647 (rs6000_function_arg_advance_1): Adjust.
9648 (rs6000_function_arg): Adjust.
9649 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9651 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9653 PR rtl-optimization/84169
9654 PR rtl-optimization/84780
9655 * combine.c (can_combine_p): Check for a 2-insn combination whether
9656 the destination register is used between the two insns, too.
9658 2018-03-12 Richard Biener <rguenther@suse.de>
9660 PR tree-optimization/84803
9661 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9662 for refs DR analysis didn't process.
9664 2018-03-12 Richard Biener <rguenther@suse.de>
9666 PR tree-optimization/84777
9667 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9668 force-vectorize loops ignore whether we are optimizing for size.
9670 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9672 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9673 (TARGET_MD_ASM_ADJUST): Define.
9675 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9676 Kito Cheng <kito.cheng@gmail.com>
9677 Chung-Ju Wu <jasonwucj@gmail.com>
9679 * config/nds32/nds32.c (nds32_compute_stack_frame,
9680 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9681 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9682 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9683 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9684 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9685 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9686 * config/nds32/nds32.md (prologue, epilogue): Use macro
9687 NDS32_V3PUSH_AVAILABLE_P to do checking.
9689 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9692 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9693 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9694 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9695 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9696 addition of most attributes on !orig_type_die or the attribute not
9697 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9699 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9700 Chung-Ju Wu <jasonwucj@gmail.com>
9702 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9704 * config/nds32/nds32.opt (mvh): New option.
9706 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9707 Chung-Ju Wu <jasonwucj@gmail.com>
9709 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9711 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9712 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9715 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9716 Chung-Ju Wu <jasonwucj@gmail.com>
9718 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9720 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9721 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9723 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9724 Kito Cheng <kito.cheng@gmail.com>
9725 Chung-Ju Wu <jasonwucj@gmail.com>
9727 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9728 UNSPEC_FFMISM and UNSPEC_FLMISM.
9729 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9730 for ffb, ffmism and flmism.
9731 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9732 (unspec_ffmism): Ditto.
9733 (unspec_flmism): Ditto.
9734 (nds32_expand_builtin_impl): Check if string extension is available.
9735 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9736 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9738 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9741 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9744 * lra-assigns.c (assign_by_spills): Return a flag of reload
9745 assignment failure. Do not process the reload assignment
9746 failures. Do not spill other reload pseudos if they has the same
9748 (lra_assign): Add a return arg. Set up from the result of
9749 assign_by_spills call.
9750 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9751 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9752 usage_insns if it is not NULL.
9753 (spill_hard_reg_in_range): New function.
9754 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9755 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9756 function prototypes.
9757 (lra_assign): Change prototype.
9758 * lra.c (lra): Add code to deal with fails by splitting hard reg
9761 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9764 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9766 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9769 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9770 (dw_val_node): Add val_symbolic_view.
9771 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9772 (symview_upper_bound): New.
9773 (new_line_info_table): Initialize symviews_since_reset.
9774 (dwarf2out_source_line): Count symviews_since_reset and set
9775 symview_upper_bound.
9776 (dw_val_equal_p): Handle symview.
9777 (add_AT_symview): New.
9778 (print_dw_val): Handle symview.
9779 (attr_checksum, attr_checksum_ordered): Likewise.
9780 (same_dw_val_p, size_of_die): Likewise.
9781 (value_format, output_die): Likewise.
9782 (add_high_low_attributes): Use add_AT_symview for entry_view.
9783 (dwarf2out_finish): Reset symview_upper_bound, clear
9786 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9789 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9790 Add strict argument and use it.
9791 (rs6000_split_multireg_move): Update for new strict argument.
9792 (mem_operand_gpr): Disallow all non-offsettable addresses.
9793 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9795 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9798 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9799 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9800 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9803 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9804 decl, use remap_type if we want to use the type.
9806 2018-03-09 Martin Sebor <msebor@redhat.com>
9808 PR tree-optimization/84526
9809 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9811 (builtin_access::generic_overlap): Be prepared to handle non-array
9814 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9816 PR rtl-optimization/84682
9817 * lra-constraints.c (process_address_1): Check is_address flag
9818 for address constraints.
9819 (process_alt_operands): Likewise.
9820 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9821 preprocess_constraints.
9822 * recog.h (preprocess_constraints): Add oploc parameter.
9824 * recog.c (preprocess_constraints): Test address_operand for
9825 CT_ADDRESS constraints.
9827 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9830 * lra-assigns.c (assign_by_spills): Return a flag of reload
9831 assignment failure. Do not process the reload assignment
9832 failures. Do not spill other reload pseudos if they has the same
9834 (lra_assign): Add a return arg. Set up from the result of
9835 assign_by_spills call.
9836 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9837 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9838 usage_insns if it is not NULL.
9839 (spill_hard_reg_in_range): New function.
9840 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9841 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9842 function prototypes.
9843 (lra_assign): Change prototype.
9844 * lra.c (lra): Add code to deal with fails by splitting hard reg
9847 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9850 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9851 Accept complain bool parameter. Only emit errors if it is true.
9852 (arm_parse_cpu_option_name): Likewise.
9853 (arm_target_thumb_only): Adjust callers of the above.
9854 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9855 prototype to take a default true bool parameter.
9856 (arm_parse_arch_option_name): Likewise.
9858 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9859 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9863 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9864 * configure: Regenerate.
9865 * configure.ac ("linker --version-script option"): New.
9866 ("linker soname option"): New.
9868 2018-03-09 Richard Biener <rguenther@suse.de>
9870 PR tree-optimization/84775
9871 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9872 immediate uses of predicate stmts and mark them modified.
9875 PR tree-optimization/84178
9876 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9878 (version_loop_for_if_conversion): Delay update_ssa call.
9879 (tree_if_conversion): Delay update_ssa until after predicate
9882 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9885 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9886 when the function accesses prior frames.
9888 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9891 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9892 gen_llsym, otherwise call maybe_gen_llsym.
9895 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9896 has ',' character inside of it.
9898 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9901 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9902 as clobbering CC_REGNUM.
9904 2018-03-08 Richard Biener <rguenther@suse.de>
9907 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9908 (follow_copies_to_constant): Do not follow SSA names registered
9911 2018-03-08 Richard Biener <rguenther@suse.de>
9913 PR tree-optimization/84178
9914 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9916 (version_loop_for_if_conversion): Delay update_ssa call.
9917 (tree_if_conversion): Delay update_ssa until after predicate
9920 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9922 PR tree-optimization/84178
9923 * tree-if-conv.c (release_bb_predicate): Remove the
9924 the assertion that the stmts have NULL use_ops.
9925 Discard the statements, asserting that they haven't
9926 yet been added to a BB.
9928 2018-03-08 Richard Biener <rguenther@suse.de>
9930 PR tree-optimization/84746
9931 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9932 (phi_translate): Pass in destination ANTIC_OUT set.
9933 (phi_translate_1): Likewise. For a simplified result lookup
9934 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9935 (phi_translate_set): Adjust.
9936 (do_pre_regular_insertion): Likewise.
9937 (do_pre_partial_partial_insertion): Likewise.
9939 2018-03-08 Martin Liska <mliska@suse.cz>
9941 PR gcov-profile/84735
9942 * doc/gcov.texi: Document usage of profile files.
9943 * gcov-io.h: Document changes in the format.
9945 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9949 * dwarf2out.c (struct dw_line_info_table): Update comments for
9951 (FORCE_RESET_NEXT_VIEW): New.
9952 (FORCE_RESETTING_VIEW_P): New.
9953 (RESETTING_VIEW_P): Check for -1 too.
9954 (ZERO_VIEW_P): Likewise.
9955 (new_line_info_table): Force-reset next view.
9956 (dwarf2out_begin_function): Likewise.
9957 (dwarf2out_source_line): Simplify zero_view_p initialization.
9958 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9959 view directly. Omit view when omitting .loc at line 0.
9961 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9963 PR tree-optimization/84740
9964 * tree-switch-conversion.c (process_switch): Call build_constructors
9965 only if info.phi_count is non-zero.
9967 PR tree-optimization/84739
9968 * tree-tailcall.c (find_tail_calls): Check call arguments against
9969 DECL_ARGUMENTS (current_function_decl) rather than
9970 DECL_ARGUMENTS (func) when checking for tail recursion.
9972 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9974 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9975 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9976 Volker Reichelt's entry and add entries for people that perform
9977 GCC fuzzy testing and report numerous bugs.
9979 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9982 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9983 readonly data in sdata, if that is disabled.
9984 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9985 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9986 -mreadonly-in-sdata option.
9988 2018-03-07 Martin Sebor <msebor@redhat.com>
9990 PR tree-optimization/84468
9991 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9992 basic block when looking for nul assignment.
9994 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9997 * except.h (output_function_exception_table): Adjust prototype.
9998 * except.c (output_function_exception_table): Remove FNNAME parameter
9999 and add SECTION parameter. Ouput one part of the table at a time.
10000 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10001 the first part of the exception table and emit unwind directives.
10002 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10003 (i386_pe_seh_cold_init): Likewise.
10004 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10005 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10006 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10007 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10008 * config/i386/winnt.c: Include except.h.
10009 (struct seh_frame_state): Add reg_offset, after_prologue and
10010 in_cold_section fields.
10011 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10012 (i386_pe_seh_cold_init): New function.
10013 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10014 to seh->in_cold_section.
10015 (seh_emit_push): Record the offset of the push.
10016 (seh_emit_save): Record the offet of the save.
10017 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10018 Test seh->after_prologue to disregard the epilogue.
10019 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10020 (i386_pe_end_cold_function): New function.
10022 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10025 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10026 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10029 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10031 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10032 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10034 PR middle-end/84723
10035 * multiple_target.c: Include tree-inline.h and intl.h.
10036 (expand_target_clones): Diagnose and fail if node->definition and
10037 !tree_versionable_function_p (node->decl).
10039 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10041 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10043 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10044 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10045 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10047 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10050 * combine.c (try_combine): Use reg_or_subregno instead of handling
10051 just paradoxical SUBREGs and REGs.
10053 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10055 * config/arc/arc.c (arc_finalize_pic): Remove function.
10056 (arc_must_save_register): We use single base PIC register, remove
10057 checks to save/restore the PIC register.
10058 (arc_expand_prologue): Likewise.
10059 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10061 (arc_verify_short): Likewise.
10062 (arc_attr_type): Likewise.
10063 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10064 (walk_stores): Likewise.
10065 (arc_address_cost): Make it static.
10066 (arc_verify_short): Likewise.
10067 (branch_dest): Likewise.
10068 (arc_attr_type): Likewise.
10069 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10070 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10071 (arc_final_prescan_insn): Remove inserting the nops due to
10072 hardware hazards. It is done in reorg step.
10073 (insn_length_variant_t): Remove.
10074 (insn_length_parameters_t): Likewise.
10075 (arc_insn_length_parameters): Likewise.
10076 (arc_get_insn_variants): Likewise.
10077 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10079 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10081 PR inline-asm/84683
10082 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10085 PR tree-optimization/84687
10086 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10088 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10090 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10092 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10093 Rename to ppc_speculation_barrier.
10094 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10095 __builtin_ppc_speculation_barrier.
10097 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10100 * combine.c (combine_simplify_rtx): Don't try to simplify if
10101 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10104 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10106 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10107 to 32 bytes when compiling for POWER9.
10109 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10112 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10113 regparm >= 3 with no arg reg available also for calls with
10114 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10117 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10119 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10121 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10124 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10126 2018-03-05 Richard Biener <rguenther@suse.de>
10128 PR tree-optimization/84486
10129 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10130 When inserting a __builtin_assume_aligned call set the LHS
10131 SSA name alignment info accordingly.
10133 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10135 PR tree-optimization/84114
10136 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10137 Avoid reassociation of FLOAT_MODE addition.
10139 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10141 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10142 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10143 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10144 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10145 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10146 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10147 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10149 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10150 __builtin_ia32_wbinvd): New builtins.
10151 (SPECIAL_ARGS2): New.
10152 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
10153 (SPECIAL_ARGS2): New.
10154 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
10155 (ix86_valid_target_attribute_inner_p): Ditto.
10156 (ix86_init_mmx_sse_builtins): Add special_args2.
10157 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
10158 TARGET_WBNOINVD_P): New.
10159 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
10160 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
10161 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
10162 * config/i386/immintrin.h (_wbinvd): New intrinsic.
10163 * config/i386/pconfigintrin.h: New file.
10164 * config/i386/wbnoinvdintrin.h: Ditto.
10165 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
10167 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
10169 2018-03-05 Richard Biener <rguenther@suse.de>
10171 PR tree-optimization/84670
10172 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10174 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10175 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10176 (compute_antic_aux): Only assert the number of values in ANTIC_IN
10177 doesn't grow if all successors (recursively) were visited at least
10180 2018-03-05 Richard Biener <rguenther@suse.de>
10182 PR tree-optimization/84650
10183 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
10184 if executed in the loop pipeline.
10186 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
10188 * doc/configfiles.texi (Configuration Files): Move info about
10189 conditionalizing $target-protos.h to...
10190 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
10191 differs from $target-protos.h.
10193 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
10194 Chung-Ju Wu <jasonwucj@gmail.com>
10196 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
10197 * config/nds32/nds32-multiple.md (setmemsi): Define.
10198 * config/nds32/nds32-memory-manipulation.c
10199 (nds32_gen_dup_4_byte_to_word_value): New.
10200 (emit_setmem_word_loop): New.
10201 (emit_setmem_byte_loop): New.
10202 (nds32_expand_setmem_loop): New.
10203 (nds32_expand_setmem_loop_v3m): New.
10204 (nds32_expand_setmem_unroll): New.
10205 (nds32_expand_setmem): New.
10207 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10208 Chung-Ju Wu <jasonwucj@gmail.com>
10210 * config/nds32/nds32-memory-manipulation.c
10211 (nds32_emit_load_store): New.
10212 (nds32_emit_post_inc_load_store): New.
10213 (nds32_emit_mem_move): New.
10214 (nds32_emit_mem_move_block): New.
10215 (nds32_expand_movmemsi_loop_unknown_size): New.
10216 (nds32_expand_movmemsi_loop_known_size): New.
10217 (nds32_expand_movmemsi_loop): New.
10218 (nds32_expand_movmemsi_unroll): New.
10219 (nds32_expand_movmemqi): Rename ...
10220 (nds32_expand_movmemsi): ... to this.
10221 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
10222 (movmemsi): ... to this.
10223 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
10224 (nds32_expand_movmemsi): ... to this.
10226 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10227 Monk Chiang <sh.chiang04@gmail.com>
10228 Chung-Ju Wu <jasonwucj@gmail.com>
10230 * config/nds32/nds32-protos.h
10231 (nds32_expand_load_multiple): New arguments.
10232 (nds32_expand_store_multiple): Ditto.
10233 (nds32_valid_multiple_load_store): Rename ...
10234 (nds32_valid_multiple_load_store_p): ... to this.
10235 * config/nds32/nds32-memory-manipulation.c
10236 (nds32_expand_load_multiple): Refine implementation.
10237 (nds32_expand_store_multiple): Ditto.
10238 * config/nds32/nds32-multiple.md
10239 (load_multiple): Update nds32_expand_load_multiple interface.
10240 (store_multiple): Update nds32_expand_store_multiple interface.
10241 * config/nds32/nds32-predicates.c
10242 (nds32_valid_multiple_load_store): Rename ...
10243 (nds32_valid_multiple_load_store_p): ... to this and refine
10245 * config/nds32/predicates.md
10246 (nds32_load_multiple_and_update_address_operation): New predicate.
10247 (nds32_store_multiple_and_update_address_operation): New predicate.
10249 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
10250 Chung-Ju Wu <jasonwucj@gmail.com>
10252 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
10253 (combo): New attribute.
10254 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
10256 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
10258 * config/nds32/nds32.opt: Change -mcmodel= default value.
10260 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
10261 Monk Chiang <sh.chiang04@gmail.com>
10262 Chung-Ju Wu <jasonwucj@gmail.com>
10264 * config/nds32/constants.md (unspec_element): New enum.
10265 * config/nds32/constraints.md (Umw): New constraint.
10266 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
10267 * config/nds32/nds32-intrinsic.md: Likewise.
10268 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
10269 (nds32_valid_smw_lwm_base_p): New.
10270 (nds32_output_smw_single_word): New.
10271 (nds32_output_lmw_single_word): New.
10272 (nds32_expand_unaligned_load): New.
10273 (nds32_expand_unaligned_store): New.
10274 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
10275 (nds32_output_smw_single_word): Declare.
10276 (nds32_output_lmw_single_word): Declare.
10277 (nds32_expand_unaligned_load): Declare.
10278 (nds32_expand_unaligned_store): Declare.
10279 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
10280 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
10281 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
10282 NDS32_BUILTIN_UASTORE_DW.
10283 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
10286 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
10287 Kito Cheng <kito.cheng@gmail.com>
10288 Chung-Ju Wu <jasonwucj@gmail.com>
10290 * config/nds32/nds32-intrinsic.c
10291 (nds32_expand_builtin_null_ftype_reg): Delete.
10292 (nds32_expand_builtin_reg_ftype_imm): Ditto.
10293 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
10294 (nds32_read_argument): New.
10295 (nds32_legitimize_target): Ditto.
10296 (nds32_legitimize_argument): Ditto.
10297 (nds32_check_constant_argument): Ditto.
10298 (nds32_expand_unop_builtin): Ditto.
10299 (nds32_expand_unopimm_builtin): Ditto.
10300 (nds32_expand_binop_builtin): Ditto.
10301 (nds32_builtin_decl_impl): Ditto.
10302 (builtin_description): Ditto.
10303 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
10304 (nds32_init_builtins_impl): Ditto.
10305 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
10306 (nds32_builtin_decl): New.
10307 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
10308 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
10310 2018-03-02 Jeff Law <law@redhat.com>
10312 * reorg.c (stop_search_p): Handle DEBUG_INSN.
10313 (redundant_insn, fill_simple_delay_slots): Likewise.
10314 (fill_slots_from_thread): Likewise.
10315 * resource.c (mark_referenced_resources): Likewise.
10316 (mark_set_resources, find_dead_or_set_registers): Likewise.
10318 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10320 * substring-locations.h (format_warning_va): Formatting fix for
10321 ATTRIBUTE_GCC_DIAG.
10322 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
10324 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
10325 * substring-locations.c: Include intl.h.
10326 (format_warning_va): Turned into small wrapper around
10327 format_warning_n_va, renamed to ...
10328 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
10329 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
10331 (format_warning_at_substring_n): New function.
10332 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
10333 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
10334 format_warning_at_substring with just a shorter name instead of
10335 const function pointer.
10336 (fmtwarn_n): New function.
10337 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
10338 appropriate, get rid of all the fmtstr temporaries, move conditionals
10339 with G_() wrapped string literals directly into fmtwarn arguments,
10340 cast dir.len to (int), formatting fixes.
10342 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
10344 * doc/invoke.texi: Remove "Cilk Plus" references.
10346 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10347 Richard Biener <rguenther@suse.de>
10350 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
10351 for error or warning attributes if CALL_FROM_THUNK_P is set.
10354 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10357 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
10358 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
10361 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
10362 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
10364 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
10365 instead of -1U in last predictors element's probability member.
10367 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10370 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10371 arguments if they are comparable.
10373 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10375 PR tree-optimization/84634
10376 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
10377 masks and masked_loop_p with a single loop_masks, making sure it's
10378 null for bb vectorization.
10380 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
10382 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10383 (vect_analyze_data_ref_access): Use loop->safe_len rather than
10384 loop->force_vectorize to check whether there is no alias.
10386 2018-03-02 Jakub Jelinek <jakub@redhat.com>
10389 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
10391 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
10393 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
10394 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
10395 instead of a loop around prev_real_insn.
10396 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
10399 PR inline-asm/84625
10400 * config/i386/i386.c (ix86_print_operand): Use conditional
10401 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
10404 2018-03-02 Richard Biener <rguenther@suse.de>
10406 PR tree-optimization/84427
10407 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
10408 (bitmap_set_subtract_values): Rewrite to handle multiple
10411 (prune_clobbered_mems): Likewise.
10412 (phi_translate): Take edge instead of pred/phiblock.
10413 (phi_translate_1): Likewise.
10414 (phi_translate_set): Likewise. Insert all translated
10415 exprs for a value into the set, keeping possibly multiple
10416 expressions per value.
10417 (compute_antic_aux): Adjust for phi_translate changes.
10418 When intersecting union the expressions and prune those
10419 not in the final value set, keeping possibly multiple
10420 expressions per value. Do not use value-insertion
10421 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
10422 all expressions. Add verification that the value-sets
10423 only shrink during iteration.
10424 (compute_partial_antic_aux): Adjust for the phi_translate changes.
10425 (do_pre_regular_insertion): Likewise.
10426 (do_pre_partial_partial_insertion): Likewise.
10428 2018-03-02 Richard Biener <rguenther@suse.de>
10431 * config/darwin.c (saved_debug_info_level): New static global.
10432 (darwin_asm_lto_start): Disable debug info generation for LTO out.
10433 (darwin_asm_lto_end): Restore debug info generation settings.
10435 2018-03-01 Martin Liska <mliska@suse.cz>
10438 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
10439 volatile arguments.
10441 2018-03-01 Richard Biener <rguenther@suse.de>
10444 * dwarf2out.c (gen_variable_die): Properly handle late VLA
10445 type annotation with LTO when debug was disabled at compile-time.
10447 2018-03-01 Matthew Fortune <mfortune@gmail.com>
10449 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
10451 (mips_final_postscan_insn): Likewise.
10453 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
10455 PR rtl-optimization/84528
10456 * alias.c (init_alias_target): Add commentary.
10457 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
10458 a unique base value if the frame pointer is not eliminated
10459 to the stack pointer.
10461 2018-03-01 Tom de Vries <tom@codesourcery.com>
10463 PR rtl-optimization/83327
10464 * lra-int.h (hard_regs_spilled_into): Declare.
10465 * lra.c (hard_regs_spilled_into): Define.
10466 (init_reg_info): Init hard_regs_spilled_into.
10467 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
10468 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
10469 (process_bb_lives): Handle hard_regs_spilled_into.
10470 (lra_create_live_ranges_1): Before doing liveness propagation, clear
10471 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
10473 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
10475 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
10476 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
10477 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
10478 * config/rs6000/aix72.h: New file.
10480 2018-02-28 Jakub Jelinek <jakub@redhat.com>
10482 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
10483 instead of warning_at with conditional singular and plural messages
10487 * stor-layout.c (update_alignment_for_field): For
10488 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
10489 && !DECL_PACKED (field), do the alignment update, just use
10490 only desired_align instead of MAX (type_align, desired_align)
10492 (place_field): Don't do known_align < desired_align handling
10493 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
10494 is non-NULL, instead do it after rli->prev_field handling and
10495 only if not within a bitfield word. For DECL_PACKED (field)
10496 use type_align of BITS_PER_UNIT.
10498 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
10500 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
10501 superfluous parentheses and trailing spaces.
10503 2018-02-28 Richard Biener <rguenther@suse.de>
10505 PR tree-optimization/84584
10506 * graphite-scop-detection.c (scop_detection::add_scop): Discard
10507 SCoPs with fake exit edge.
10509 2018-02-28 Martin Liska <mliska@suse.cz>
10512 * timevar.c (timer::print): Fix format to properly print 100%
10515 2018-02-28 Richard Biener <rguenther@suse.de>
10517 PR middle-end/84607
10518 * genmatch.c (capture_info::walk_match): Do not mark
10519 captured expressions without operands as expr_p given
10520 they act more like predicates and should be subject to
10521 "lost tail" side-effect preserving.
10523 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
10525 PR rtl-optimization/81611
10526 * auto-inc-dec.c (attempt_change): Move dead note from
10527 mem_insn if it's the next use of regno
10528 (find_address): Take address use of reg holding
10529 non-incremented value. Add parm to limit search to the named
10531 (merge_in_block): Attempt to use a mem insn that is the next
10532 use of the original regno.
10534 2018-02-27 Martin Sebor <msebor@redhat.com>
10537 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10538 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10540 2018-02-27 Martin Sebor <msebor@redhat.com>
10542 PR translation/84207
10543 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10544 n argument to unsigned HOST_WIDE_INT.
10545 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10546 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10547 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10548 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10550 2018-02-27 Richard Biener <rguenther@suse.de>
10552 PR tree-optimization/84512
10553 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10554 Do not use the estimate returned from record_stmt_cost for
10555 the scalar iteration cost but sum properly using add_stmt_cost.
10557 2018-02-27 Richard Biener <rguenther@suse.de>
10559 PR tree-optimization/84466
10560 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10561 Adjust last change to less strictly validate use operands.
10563 2018-02-27 Martin Liska <mliska@suse.cz>
10565 PR gcov-profile/84548
10566 * gcov.c (process_file): Allow partial overlap and consider it
10567 also as group functions.
10568 (output_lines): Properly calculate range of lines for a group.
10570 2018-02-27 Martin Liska <mliska@suse.cz>
10572 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10573 'ggc' suffixes. Change first column width.
10574 (timer::print): Fix formatting of the column.
10576 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10578 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10579 preserve inline entry blocks for the sake of debug inline
10580 entry point markers alone.
10581 (remove_unused_locals): Suggest in comments a better place to
10582 force the preservation of inline entry blocks that are
10583 otherwise unused, but do not preserve them.
10585 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10587 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10589 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10592 * config/i386/constraints.md (Bs): Replace
10593 ix86_indirect_branch_register with
10594 TARGET_INDIRECT_BRANCH_REGISTER.
10596 * config/i386/i386.md (indirect_jump): Likewise.
10597 (tablejump): Likewise.
10598 (*sibcall_memory): Likewise.
10599 (*sibcall_value_memory): Likewise.
10600 Peepholes of indirect call and jump via memory: Likewise.
10601 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10602 (*sibcall_value_GOT_32): Likewise.
10603 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10604 (GOT_memory_operand): Likewise.
10605 (call_insn_operand): Likewise.
10606 (sibcall_insn_operand): Likewise.
10607 (GOT32_symbol_operand): Likewise.
10608 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10610 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10612 PR rtl-optimization/83496
10613 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10614 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10615 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10616 redundant insn, if any.
10617 (relax_delay_slots): Likewise.
10618 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10620 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10622 PR tree-optimization/83965
10623 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10624 that grouped statements are part of a reduction chain. Return
10625 true if the statement is not marked as a reduction itself but
10626 is part of a group.
10627 (vect_recog_dot_prod_pattern): Don't check whether the statement
10628 is part of a group here.
10629 (vect_recog_sad_pattern): Likewise.
10630 (vect_recog_widen_sum_pattern): Likewise.
10632 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10635 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10637 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10640 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10642 (ix86_output_indirect_function_return): New prototype.
10643 (ix86_split_simple_return_pop_internal): Likewise.
10644 * config/i386/i386.c (indirect_return_via_cx): New.
10645 (indirect_return_via_cx_bnd): Likewise.
10646 (indirect_thunk_name): Handle return va CX_REG.
10647 (output_indirect_thunk_function): Create alias for
10648 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10649 (ix86_output_indirect_jmp): Remove the bool argument.
10650 (ix86_output_indirect_function_return): New function.
10651 (ix86_split_simple_return_pop_internal): Likewise.
10652 * config/i386/i386.md (*indirect_jump): Don't pass false
10653 to ix86_output_indirect_jmp.
10654 (*tablejump_1): Likewise.
10655 (simple_return_pop_internal): Change it to define_insn_and_split.
10656 Call ix86_split_simple_return_pop_internal to split it for
10657 -mfunction-return=.
10658 (simple_return_indirect_internal): Call
10659 ix86_output_indirect_function_return instead of
10660 ix86_output_indirect_jmp.
10662 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10665 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10666 memset and value initialization afterwards.
10668 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10670 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10672 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10675 * common/config/aarch64/aarch64-common.c
10676 (aarch_option_optimization_table[]): Switch
10677 off fomit-frame-pointer
10679 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10680 Chung-Ju Wu <jasonwucj@gmail.com>
10682 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10684 (store_multiple): Ditto.
10686 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10688 * config.gcc: Add --with-cpu support for nds32 target.
10689 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10690 * config/nds32/nds32.opt: Add -mcpu= option.
10692 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10694 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10695 isel=yes): Warn for these deprecated options.
10697 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10699 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10700 ISA_2_5_MASKS_EMBEDDED.
10702 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10704 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10705 p->max as pointers rather than using iterative_hash_expr.
10707 2018-02-23 Carl Love <cel@us.ibm.com>
10709 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10710 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10712 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10713 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10714 P8V_BUILTIN_VEC_VUNSIGNED2.
10716 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10719 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10720 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10722 * lra-constraints.c (curr_insn_transform): Set up
10723 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10725 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10726 * lra-eliminations.c (spill_pseudos): Ditto.
10727 (process_insn_for_elimination): Ditto.
10728 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10729 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10730 LRA_NON_CLOBBERED_ALT.
10732 2018-02-22 Martin Sebor <msebor@redhat.com>
10734 PR tree-optimization/84480
10735 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10736 to maybe_diag_stxncpy_trunc. Call it.
10737 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10738 from gimple_fold_builtin_strcpy. Print inlining stack.
10739 (handle_builtin_stxncpy): Print inlining stack.
10740 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10742 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10745 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10746 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10747 and -fcheck-pointer-bounds are used together.
10748 (indirect_thunk_prefix): New enum.
10749 (indirect_thunk_need_prefix): New function.
10750 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10751 "_nt" instead of "_bnd" for NOTRACK prefix.
10752 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10753 (output_indirect_thunk_function): Likewise.
10755 (ix86_code_end): Update output_indirect_thunk_function calls.
10756 (ix86_output_indirect_branch_via_reg): Replace
10757 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10758 (ix86_output_indirect_branch_via_push): Likewise.
10759 (ix86_output_function_return): Likewise.
10760 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10761 incompatible with -fcf-protection=branch and
10762 -fcheck-pointer-bounds.
10764 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10767 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10768 Change gcc_assert call to output_operand_lossage.
10770 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10772 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10774 2018-02-22 DJ Delorie <dj@redhat.com>
10775 Sebastian Perta <sebastian.perta@renesas.com>
10776 Oleg Endo <olegendo@gcc.gnu.org>
10778 * config/rx/rx.c (rx_rtx_costs): New function.
10779 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10781 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10783 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10785 2018-02-22 Martin Liska <mliska@suse.cz>
10788 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10789 Add "native" as a possible value.
10791 2018-02-22 Martin Liska <mliska@suse.cz>
10794 * config/i386/i386.c (ix86_option_override_internal):
10795 Add "native" as a possible value for -march and -mtune.
10797 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10800 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10801 to all type variants.
10803 PR tree-optimization/84503
10804 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10805 width as info->bitpos + info->bitsize - start.
10806 (merged_store_group::merge_overlapping): Simplify width computation.
10807 (check_no_overlap): New function.
10808 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10809 start + width and last_order of the group, fail if check_no_overlap
10811 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10812 to group if check_no_overlap fails.
10814 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10816 * config/rs6000/altivec.md: Delete contraint arguments to
10817 define_expand, define_split, and define_peephole2, and in
10818 define_insn_and_split if always unused.
10819 * config/rs6000/darwin.md: Ditto.
10820 * config/rs6000/dfp.md: Ditto.
10821 * config/rs6000/rs6000.md: Ditto.
10822 * config/rs6000/sync.md: Ditto.
10823 * config/rs6000/vector.md: Ditto.
10824 * config/rs6000/vsx.md: Ditto.
10826 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10828 * config/rs6000/altivec.md: Write output control strings as braced
10829 blocks instead of double-quoted strings.
10830 * config/rs6000/darwin.md: Ditto.
10831 * config/rs6000/rs6000.md: Ditto.
10832 * config/rs6000/vector.md: Ditto.
10833 * config/rs6000/vsx.md: Ditto.
10835 2018-02-21 Jason Merrill <jason@redhat.com>
10837 PR c++/84314 - ICE with templates and fastcall attribute.
10838 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10840 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10842 * ipa-cp.c (determine_versionability): Fix comment typos.
10844 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10847 * ipa-cp.c (determine_versionability): Do not version functions caling
10850 2018-02-21 Martin Liska <mliska@suse.cz>
10853 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10854 Add "native" as a possible value.
10855 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10856 the macro when native cpu detection is available.
10858 2018-02-21 Martin Liska <mliska@suse.cz>
10861 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10862 Add "native" as a possible value.
10863 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10864 when native cpu detection is available.
10866 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10867 Martin Sebor <msebor@redhat.com>
10869 PR tree-optimization/84478
10870 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10872 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10873 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10874 support which is conservatively correct, for 2 only stay conservative
10875 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10876 argument to the 2 argument get_range_strlen, adjust 6 arg
10877 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10879 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10880 (gimple_fold_builtin_strlen): Pass true as last argument to
10883 2018-02-20 Martin Sebor <msebor@redhat.com>
10885 PR middle-end/84095
10886 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10887 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10888 (builtin_memref::builtin_memref): Factor out parts into
10889 set_base_and_offset and call it.
10891 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10893 PR middle-end/84406
10894 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10895 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10896 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10897 search at the associated MODE_INT.
10899 2018-02-20 Jeff Law <law@redhat.com>
10901 PR middle-end/82123
10902 PR tree-optimization/81592
10903 PR middle-end/79257
10904 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10905 for range data rather than using global data.
10906 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10907 range data rather than using global data.
10908 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10909 pass it to children as needed.
10910 (struct directive::fmtresult): Similarly.
10911 (struct directive::set_width): Similarly.
10912 (struct directive::set_precision): Similarly.
10913 (format_integer, format_directive, parse_directive): Similarly.
10914 (format_none): Accept unnamed vr_values parameter.
10915 (format_percent, format_floating, format_character): Similarly.
10916 (format_string, format_plain): Similarly.
10917 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10918 the EVRP range analyzer for range data rather than using global data.
10919 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10920 gimple-ssa-evrp-analyze.h
10921 (class sprintf_dom_walker): Add after_dom_children member function.
10922 Add evrp_range_analyzer member.
10923 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10924 range analyzer as needed.
10925 (sprintf_dom_walker::after_dom_children): New member function.
10926 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10928 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10929 (evrp_range_analyzer::pop_to_marker): Likewise.
10931 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10933 PR tree-optimization/84419
10934 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10935 with the required type if its current type is compatible but
10938 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10940 PR middle-end/82004
10941 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10942 after vectorization.
10944 2018-02-20 Martin Liska <mliska@suse.cz>
10947 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10948 possible values if we don't have a hint.
10950 2018-02-20 Martin Liska <mliska@suse.cz>
10954 * final.c (shorten_branches): Build align_tab array with one
10956 * opts.c (finish_options): Add alignment option limit check.
10957 (MAX_CODE_ALIGN): Likewise.
10958 (MAX_CODE_ALIGN_VALUE): Likewise.
10959 * doc/invoke.texi: Document maximum allowed option value for
10960 all -falign-* options.
10962 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10965 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10966 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10967 * var-tracking.c (emit_note_insn_var_location): Remove all references
10968 to NOTE_INSN_CALL_ARG_LOCATION.
10969 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10970 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10971 Use copy_rtx_if_shared.
10972 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10973 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10974 (dwarf2out_var_location): Remove handling of
10975 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10977 * final.c (final_scan_insn): Remove all references to
10978 NOTE_INSN_CALL_ARG_LOCATION.
10979 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10980 before dumping final insns.
10981 * except.c (emit_note_eh_region_end): Remove all references to
10982 NOTE_INSN_CALL_ARG_LOCATION.
10983 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10984 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10985 * config/arc/arc.c (hwloop_optimize): Likewise.
10986 * config/arm/arm.c (create_fix_barrier): Likewise.
10987 * config/s390/s390.c (s390_chunkify_start): Likewise.
10988 * config/sh/sh.c (find_barrier): Likewise.
10989 * config/i386/i386.c (rest_of_insert_endbranch,
10990 ix86_seh_fixup_eh_fallthru): Likewise.
10991 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10992 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10993 * config/frv/frv.c (frv_function_prologue): Likewise.
10994 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10996 (note_outside_basic_block_p): Remove all references to
10997 NOTE_INSN_CALL_ARG_LOCATION.
10998 * gengtype.c (adjust_field_rtx_def): Likewise.
10999 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11001 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11002 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11005 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11008 PR tree-optimization/84452
11009 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11010 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11013 2018-02-19 Martin Liska <mliska@suse.cz>
11016 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11018 2018-02-19 Martin Liska <mliska@suse.cz>
11019 Richard Sandiford <richard.sandiford@linaro.org>
11021 PR tree-optimization/82491
11022 * gimple-fold.c (get_base_constructor): Make earlier bail out
11025 2018-02-19 Carl Love <cel@us.ibm.com>
11027 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11028 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11030 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11031 P8V_BUILTIN_VEC_NEG.
11033 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11035 * config/rl78/rl78.md (movdf): New define expand.
11037 2018-02-19 Martin Liska <mliska@suse.cz>
11040 * doc/invoke.texi: Fix typo.
11041 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11043 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11045 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11046 handle rs6000_single_float and rs6000_double_float specially for
11049 2018-02-16 Jeff Law <law@redhat.com>
11051 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11052 (pop_regs): Use it.
11054 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11057 * ipa-inline.c (inline_small_functions): Fix a typo.
11059 2018-02-16 Nathan Sidwell <nathan@acm.org>
11061 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11063 2018-02-16 Carl Love <cel@us.ibm.com>
11065 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11066 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11067 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11068 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11069 expansion to P8V_BUILTIN_VEC_FLOAT2.
11071 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11073 PR rtl-optimization/70023
11074 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11075 src_regno into account.
11077 2018-02-16 Carl Love <cel@us.ibm.com>
11079 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11080 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11081 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11082 * config/rs6000/rs6000.c: Remove case statements for
11083 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11084 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11085 and P9V_BUILTIN_VEC_VINSERT4B.
11086 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11087 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11088 * config/rs6000/vsx.md:
11089 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11092 2018-02-16 Carl Love <cel@us.ibm.com>
11094 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11096 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11098 * config/rs6000/rs6000-c.c: Add the definitions for
11099 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11100 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11101 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11102 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11103 definition for insert4b and define insn *insert3b_internal.
11104 * doc/extend.texi: Add documentation for vec_extract4b.
11106 2018-02-16 Nathan Sidwell <nathan@acm.org>
11108 * doc/extend.texi (Backwards Compatibility): Mention friend
11109 injection. Note for-scope is deprecated.
11110 * doc/invoke.texi (-ffriend-injection): Deprecate.
11112 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11114 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11115 that moved to I2, also allow destinations that are a paradoxical
11116 subreg (instead of a normal reg).
11118 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11121 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11124 2018-02-16 Richard Biener <rguenther@suse.de>
11126 PR tree-optimization/84037
11127 PR tree-optimization/84016
11129 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11130 Adjust vec_construct for the fact we need additional higher latency
11131 128bit inserts for AVX256 and AVX512 vector builds.
11132 (ix86_add_stmt_cost): Scale vector construction cost for
11135 2018-02-16 Richard Biener <rguenther@suse.de>
11137 PR tree-optimization/84417
11138 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11139 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11140 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11142 2018-02-16 Martin Liska <mliska@suse.cz>
11145 * internal-fn.def (ASAN_CHECK): Set proper flags.
11146 (ASAN_MARK): Likewise.
11148 2018-02-16 Julia Koval <julia.koval@intel.com>
11150 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
11151 from PTA_CANNONLAKE.
11153 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11156 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
11157 Use ++iter rather than iter++ for std::list iterators.
11158 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
11159 defer deleting them until all nodes in the forest are processed. Do
11160 free even leaf nodes. Change to_process into auto_vec.
11163 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
11164 * vec.h (vec_default_construct): Use memset instead of placement new
11165 if BROKEN_VALUE_INITIALIZATION is defined.
11166 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
11167 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
11170 PR rtl-optimization/83723
11171 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
11172 * lra.c (lra_substitute_pseudo): Likewise. If true, use
11173 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
11175 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
11177 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
11179 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
11181 PR rtl-optimization/81443
11182 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
11183 from inner REGs to paradoxical SUBREGs.
11185 2018-02-16 Richard Biener <rguenther@suse.de>
11187 PR tree-optimization/84399
11188 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11189 For operands we can analyze at their definition make sure we can
11190 analyze them at each use as well.
11192 2018-02-16 Richard Biener <rguenther@suse.de>
11194 PR tree-optimization/84190
11195 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
11196 volatile accesses if the decl isn't volatile.
11198 2018-02-15 Jason Merrill <jason@redhat.com>
11200 PR c++/84314 - ICE with templates and fastcall attribute.
11201 * attribs.c (build_type_attribute_qual_variant): Don't clobber
11202 TYPE_CANONICAL on an existing type.
11204 2018-02-15 Jakub Jelinek <jakub@redhat.com>
11206 PR tree-optimization/84383
11207 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
11208 dstoff nor call operand_equal_p if dstbase is NULL.
11210 PR tree-optimization/84334
11211 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
11212 also a CONSTANT_CLASS_P, punt.
11214 2018-02-14 Jim Wilson <jimw@sifive.com>
11216 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
11217 first SMALL_OPERAND check. New local min_second_step. Move assert
11218 to where locals are set. Add TARGET_RVC support.
11219 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
11221 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
11223 * doc/invoke.texi: Correct -Wformat-overflow code sample.
11225 2018-02-14 Martin Sebor <msebor@redhat.com>
11227 PR tree-optimization/83698
11228 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
11229 arrays constrain the offset range to their bounds.
11230 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
11231 (builtin_access::overlap): Avoid setting the size of overlap if it's
11233 (maybe_diag_overlap): Also consider arrays when deciding what values
11234 of offsets to include in diagnostics.
11236 2018-02-14 Martin Sebor <msebor@redhat.com>
11239 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
11240 that correspond to the kind of a declaration.
11242 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
11245 * config/pa/pa.md: Load address of PIC label using the linkage table
11246 if the label is nonlocal.
11248 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
11250 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
11251 warning message if user requests -maltivec=be.
11252 * doc/invoke.texi: Document deprecation of -maltivec=be.
11254 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
11257 * config/rs6000/rs6000-c.c: Update definitions for
11258 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
11259 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
11261 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11264 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
11265 add _get_ssp intrinsics. Remove argument from
11266 __builtin_ia32_rdssp[d|q].
11267 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
11268 * config/i386/i386-builtin.def: Remove argument from
11269 __builtin_ia32_rdssp[d|q].
11270 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
11271 ix86_expand_special_args_builtin for _rdssp[d|q].
11272 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
11273 Clear register before usage.
11274 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
11275 Add documentation for new _get_ssp and _inc_ssp intrinsics.
11277 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
11279 PR tree-optimization/84357
11280 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
11281 operand 1 of an ARRAY_REF too.
11283 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
11286 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
11287 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11289 (set_of_reg): New struct.
11290 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
11291 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
11292 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
11294 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
11295 Split into bitclr, bitset, bitinvert patterns if appropriate.
11296 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
11297 use rx_fuse_in_memory_bitop.
11298 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
11299 to named insn, correct maximum insn length.
11301 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
11304 * machmode.def: Define a complex mode for PARTIAL_INT.
11305 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
11307 * doc/rtl.texi: Document CSPImode.
11308 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
11310 (msp430_hard_regno_nregs_with_padding): Likewise.
11312 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
11315 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
11317 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
11319 PR rtl-optimization/84169
11320 * combine.c (try_combine): New variable split_i2i3. Set it to true if
11321 we generated a parallel as new i3 and we split that to new i2 and i3
11322 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
11323 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
11324 those to i2, not i1. Partially rewrite this scan code.
11326 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11329 * stor-layout.c (place_field): For variable length fields, adjust
11330 offset_align afterwards not just based on the field's alignment,
11331 but also on the size.
11333 PR middle-end/84309
11334 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
11335 of exps and logs in the use_exp2 case.
11337 2018-02-13 Jeff Law <law@redhat.com>
11339 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
11340 entry for "vector".
11342 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
11345 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
11349 * common.opt (gas-loc-support, gas-locview-support): New.
11350 (ginline-points, ginternal-reset-location-views): New.
11351 * doc/invoke.texi: Document them. Use @itemx where intended.
11352 (gvariable-location-views): Adjust.
11353 * target.def (reset_location_view): New.
11354 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
11355 (TARGET_RESET_LOCATION_VIEW): New.
11356 * doc/tm.texi: Rebuilt.
11357 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
11358 (dwarf2out_default_as_locview_support): New.
11359 (output_asm_line_debug_info): Use option variables.
11360 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
11361 (output_loc_list): Likewise.
11362 (add_high_low_attributes): Check option variables.
11363 Don't output entry view attribute in strict mode.
11364 (gen_inlined_subroutine_die): Check option variables.
11365 (dwarf2out_inline_entry): Likewise.
11366 (init_sections_and_labels): Likewise.
11367 (dwarf2out_early_finish): Likewise.
11368 (maybe_reset_location_view): New, from...
11369 (dwarf2out_var_location): ... here. Call it.
11370 * debug.h (dwarf2out_default_as_loc_support): Declare.
11371 (dwarf2out_default_as_locview_support): Declare.
11372 * hooks.c (hook_int_rtx_insn_0): New.
11373 * hooks.h (hook_int_rtx_insn_0): Declare.
11374 * toplev.c (process_options): Take -gas-loc-support and
11375 -gas-locview-support from dwarf2out. Enable
11376 -gvariable-location-views by default only with locview
11377 assembler support. Enable -ginternal-reset-location-views by
11378 default only if the target defines the corresponding hook.
11379 Enable -ginline-points by default if location views are
11380 enabled; force it disabled if statement frontiers are
11382 * tree-inline.c (expand_call_inline): Check option variables.
11383 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
11385 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
11387 PR tree-optimization/84321
11388 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
11389 handling. Also check whether the anti-range contains any values
11390 that satisfy the mask; switch to a VR_RANGE if not.
11392 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
11395 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
11397 2018-02-13 Martin Jambor <mjambor@suse.cz>
11400 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
11401 of call statements, also set location of a load to a temporary.
11403 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
11405 * config/rl78/rl78.c (add_vector_labels): New function.
11406 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
11407 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
11408 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
11409 which checks that no arguments are passed.
11410 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
11411 * doc/extend.texi: Documentation for the new attribute.
11413 2018-02-13 Andreas Schwab <schwab@suse.de>
11415 * config/riscv/linux.h (CPP_SPEC): Define.
11417 2018-02-13 Jakub Jelinek <jakub@redhat.com>
11420 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
11421 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
11422 OPTION_MASK_ISA_AES as first argument to def_builtin_const
11423 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
11424 instead of OPTION_MASK_ISA_PCLMUL as first argument to
11425 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
11426 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
11427 temporarily for AES and PCLMUL builtins.
11429 PR tree-optimization/84339
11430 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
11431 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
11434 PR middle-end/84309
11435 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
11436 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
11437 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
11439 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
11441 * omp-simd-clone.h: New file.
11442 * omp-simd-clone.c: Include omp-simd-clone.h.
11443 (expand_simd_clones): No longer static.
11444 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
11445 cgraph.h and omp-simd-clone.h.
11446 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
11447 (vect_recog_widen_shift_pattern): Formatting fix.
11448 (vect_pattern_recog_1): Don't check optab for calls.
11451 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
11452 operands[2] into a REG before using gen_lowpart on it.
11454 2018-02-12 Jeff Law <law@redhat.com>
11457 * config/sh/sh.c (find_barrier): Consider a sibling call
11460 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
11461 successfully back substituting a reg.
11463 2018-02-12 Richard Biener <rguenther@suse.de>
11465 PR tree-optimization/84037
11466 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
11467 parameter, move visited init to caller.
11468 (vect_slp_analyze_operations): Separate cost from validity
11469 check, initialize visited once for all instances.
11470 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
11472 * tree-vect-stmts.c (vect_model_simple_cost): Make early
11474 (vect_model_promotion_demotion_cost): Likewise.
11475 (vectorizable_bswap): Guard cost modeling with !slp_node
11476 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
11478 (vectorizable_call): Likewise.
11479 (vectorizable_conversion): Likewise.
11480 (vectorizable_assignment): Likewise.
11481 (vectorizable_shift): Likewise.
11482 (vectorizable_operation): Likewise.
11483 (vectorizable_store): Likewise.
11484 (vectorizable_load): Likewise.
11485 (vectorizable_condition): Likewise.
11486 (vectorizable_comparison): Likewise.
11488 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
11491 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
11492 (ASAN_MARK): Fix fnspec to account for return value, change pointer
11493 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
11495 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11497 PR middle-end/83665
11498 * params.def (inline-min-speedup): Increase from 8 to 15.
11499 (max-inline-insns-auto): Decrease from 40 to 30.
11500 * ipa-split.c (consider_split): Add some buffer for function to
11501 be considered inlining candidate.
11502 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
11505 2018-02-12 Richard Biener <rguenther@suse.de>
11507 PR tree-optimization/84037
11508 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
11509 matched stmts if we cannot swap the non-matched ones.
11511 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11513 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
11514 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
11515 _mm_maskz_scalef_round_ss): New intrinsics.
11516 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
11517 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
11518 __builtin_ia32_scalefss_round): Remove.
11519 (__builtin_ia32_scalefsd_mask_round,
11520 __builtin_ia32_scalefss_mask_round): New intrinsics.
11521 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
11522 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
11523 ((match_operand:VF_128 2 "<round_nimm_predicate>"
11524 "<round_constraint>")): Changed to ...
11525 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11526 "<round_scalar_constraint>")): ... this.
11527 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11528 %0, %1, %2<round_op3>}"): Changed to ...
11529 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11530 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11531 %2<round_scalar_mask_op3>}"): ... this.
11532 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11534 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11536 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11537 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11538 (_mm_maskz_sqrt_round_ss): New intrinsics.
11539 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11540 (__builtin_ia32_sqrtsd_mask_round)
11541 (__builtin_ia32_sqrtss_mask_round): New builtins.
11542 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11543 (__builtin_ia32_sqrtss_round): Remove.
11544 (__builtin_ia32_sqrtsd_mask_round)
11545 (__builtin_ia32_sqrtss_mask_round): New builtins.
11546 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11547 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11548 ((match_operand:VF_128 1 "vector_operand"
11549 "xBm,<round_constraint>")): Changed to ...
11550 ((match_operand:VF_128 1 "vector_operand"
11551 "xBm,<round_scalar_constraint>")): ... this.
11552 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11553 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11554 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11555 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11556 %<iptr>1<round_scalar_mask_op3>}): ... this.
11557 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11558 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11560 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11563 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11564 Cast vec_cmpeq result to correct type.
11565 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11566 Cast vec_cmpgt result to correct type.
11568 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11570 * final.c (final_scan_insn_1): Renamed from...
11571 (final_scan_insn): ... this. New wrapper, to recover
11572 seen from the outermost call in recursive ones.
11573 * config/sparc/sparc.c (output_return): Drop seen from call.
11574 (output_sibcall): Likewise.
11575 * config/visium/visium.c (output_branch): Likewise.
11577 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11579 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11582 2018-02-10 Alan Modra <amodra@gmail.com>
11585 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11586 Specify LR as an input.
11588 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11591 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11592 remove_member_access_dummy_vars): New functions.
11593 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11594 lower_omp_1, execute_lower_omp): Use them.
11596 PR rtl-optimization/84308
11597 * shrink-wrap.c (spread_components): Release todo vector.
11599 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11601 PR rtl-optimization/57193
11602 * ira-color.c (struct allocno_color_data): Add member
11603 conflict_allocno_hard_prefs.
11604 (update_conflict_allocno_hard_prefs): New.
11605 (bucket_allocno_compare_func): Add a preference based on
11606 conflict_allocno_hard_prefs.
11607 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11608 (color_allocnos): Remove a dead code. Initiate
11609 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11611 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11614 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11615 constraint from =wa to wa. Avoid a subreg on the output operand,
11616 instead use a pseudo and subreg it in a move.
11617 (p9_xxbrd_<mode>): Changed to ...
11618 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11619 (p9_xxbrd_v2df): New expander.
11620 (p9_xxbrw_<mode>): Changed to ...
11621 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11622 (p9_xxbrw_v4sf): New expander.
11624 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11626 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11628 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11631 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11632 multiply in 32-bit mode.
11633 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11634 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11637 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11639 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11640 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11641 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11642 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11644 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11647 * dwarf2out.c (is_trivial_indirect_ref): New function.
11648 (dwarf2out_late_global_decl): Do not generate a location
11649 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11650 and that are not defined in the current unit.
11652 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11654 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11655 instead of a libcall for UNORDERED.
11657 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11660 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11661 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11663 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11666 * config/s390/s390.c (s390_set_current_function): Invoke
11667 s390_indirect_branch_settings also if fndecl didn't change.
11669 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11671 * config/rs6000/rs6000.md (blockage): Set length to zero.
11673 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11675 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11677 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11680 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11681 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11685 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11686 PARALLEL incoming that failed vt_get_decl_and_offset check.
11688 PR middle-end/84237
11689 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11690 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11692 (get_variable_section): For decls in named .bss* sections pass true as
11693 second argument to bss_initializer_p.
11695 2018-02-09 Marek Polacek <polacek@redhat.com>
11696 Jakub Jelinek <jakub@redhat.com>
11699 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11700 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11701 Sync some changes from cxx_fold_indirect_ref.
11703 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11705 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11707 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11708 (BLOCK_INLINE_ENTRY_LABEL): New.
11709 (dwarf2out_var_location): Disregard inline entry markers.
11710 (inline_entry_data): New struct.
11711 (inline_entry_data_hasher): New hashtable type.
11712 (inline_entry_data_hasher::hash): New.
11713 (inline_entry_data_hasher::equal): New.
11714 (inline_entry_data_table): New variable.
11715 (add_high_low_attributes): Add DW_AT_entry_pc and
11716 DW_AT_GNU_entry_view attributes if a pending entry is found
11717 in inline_entry_data_table. Add old entry_pc attribute only
11718 if debug nonbinding markers are disabled.
11719 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11720 markers are enabled.
11721 (block_within_block_p, dwarf2out_inline_entry): New.
11722 (dwarf2out_finish): Check that no entries remained in
11723 inline_entry_data_table.
11724 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11725 (final_scan_insn, notice_source_line): Likewise.
11726 (rest_of_clean_state): Skip inline entry markers.
11727 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11729 * gimple.c (gimple_build_debug_inline_entry): New.
11730 * gimple.h (enum gimple_debug_subcode): Add
11731 GIMPLE_DEBUG_INLINE_ENTRY.
11732 (gimple_build_debug_inline_entry): Declare.
11733 (gimple_debug_inline_entry_p): New.
11734 (gimple_debug_nonbind_marker_p): Adjust.
11735 * insn-notes.def (INLINE_ENTRY): New.
11736 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11737 inline entry marker notes.
11738 (print_insn): Likewise.
11739 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11740 (INSN_DEBUG_MARKER_KIND): Likewise.
11741 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11742 * tree-inline.c (expand_call_inline): Build and insert
11743 debug_inline_entry stmt.
11744 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11745 inline entry blocks early, if nonbind markers are enabled.
11746 (dump_scope_block): Dump fragment info.
11747 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11748 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11749 (gimple_build_debug_inline_entry): New.
11750 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11751 Enable/disable inline entry points too.
11752 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11753 (DEBUG_INSN): Describe inline entry markers.
11755 * common.opt (gvariable-location-views): New.
11756 (gvariable-location-views=incompat5): New.
11757 * config.in: Rebuilt.
11758 * configure: Rebuilt.
11759 * configure.ac: Test assembler for view support.
11760 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11761 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11762 * dwarf2out.c (var_loc_view): New typedef.
11763 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11764 (dwarf2out_locviews_in_attribute): New.
11765 (dwarf2out_locviews_in_loclist): New.
11766 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11767 (enum dw_line_info_opcode): Add LI_adv_address.
11768 (struct dw_line_info_table): Add view.
11769 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11770 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11771 (zero_view_p): New variable.
11772 (ZERO_VIEW_P): New macro.
11773 (output_asm_line_debug_info): New.
11774 (struct var_loc_node): Add view.
11775 (add_AT_view_list, AT_loc_list): New.
11776 (add_var_loc_to_decl): Add view param. Test it against last.
11777 (new_loc_list): Add view params. Record them.
11778 (AT_loc_list_ptr): Handle loc and view lists.
11779 (view_list_to_loc_list_val_node): New.
11780 (print_dw_val): Handle dw_val_class_view_list.
11781 (size_of_die): Likewise.
11782 (value_format): Likewise.
11783 (loc_list_has_views): New.
11784 (gen_llsym): Set vl_symbol too.
11785 (maybe_gen_llsym, skip_loc_list_entry): New.
11786 (dwarf2out_maybe_output_loclist_view_pair): New.
11787 (output_loc_list): Output view list or entries too.
11788 (output_view_list_offset): New.
11789 (output_die): Handle dw_val_class_view_list.
11790 (output_dwarf_version): New.
11791 (output_compilation_unit_header): Use it.
11792 (output_skeleton_debug_sections): Likewise.
11793 (output_rnglists, output_line_info): Likewise.
11794 (output_pubnames, output_aranges): Update version comments.
11795 (output_one_line_info_table): Output view numbers in asm comments.
11796 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11797 Call maybe_gen_llsym.
11798 (loc_list_from_tree_1): Adjust.
11799 (add_AT_location_description): Create view list attribute if
11800 needed, check it's absent otherwise.
11801 (convert_cfa_to_fb_loc_list): Adjust.
11802 (maybe_emit_file): Call output_asm_line_debug_info for test.
11803 (dwarf2out_var_location): Reset views as needed. Precompute
11804 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11805 attribute. Set view.
11806 (new_line_info_table): Reset next view.
11807 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11808 (dwarf2out_source_line): Likewise. Output view resets and labels to
11809 the assembler, or select appropriate line info opcodes.
11810 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11811 (optimize_string_length): Catch it. Adjust.
11812 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11813 dw_val_class_view_list, and remove it if no longer needed.
11814 (hash_loc_list): Hash view numbers.
11815 (loc_list_hasher::equal): Compare them.
11816 (optimize_location_lists): Check whether a view list symbol is
11817 needed, and whether the locview attribute is present, and
11818 whether they match. Remove the locview attribute if no longer
11820 (index_location_lists): Call skip_loc_list_entry for test.
11821 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11822 Use output_dwarf_version.
11823 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11824 (struct dw_val_node): Add val_view_list.
11825 * final.c (SEEN_NEXT_VIEW): New.
11826 (set_next_view_needed): New.
11827 (clear_next_view_needed): New.
11828 (maybe_output_next_view): New.
11829 (final_start_function): Rename to...
11830 (final_start_function_1): ... this. Take pointer to FIRST,
11831 add SEEN parameter. Emit param bindings in the initial view.
11832 (final_start_function): Reintroduce SEEN-less interface.
11833 (final): Rename to...
11834 (final_1): ... this. Take SEEN parameter. Output final pending
11835 next view at the end.
11836 (final): Reintroduce seen-less interface.
11837 (final_scan_insn): Output pending next view before switching
11838 sections or ending a block. Mark the next view as needed when
11839 outputting variable locations. Notify debug backend of section
11840 changes, and of location view changes.
11841 (rest_of_handle_final): Adjust.
11842 * toplev.c (process_options): Autodetect value for debug variable
11843 location views option. Warn on incompat5 without -gdwarf-5.
11844 * doc/invoke.texi (gvariable-location-views): New.
11845 (gvariable-location-views=incompat5): New.
11846 (gno-variable-location-views): New.
11848 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11850 PR tree-optimization/84136
11851 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11852 that the result of find_edge is non-NULL.
11854 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11857 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11858 storing integer register in SImode. Fix cost of 256 and 512
11859 byte aligned SSE register store.
11861 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11863 * config/i386/i386.c (ix86_multiplication_cost): Fix
11864 multiplication cost for TARGET_AVX512DQ.
11866 2018-02-08 Marek Polacek <polacek@redhat.com>
11868 PR tree-optimization/84238
11869 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11872 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11874 PR tree-optimization/84265
11875 * tree-vect-stmts.c (vectorizable_store): Don't treat
11876 VMAT_CONTIGUOUS accesses as grouped.
11877 (vectorizable_load): Likewise.
11879 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11881 PR tree-optimization/81635
11882 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11883 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11884 (test_round_for_mask): New functions.
11885 (wide_int_cc_tests): Call test_round_for_mask.
11886 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11887 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11888 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11889 range returned by get_range_info.
11891 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11894 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11895 * symtab.c: Include builtins.h
11896 (symtab_node::output_to_lto_symbol_table_p): Move here
11897 from lto-streamer-out.c:output_symbol_p.
11898 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11899 (output_symbol_p): Move all logic to symtab.c
11900 (produce_symtab): Update.
11902 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11904 * config/s390/s390-opts.h (enum indirect_branch): Define.
11905 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11906 (s390_indirect_branch_via_thunk)
11907 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11908 (enum s390_indirect_branch_type): Define.
11909 * config/s390/s390.c (struct s390_frame_layout, struct
11910 machine_function): Remove.
11911 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11912 (indirect_branch_table_label_no, indirect_branch_table_name):
11914 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11915 (enum s390_indirect_branch_option): Define.
11916 (s390_return_addr_from_memory): New function.
11917 (s390_handle_string_attribute): New function.
11918 (s390_attribute_table): Add new attribute handler.
11919 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11920 (s390_indirect_branch_via_thunk): New function.
11921 (s390_indirect_branch_via_inline_thunk): New function.
11922 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11923 sibling call optimization for non z10 compiles.
11924 (s390_emit_call): Force indirect branch target to be a single
11925 register. Add r1 clobber for non-z10 compiles.
11926 (s390_emit_epilogue): Emit return jump via return_use expander.
11927 (s390_reorg): Handle JUMP_INSNs as execute targets.
11928 (s390_option_override_internal): Perform validity checks for the
11929 new command line options.
11930 (s390_indirect_branch_attrvalue): New function.
11931 (s390_indirect_branch_settings): New function.
11932 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11933 (s390_output_indirect_thunk_function): New function.
11934 (s390_code_end): Implement target hook.
11935 (s390_case_values_threshold): Implement target hook.
11936 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11938 * config/s390/s390.h (struct s390_frame_layout)
11939 (struct machine_function): Move here from s390.c.
11940 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11941 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11942 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11943 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11944 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11945 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11946 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11947 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11948 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11949 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11950 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11951 (mnemonic attribute): Add values which aren't recognized
11953 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11954 pattern for branch conversion. Fix mnemonic attribute.
11955 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11956 indirect branch via thunk if requested.
11957 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11958 ("*indirect_jump"): Disable for branch conversion using out of
11960 ("indirect_jump_via_thunk<mode>_z10")
11961 ("indirect_jump_via_thunk<mode>")
11962 ("indirect_jump_via_inlinethunk<mode>_z10")
11963 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11964 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11965 ("casesi_jump_via_inlinethunk<mode>_z10")
11966 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11967 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11968 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11969 ("*indirect2_jump"): Disable for branch conversion.
11970 ("casesi_jump"): Turn into expander and expand patterns for branch
11972 ("return_use"): New expander.
11973 ("*return"): Emit return via thunk and rename it to ...
11974 ("*return<mode>"): ... this one.
11975 * config/s390/s390.opt: Add new options and and enum for the
11978 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11980 * lra-constraints.c (match_reload): Unconditionally use
11981 gen_lowpart_SUBREG, rather than selecting between that
11982 and equivalent gen_rtx_SUBREG code.
11984 2018-02-08 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/84233
11987 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11988 changed flag instead of boguously re-using phi_inserted.
11990 2018-02-08 Martin Jambor <mjambor@suse.cz>
11992 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11993 static local variables.
11995 2018-02-08 Richard Biener <rguenther@suse.de>
11997 PR tree-optimization/84278
11998 * tree-vect-stmts.c (vectorizable_store): When looking for
11999 smaller vector types to perform grouped strided loads/stores
12000 make sure the mode is supported by the target.
12001 (vectorizable_load): Likewise.
12003 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12005 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12006 Increase LDP/STP opportunities by adding adjacent callee-saves.
12008 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12010 PR rtl-optimization/84068
12011 PR rtl-optimization/83459
12012 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12014 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12016 PR tree-optimization/84224
12017 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12018 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12019 non-zero arguments.
12021 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12024 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12025 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12026 count, remove one USE.
12028 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12030 * doc/install.texi (Configuration): Document the
12031 --with-long-double-format={ibm,ieee} PowerPC configuration
12035 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12036 Convert from define_expand to be define_insn_and_split. Rework
12037 float/double/_Float128 conversions to QI/HI/SImode to work with
12038 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12039 conversions to QI/HImode types did a store and then a load to
12040 truncate the value. For conversions to VSX registers, don't split
12041 the insn, instead emit the code directly. Use the code iterator
12042 any_fix to combine signed and unsigned conversions.
12043 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12044 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12045 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12046 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12047 (fix_<mode>di2_hw): Likewise.
12048 (fixuns_<mode>di2_hw): Likewise.
12049 (fix_<mode>si2_hw): Likewise.
12050 (fixuns_<mode>si2_hw): Likewise.
12051 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12052 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12053 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12054 fix<uns>_trunc<SFDF:mode>si2_p8.
12055 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12056 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12057 (fix<uns>_<mode>_mem): Likewise.
12058 (fctiw<u>z_<mode>_mem): Likewise.
12059 (fix<uns>_<mode>_mem): Likewise.
12060 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12061 the register allocator from doing a direct move to the GPRs to do
12062 a store, and instead use the ISA 3.0 store byte/half-word from
12063 vector register instruction. For IEEE 128-bit floating point,
12064 also optimize stores of 32-bit ints.
12065 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12067 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12069 * genextract.c (push_pathstr_operand): New function to support
12071 (walk_rtx): Call push_pathstr_operand.
12072 (print_path): Support [a-zA-Z].
12074 2018-02-07 Richard Biener <rguenther@suse.de>
12076 PR tree-optimization/84037
12077 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12078 (cse_and_gimplify_to_preheader): Declare.
12079 (vect_get_place_in_interleaving_chain): Likewise.
12080 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12082 (_loop_vec_info::~_loop_vec_info): Delete it.
12083 (cse_and_gimplify_to_preheader): New function.
12084 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12085 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12086 (vectorizable_load): Likewise. For grouped stores always base
12087 the IV on the first element.
12088 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12089 condition before gimplifying.
12091 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12093 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12094 *DIV_EXPR and *MOD_EXPR.
12096 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12099 * config/i386/i386.c (ix86_option_override_internal): Mask out
12100 the CF_SET bit when checking -fcf-protection.
12102 2018-02-07 Tom de Vries <tom@codesourcery.com>
12105 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12108 2018-02-07 Richard Biener <rguenther@suse.de>
12110 PR tree-optimization/84204
12111 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12114 PR tree-optimization/84205
12115 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12116 special-case isl_ast_op_zdiv_r.
12118 PR tree-optimization/84223
12119 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12120 Only add conditions from within the region.
12121 (gather_bbs::after_dom_children): Adjust.
12123 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12126 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12127 * config/avr/avr.md: Only post-reload split REG-REG moves if
12128 either register is GENERAL_REG_P.
12130 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12132 PR tree-optimization/84235
12133 * tree-ssa-scopedtables.c
12134 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12135 if the subtraction is performed in floating point type where NaNs are
12136 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12137 build 1. Formatting fix.
12139 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12142 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12143 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12144 and skip it regardless of bb boundaries. Use CALL_P macro,
12145 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12146 unnecessarily, formatting fix.
12148 2018-02-06 Michael Collison <michael.collison@arm.com>
12150 * config/arm/thumb2.md:
12151 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
12152 (*thumb_mov_notscc): Ditto.
12154 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
12157 * config/rs6000/rs6000.md (su code attribute): Use "u" for
12158 unsigned_fix, not "s".
12160 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12162 * configure.ac (gcc_fn_eh_frame_ro): New function.
12163 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
12164 correct .eh_frame permissions.
12165 * configure: Regenerate.
12167 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
12169 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
12170 irrelevant options.
12172 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12174 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12175 Display warning message for -mno-speculate-indirect-jumps.
12177 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
12179 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
12181 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
12183 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
12185 PR tree-optimization/84225
12186 * tree-eh.c (find_trapping_overflow): Only call
12187 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
12189 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12192 * config/i386/i386.c: Reimplement the check of possible options
12193 -mibt/-mshstk conbination. Change error messages.
12194 * doc/invoke.texi: Fix a typo: remove extra '='.
12196 2018-02-06 Marek Polacek <polacek@redhat.com>
12198 PR tree-optimization/84228
12199 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
12201 2018-02-06 Tamar Christina <tamar.christina@arm.com>
12204 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
12205 emitted arch directives.
12206 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
12207 __ARM_FEATURE_COPROC before changing architectures.
12209 2018-02-06 Richard Biener <rguenther@suse.de>
12211 * config/i386/i386.c (print_reg): Fix typo.
12212 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
12214 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
12216 * configure: Regenerate.
12218 2018-02-05 Martin Sebor <msebor@redhat.com>
12220 PR tree-optimization/83369
12221 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
12224 2018-02-05 Martin Liska <mliska@suse.cz>
12226 * doc/invoke.texi: Cherry-pick upstream r323995.
12228 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
12230 * ira.c (ira_init_register_move_cost): Adjust comment.
12232 2018-02-05 Martin Liska <mliska@suse.cz>
12234 PR gcov-profile/84137
12235 * doc/gcov.texi: Fix typo in documentation.
12237 2018-02-05 Martin Liska <mliska@suse.cz>
12239 PR gcov-profile/83879
12240 * doc/gcov.texi: Document necessity of --dynamic-list-data when
12241 using dlopen functionality.
12243 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
12245 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
12246 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
12247 _mm_maskz_range_ss, _mm_mask_range_round_ss,
12248 _mm_maskz_range_round_ss): New intrinsics.
12249 (__builtin_ia32_rangesd128_round)
12250 (__builtin_ia32_rangess128_round): Remove.
12251 (__builtin_ia32_rangesd128_mask_round,
12252 __builtin_ia32_rangess128_mask_round): New builtins.
12253 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
12254 __builtin_ia32_rangess128_round): Remove.
12255 (__builtin_ia32_rangesd128_mask_round,
12256 __builtin_ia32_rangess128_mask_round): New builtins.
12257 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
12258 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
12259 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
12260 "<round_saeonly_constraint>")): Changed to ...
12261 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
12262 "<round_saeonly_scalar_constraint>")): ... this.
12263 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
12264 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
12265 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
12266 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
12267 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
12269 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
12271 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
12273 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
12274 Remove all values except native, 8540 and 8548.
12276 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
12278 * config/i386/i386.c (ix86_output_function_return): Pass
12279 INVALID_REGNUM, instead of -1, as invalid register number to
12280 indirect_thunk_name and output_indirect_thunk.
12282 2018-02-02 Julia Koval <julia.koval@intel.com>
12284 * config.gcc: Add -march=icelake.
12285 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
12286 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
12287 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
12288 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
12289 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
12290 (processor_target_table): Add icelake.
12291 (ix86_option_override_internal): Handle new PTAs.
12292 (get_builtin_code_for_version): Handle icelake.
12293 (M_INTEL_COREI7_ICELAKE): New.
12294 (fold_builtin_cpu): Handle icelake.
12295 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
12296 * doc/invoke.texi: Add -march=icelake.
12298 2018-02-02 Julia Koval <julia.koval@intel.com>
12300 * config/i386/i386.c (ix86_option_override_internal): Change flags type
12301 to wide_int_bitmask.
12302 * wide-int-bitmask.h: New.
12304 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12307 * config/i386/i386.md: Replace Pmode with word_mode in
12308 builtin_setjmp_setup and builtin_longjmp to support x32.
12310 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
12314 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
12316 (rs6000_supported_cpu_names): New static variable.
12317 (linux_cpu_translation_table): Likewise.
12318 (elf_platform) <cpu>: Define new static variable and use it.
12319 Translate kernel AT_PLATFORM name to canonical name if needed.
12320 Error if platform name is unknown.
12322 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
12325 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
12327 2018-02-01 Jeff Law <law@redhat.com>
12330 * config/i386/i386.c (release_scratch_register_on_entry): Add new
12331 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
12332 the scratch if RELEASE_VIA_POP is false.
12333 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
12334 If we have to save a temporary register, decrement SIZE appropriately.
12335 Pass new arguments to release_scratch_register_on_entry.
12336 (ix86_adjust_stack_and_probe): Likewise.
12337 (ix86_emit_probe_stack_range): Pass new arguments to
12338 release_scratch_register_on_entry.
12340 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
12342 PR rtl-optimization/84157
12343 * combine.c (change_zero_ext): Use REG_P predicate in
12344 front of HARD_REGISTER_P predicate.
12346 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
12348 * config/avr/avr.c (avr_option_override): Move disabling of
12349 -fdelete-null-pointer-checks to...
12350 * common/config/avr/avr-common.c (avr_option_optimization_table):
12353 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12355 PR tree-optimization/81635
12356 * tree-data-ref.c (split_constant_offset_1): For types that
12357 wrap on overflow, try to use range info to prove that wrapping
12360 2018-02-01 Renlin Li <renlin.li@arm.com>
12363 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
12364 TAILCALL_ADDR_REGS.
12365 (aarch64_register_move_cost): Likewise.
12366 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
12367 TAILCALL_ADDR_REGS.
12368 (REG_CLASS_NAMES): Likewise.
12369 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
12370 TAILCALL_ADDR_REGS. Remove IP registers.
12371 * config/aarch64/aarch64.md (Ucs): Update register constraint.
12373 2018-02-01 Richard Biener <rguenther@suse.de>
12375 * domwalk.h (dom_walker::dom_walker): Add additional constructor
12376 for specifying RPO order and allow NULL for that.
12377 * domwalk.c (dom_walker::dom_walker): Likewise.
12378 (dom_walker::walk): Handle NULL RPO order.
12379 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
12381 (rewrite_update_dom_walker): Likewise.
12382 (mark_def_dom_walker): Likewise.
12384 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12386 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
12387 (aarch64_maybe_expand_sve_subreg_move): Declare.
12388 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
12389 * config/aarch64/predicates.md (aarch64_any_register_operand): New
12391 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
12392 that are semantically a reverse operation.
12393 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
12394 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
12395 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
12397 (aarch64_can_change_mode_class): For big-endian, forbid changes
12398 between two SVE modes if they have different element sizes.
12400 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12402 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
12403 the TImode handling for big-endian targets.
12405 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12407 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
12408 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
12410 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
12411 Remove BSWAP handing for big-endian targets and use the form of
12412 LD1RQ appropariate for the mode.
12414 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12416 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
12417 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
12418 duplicated element.
12420 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
12423 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
12424 check for operands that need to go through aarch64_sve_reload_be.
12426 2018-02-01 Jakub Jelinek <jakub@redhat.com>
12428 PR tree-optimization/81661
12429 PR tree-optimization/84117
12430 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
12431 * tree-eh.c: Include gimplify.h.
12432 (find_trapping_overflow, replace_trapping_overflow,
12433 rewrite_to_non_trapping_overflow): New functions.
12434 * tree-vect-loop.c: Include tree-eh.h.
12435 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
12436 * tree-data-ref.c: Include tree-eh.h.
12437 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
12439 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
12441 PR rtl-optimization/84123
12442 * combine.c (change_zero_ext): Check if hard register satisfies
12443 can_change_dest_mode before calling gen_lowpart_SUBREG.
12445 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
12448 * ira.c (ira_init_register_move_cost): Remove assert.
12450 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12452 PR rtl-optimization/84071
12453 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
12454 * doc/tm.texi: Regenerate.
12456 2018-01-31 Richard Biener <rguenther@suse.de>
12458 PR tree-optimization/84132
12459 * tree-data-ref.c (analyze_miv_subscript): Properly
12460 check whether evolution_function_is_affine_multivariate_p
12461 before calling gcd_of_steps_may_divide_p.
12463 2018-01-31 Julia Koval <julia.koval@intel.com>
12466 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
12467 * config/i386/i386.md (rdpid_rex64) New.
12468 (rdpid): Make 32bit only.
12470 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
12473 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
12474 an IDENTIFIER_NODE for FUNCTION_TYPE's.
12476 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12479 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12481 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12483 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
12485 PR rtl-optimization/84071
12486 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
12487 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
12489 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12491 * config/arc/arc.c (arc_handle_aux_attribute): New function.
12492 (arc_attribute_table): Add 'aux' attribute.
12493 (arc_in_small_data_p): Consider aux like variables.
12494 (arc_is_aux_reg_p): New function.
12495 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
12496 (arc_get_aux_arg): New function.
12497 (prepare_move_operands): Handle aux-register access.
12498 (arc_handle_aux_attribute): New function.
12499 * doc/extend.texi (ARC Variable attributes): Add subsection.
12501 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
12503 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
12504 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
12505 (arc_attribute_table): Add 'uncached' attribute.
12506 (arc_print_operand): Print '.di' flag for uncached memory
12508 (arc_in_small_data_p): Do not consider for small data the uncached
12510 (arc_is_uncached_mem_p): New function.
12511 * config/arc/predicates.md (compact_store_memory_operand): Check
12512 for uncached memory accesses.
12513 (nonvol_nonimm_operand): Likewise.
12514 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
12516 2018-01-31 Jakub Jelinek <jakub@redhat.com>
12519 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
12520 falign-loops=): Add Optimization flag.
12522 2018-01-30 Jeff Law <law@redhat.com>
12525 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12526 INT_REGISTERS_SAVED. Check it prior to calling
12527 get_scratch_register_on_entry.
12528 (ix86_adjust_stack_and_probe): Similarly.
12529 (ix86_emit_probe_stack_range): Similarly.
12530 (ix86_expand_prologue): Corresponding changes.
12532 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12535 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12536 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12538 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12541 * lra-constraints.c (curr_insn_transform): Process AND in the
12544 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12546 PR rtl-optimization/83986
12547 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12548 dependence against last_pending_memory_flush in addition to
12549 pending_jump_insns.
12551 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12553 PR tree-optimization/81611
12554 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12557 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12560 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12563 2018-01-30 Richard Biener <rguenther@suse.de>
12564 Jakub Jelinek <jakub@redhat.com>
12566 PR tree-optimization/84111
12567 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12568 inner loops added during recursion, as they don't have up-to-date
12571 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12574 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12575 (can_inline_edge_by_limits_p): ... here.
12576 (can_early_inline_edge_p, check_callers,
12577 update_caller_keys, update_callee_keys, recursive_inlining,
12578 add_new_edges_to_heap, speculation_useful_p,
12579 inline_small_functions,
12580 inline_small_functions, flatten_function,
12581 inline_to_all_callers_1): Update.
12583 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12585 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12588 2018-01-30 Richard Biener <rguenther@suse.de>
12590 PR tree-optimization/83008
12591 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12592 invariant and constant vector uses in stmts when they need
12593 more than one stmt.
12595 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12598 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12599 * configure: Regenerate.
12601 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12603 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12605 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12606 Use gen_rtx_REG rather than gen_lowpart.
12608 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12610 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12611 rather than 0 when creating partial subregs.
12613 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12615 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12618 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12621 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12622 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12623 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12624 flags. This restores the settings used before the 2017-07-24.
12625 Turning off pre increment/decrement/modify allows IVOPTS to
12626 optimize DF/SF loops where the index is an int.
12628 2018-01-29 Richard Biener <rguenther@suse.de>
12629 Kelvin Nilsen <kelvin@gcc.gnu.org>
12632 * tree-vect-stmts.c (vectorizable_call): Don't call
12633 targetm.vectorize_builtin_md_vectorized_function if callee is
12636 2018-01-22 Carl Love <cel@us.ibm.com>
12638 * doc/extend.tex: Fix typo in second arg in
12639 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12641 2018-01-29 Richard Biener <rguenther@suse.de>
12643 PR tree-optimization/84086
12644 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12645 (flush_ssaname_freelist): When SSA names were released reset
12646 the SCEV hash table.
12648 2018-01-29 Richard Biener <rguenther@suse.de>
12650 PR tree-optimization/84057
12651 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12652 removed paths when removing edges.
12654 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12656 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12657 -mfunction-return=@var{choice}.
12659 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12661 PR diagnostic/84034
12662 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12663 Handle CR like TAB.
12664 (layout::print_source_line): Likewise.
12665 (test_get_line_width_without_trailing_whitespace): Add test cases.
12667 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12669 PR middle-end/84040
12670 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12673 2018-01-26 Jim Wilson <jimw@sifive.com>
12675 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12677 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12680 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12682 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12683 and CMP + SUB-immediate -> SUBS.
12685 2018-01-26 Martin Sebor <msebor@redhat.com>
12687 PR tree-optimization/83896
12688 * tree-ssa-strlen.c (get_string_len): Rename...
12689 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12690 Avoid assuming length is constant.
12691 (handle_char_store): Use HOST_WIDE_INT for string length.
12693 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12696 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12697 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12699 2018-01-26 Richard Biener <rguenther@suse.de>
12701 PR rtl-optimization/84003
12702 * dse.c (record_store): Only record redundant stores when
12703 the earlier store aliases at least all accesses the later one does.
12705 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12707 PR rtl-optimization/83985
12708 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12709 REG_CFA_RESTORE insns.
12710 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12711 REG_CFA_RESTORE insns here.
12714 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12715 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12717 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12719 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12720 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12721 (arc_init): Likewise.
12722 (arc_override_options): Likewise.
12723 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12725 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12727 * config/arc/arc.h (TARGET_DBNZ): Define.
12728 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12729 properly set the tune attribute.
12730 (dbnz): Use TARGET_DBNZ guard.
12731 * config/arc/arc.opt (mtune): Add core3 option.
12733 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12735 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12736 recognize new pic like addresses.
12737 (arc_delegitimize_address): Clean up.
12739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12741 * config/arc/arc-arches.def: Option mrf16 valid for all
12743 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12744 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12745 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12746 * config/arc/arc-tables.opt: Regenerate.
12747 * config/arc/arc.c (arc_conditional_register_usage): Handle
12748 reduced register file case.
12749 (arc_file_start): Set must have build attributes.
12750 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12751 mrf16 option value.
12752 * config/arc/arc.opt (mrf16): Add new option.
12753 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12754 * config/arc/genmultilib.awk: Handle new mrf16 option.
12755 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12756 * config/arc/t-multilib: Regenerate.
12757 * doc/invoke.texi (ARC Options): Document mrf16 option.
12759 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12761 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12762 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12763 (arc_attribute_table): Add 'secure_call' attribute.
12764 (arc_print_operand): Print secure call operand.
12765 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12767 (arc_is_secure_call_p): New function. * config/arc/arc.md
12768 (call_i): Add support for sjli instruction.
12769 (call_value_i): Likewise.
12770 * config/arc/constraints.md (Csc): New constraint.
12772 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12773 John Eric Martin <John.Martin@emmicro-us.com>
12775 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12776 * config/arc/arc.c (_arc_jli_section): New struct.
12777 (arc_jli_section): New type.
12778 (rc_jli_sections): New static variable.
12779 (arc_handle_jli_attribute): New function.
12780 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12781 (arc_file_end): New function.
12782 (TARGET_ASM_FILE_END): Define.
12783 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12784 (arc_add_jli_section): New function.
12785 (jli_call_scan): Likewise.
12786 (arc_reorg): Call jli_call_scan.
12787 (arc_output_addsi): Remove 'S' from printing asm operand.
12788 (arc_is_jli_call_p): New function.
12789 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12791 (movhi_insn): Likewise.
12792 (movsi_insn): Likewise.
12793 (movsi_set_cc_insn): Likewise.
12794 (loadqi_update): Likewise.
12795 (load_zeroextendqisi_update): Likewise.
12796 (load_signextendqisi_update): Likewise.
12797 (loadhi_update): Likewise.
12798 (load_zeroextendhisi_update): Likewise.
12799 (load_signextendhisi_update): Likewise.
12800 (loadsi_update): Likewise.
12801 (loadsf_update): Likewise.
12802 (movsicc_insn): Likewise.
12803 (bset_insn): Likewise.
12804 (bxor_insn): Likewise.
12805 (bclr_insn): Likewise.
12806 (bmsk_insn): Likewise.
12807 (bicsi3_insn): Likewise.
12808 (cmpsi_cc_c_insn): Likewise.
12809 (movsi_ne): Likewise.
12810 (movsi_cond_exec): Likewise.
12811 (clrsbsi2): Likewise.
12812 (norm_f): Likewise.
12822 (call_i): Remove 'S' asm letter, add jli instruction.
12823 (call_value_i): Likewise.
12824 * config/arc/arc.op (mjli-always): New option.
12825 * config/arc/constraints.md (Cji): New constraint.
12826 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12828 (subsf3_fpx): Likewise.
12829 (mulsf3_fpx): Likewise.
12830 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12832 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12833 function attrbutes.
12834 * doc/invoke.texi (ARC): Document mjli-always option.
12836 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12838 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12839 avoid addition with 0 and use incw and decw where possible.
12841 2018-01-26 Richard Biener <rguenther@suse.de>
12843 PR tree-optimization/81082
12844 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12845 association if it requires casting to unsigned.
12846 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12847 from fold_plusminus_mult_expr to catch important cases late when
12848 range info is available.
12850 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12852 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12853 * configure.ac (hidden_linkonce): New test.
12854 * configure: Regenerate.
12855 * config.in: Regenerate.
12857 2018-01-26 Julia Koval <julia.koval@intel.com>
12859 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12860 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12861 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12862 _mm_mask_bitshuffle_epi64_mask): Fix type.
12863 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12864 USI_FTYPE_V4DI_V4DI_USI): Remove.
12865 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12866 __builtin_ia32_vpshufbitqmb256_mask,
12867 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12868 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12869 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12871 2018-01-26 Alan Modra <amodra@gmail.com>
12874 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12875 UNSPEC_VBPERMQ. Sort other unspecs.
12877 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12879 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12881 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12883 PR middle-end/83055
12884 * predict.c (drop_profile): Do not push/pop cfun; update also
12886 (handle_missing_profiles): Fix logic looking for zero profiles.
12888 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12890 PR middle-end/83977
12891 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12892 on functions with #pragma omp declare simd or functions with simd
12894 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12895 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12896 Remove trailing \n from warning_at calls.
12898 2018-01-25 Tom de Vries <tom@codesourcery.com>
12901 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12902 for neutered workers.
12904 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12907 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12908 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12910 2018-01-24 Jeff Law <law@redhat.com>
12913 * i386.c (get_probe_interval): Move to earlier point.
12914 (ix86_compute_frame_layout): If -fstack-clash-protection and
12915 the frame is larger than the probe interval, then use pushes
12916 to save registers rather than reg->mem moves.
12917 (ix86_expand_prologue): Remove conditional for int_registers_saved
12920 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12923 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12924 min/max for never referenced object.
12926 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12928 PR middle-end/83977
12929 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12931 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12932 attributes from DECL_ATTRIBUTES (decl) without affecting
12933 DECL_ATTRIBUTES (current_function_decl).
12934 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12935 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12937 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12939 PR tree-optimization/83979
12940 * fold-const.c (fold_comparison): Use constant_boolean_node
12941 instead of boolean_{true,false}_node.
12943 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12945 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12948 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12950 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12951 Simplify the clause that sets the length attribute.
12952 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12953 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12954 clause that sets the length attribute.
12955 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12957 2018-01-24 Tom de Vries <tom@codesourcery.com>
12960 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12961 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12962 Add strict parameter.
12963 (prevent_branch_around_nothing): Insert dummy insn between branch to
12964 label and label with no ptx insn inbetween.
12965 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12967 2018-01-24 Tom de Vries <tom@codesourcery.com>
12970 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12971 for neutered threads in warp.
12972 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12974 2018-01-24 Richard Biener <rguenther@suse.de>
12976 PR tree-optimization/83176
12977 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12980 2018-01-24 Richard Biener <rguenther@suse.de>
12982 PR tree-optimization/82819
12983 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12984 code generating pluses that are no-ops in the target precision.
12986 2018-01-24 Richard Biener <rguenther@suse.de>
12988 PR middle-end/84000
12989 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12991 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12993 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12994 to merge probabilities.
12995 * predict.c (probably_never_executed): Also mark as cold functions
12996 with global 0 profile and guessed local profile.
12997 * profile-count.c (profile_probability::combine_with_count): New
12999 * profile-count.h (profile_probability::operator*,
13000 profile_probability::operator*=, profile_probability::operator/,
13001 profile_probability::operator/=): Reduce precision to adjusted
13002 and set value to guessed on contradictory divisions.
13003 (profile_probability::combine_with_freq): Remove.
13004 (profile_probability::combine_wiht_count): Declare.
13005 (profile_count::force_nonzero):: Set to adjusted.
13006 (profile_count::probability_in):: Set quality to adjusted.
13007 * tree-ssa-tail-merge.c (replace_block_by): Use
13008 combine_with_count.
13010 2018-01-23 Andrew Waterman <andrew@sifive.com>
13011 Jim Wilson <jimw@sifive.com>
13013 * config/riscv/riscv.c (riscv_stack_boundary): New.
13014 (riscv_option_override): Set riscv_stack_boundary. Handle
13015 riscv_preferred_stack_boundary_arg.
13016 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13017 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13018 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13019 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13020 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13021 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13023 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13026 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13027 of struct ix86_frame.
13028 (ix86_expand_epilogue): Likewise. Add a local variable for
13029 the reg_save_offset field in struct ix86_frame.
13031 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13033 PR tree-optimization/82604
13034 * tree-loop-distribution.c (enum partition_kind): New enum item
13035 PKIND_PARTIAL_MEMSET.
13036 (partition_builtin_p): Support above new enum item.
13037 (generate_code_for_partition): Ditto.
13038 (compute_access_range): Differentiate cases that equality can be
13039 proven at all loops, the innermost loops or no loops.
13040 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13041 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13042 (finalize_partitions, distribute_loop): Don't fuse partition of
13043 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13044 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13045 parloop is enabled.
13047 2018-01-23 Martin Liska <mliska@suse.cz>
13049 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13050 order to ignore the predictor.
13051 (PRED_POLYMORPHIC_CALL): Likewise.
13052 (PRED_RECURSIVE_CALL): Likewise.
13054 2018-01-23 Martin Liska <mliska@suse.cz>
13056 * tree-profile.c (tree_profiling): Print function header to
13057 aware reader which function we are working on.
13058 * value-prof.c (gimple_find_values_to_profile): Do not print
13059 not interesting value histograms.
13061 2018-01-23 Martin Liska <mliska@suse.cz>
13063 * profile-count.h (enum profile_quality): Add
13064 profile_uninitialized as the first value. Do not number values
13065 as they are zero based.
13066 (profile_count::verify): Update sanity check.
13067 (profile_probability::verify): Likewise.
13069 2018-01-23 Nathan Sidwell <nathan@acm.org>
13071 * doc/invoke.texi (ffor-scope): Deprecate.
13073 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13075 PR tree-optimization/83510
13076 * domwalk.c (set_all_edges_as_executable): New function.
13077 (dom_walker::dom_walker): Convert bool param
13078 "skip_unreachable_blocks" to enum reachability. Move setup of
13079 edge flags to set_all_edges_as_executable and only do it when
13080 reachability is REACHABLE_BLOCKS.
13081 * domwalk.h (enum dom_walker::reachability): New enum.
13082 (dom_walker::dom_walker): Convert bool param
13083 "skip_unreachable_blocks" to enum reachability.
13084 (set_all_edges_as_executable): New decl.
13085 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13086 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13088 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13089 but converting true to REACHABLE_BLOCKS.
13090 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13092 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13093 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13094 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13096 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13097 if check_all_array_refs will be called.
13099 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13101 * tree.c (selftest::test_location_wrappers): Add more test
13104 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13106 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13107 (selftest::test_bit_in_range): Likewise.
13109 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13112 * doc/sourcebuild.texi (vect_float): Say that the selector
13113 only describes the situation when -funsafe-math-optimizations is on.
13114 (vect_float_strict): Document.
13116 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13118 PR tree-optimization/83965
13119 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13120 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13121 instead of checking only for a reduction.
13122 (vect_recog_widen_sum_pattern): Likewise.
13124 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13126 * predict.c (probably_never_executed): Only use precise profile info.
13127 (compute_function_frequency): Skip after inlining hack since we now
13128 have quality checking.
13130 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13132 * profile-count.h (profile_probability::very_unlikely,
13133 profile_probability::unlikely, profile_probability::even): Set
13134 precision to guessed.
13136 2018-01-23 Richard Biener <rguenther@suse.de>
13138 PR tree-optimization/83963
13139 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13140 Properly terminate dominator walk when crossing the exit edge not
13141 when visiting its source block.
13143 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13146 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13147 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13149 2018-01-22 Jakub Jelinek <jakub@redhat.com>
13151 PR tree-optimization/83957
13152 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
13153 semicolon after for body surrounded by braces.
13155 PR tree-optimization/83081
13156 * profile-count.h (profile_probability::split): New method.
13157 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
13158 Use profile_probability::split.
13159 (do_compare_rtx_and_jump): Fix adjustment of probabilities
13160 when splitting a single conditional jump into 2.
13162 2018-01-22 David Malcolm <dmalcolm@redhat.com>
13164 PR tree-optimization/69452
13165 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
13168 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13170 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
13171 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
13172 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
13174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13176 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
13178 * config/rl78/rl78.md (movdi): New define_expand.
13179 * config/rl78/rl78.c (rl78_split_movdi): New function.
13181 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
13184 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
13186 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
13187 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
13188 128-bit to produce an UNSPEC move to get the double word with the
13189 signbit and then a shift directly to do signbit.
13190 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
13191 implementation with a new version that just does either a direct
13192 move or a regular move. Move memory interface to separate insns.
13193 Move insns so they are next to the expander.
13194 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
13195 with signbit move. Split big and little endian case.
13196 (signbit<mode>2_dm_mem_le): Likewise.
13197 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
13198 (signbit<mode>2_dm2): Likewise.
13200 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13202 * config/rl78/rl78.md (anddi3): New define_expand.
13204 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13206 * config/rl78/rl78.md (umindi3): New define_expand.
13208 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13210 * config/rl78/rl78.md (smindi3): New define_expand.
13212 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13214 * config/rl78/rl78.md (smaxdi3): New define_expand.
13216 2018-01-22 Carl Love <cel@us.ibm.com>
13218 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
13219 LVX_V1TI): Add macro expansion.
13220 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
13221 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
13222 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
13223 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
13224 Change check to determine if the instruction is a byte reversing
13225 entry. Fix typo in comment.
13226 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
13227 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
13228 Add def_builtin calls for new builtins.
13229 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
13230 Add define_insn expansion.
13232 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13234 * config/rl78/rl78.md (umaxdi3): New define_expand.
13236 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
13238 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
13239 for non-QImode registers.
13241 2018-01-22 Richard Biener <rguenther@suse.de>
13243 PR tree-optimization/83963
13244 * graphite-scop-detection.c (scop_detection::get_sese): Delay
13245 including the loop exit block.
13246 (scop_detection::merge_sese): Likewise.
13247 (scop_detection::add_scop): Do it here instead.
13249 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13251 * doc/sourcebuild.texi (arm_softfloat): Document.
13253 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
13256 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
13257 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
13258 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
13260 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13261 David Edelsohn <dje.gcc@gmail.com>
13264 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13265 Change "crset eq" to "crset 2".
13266 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13267 (*call_indirect_aix<mode>_nospec): Likewise.
13268 (*call_value_indirect_aix<mode>_nospec): Likewise.
13269 (*call_indirect_elfv2<mode>_nospec): Likewise.
13270 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
13271 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
13272 change assembly output from . to $.
13273 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13274 (indirect_jump<mode>_nospec): Change assembly output from . to $.
13275 (*tablejump<mode>_internal1_nospec): Likewise.
13277 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
13280 * config/sh/sh_optimize_sett_clrt.cc:
13281 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
13283 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
13285 PR tree-optimization/83940
13286 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
13287 offset_dt to vect_constant_def rather than vect_unknown_def_type.
13288 (vect_check_load_store_mask): Add a mask_dt_out parameter and
13289 use it to pass back the definition type.
13290 (vect_check_store_rhs): Likewise rhs_dt_out.
13291 (vect_build_gather_load_calls): Add a mask_dt argument and use
13292 it instead of a call to vect_is_simple_use.
13293 (vectorizable_store): Update calls to vect_check_load_store_mask
13294 and vect_check_store_rhs. Use the dt returned by the latter instead
13295 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
13296 instead of calls to vect_is_simple_use. Pass the scalar rather
13297 than the vector operand to vect_is_simple_use when handling
13298 second and subsequent copies of an rhs value.
13299 (vectorizable_load): Update calls to vect_check_load_store_mask
13300 and vect_build_gather_load_calls. Use the cached mask_dt and
13301 gs_info.offset_dt instead of calls to vect_is_simple_use.
13303 2018-01-20 Jakub Jelinek <jakub@redhat.com>
13305 PR middle-end/83945
13306 * tree-emutls.c: Include gimplify.h.
13307 (lower_emutls_2): New function.
13308 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
13309 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
13310 it before further processing.
13313 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
13314 UINTVAL (trueop1) instead of INTVAL (op1).
13316 2018-01-19 Jakub Jelinek <jakub@redhat.com>
13320 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
13321 INCOMING_FRAME_SP_OFFSET if not defined.
13322 (scan_trace): Add ENTRY argument. If true and
13323 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
13324 emit a note to adjust the CFA offset.
13325 (create_cfi_notes): Adjust scan_trace callers.
13326 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
13327 INCOMING_FRAME_SP_OFFSET in the CIE.
13328 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
13329 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
13331 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
13332 * doc/tm.texi: Regenerated.
13334 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13336 PR rtl-optimization/83147
13337 * lra-constraints.c (remove_inheritance_pseudos): Use
13338 lra_substitute_pseudo_within_insn.
13340 2018-01-19 Tom de Vries <tom@codesourcery.com>
13341 Cesar Philippidis <cesar@codesourcery.com>
13344 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
13346 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
13349 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
13350 spaces for function labels.
13352 2018-01-19 Martin Liska <mliska@suse.cz>
13354 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
13355 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
13356 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
13357 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
13358 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
13359 (PRED_CONST_RETURN): Change from 69 to 65.
13360 (PRED_NULL_RETURN): Change from 91 to 71.
13361 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
13362 (PRED_LOOP_GUARD): Change from 66 to 73.
13364 2018-01-19 Martin Liska <mliska@suse.cz>
13366 * predict.c (predict_insn_def): Add new assert.
13367 (struct branch_predictor): Change type to signed integer.
13368 (test_prediction_value_range): Amend test to cover
13369 PROB_UNINITIALIZED.
13370 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
13371 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
13372 (PRED_LOOP_ITERATIONS_MAX): Likewise.
13373 (PRED_LOOP_IV_COMPARE): Likewise.
13374 * predict.h (PROB_UNINITIALIZED): Define new constant.
13376 2018-01-19 Martin Liska <mliska@suse.cz>
13378 * predict.c (dump_prediction): Add new format for
13379 analyze_brprob.py script which is enabled with -details
13381 * profile-count.h (precise_p): New function.
13383 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13385 PR tree-optimization/83922
13386 * tree-vect-loop.c (vect_verify_full_masking): Return false if
13387 there are no statements that need masking.
13388 (vect_active_double_reduction_p): New function.
13389 (vect_analyze_loop_operations): Use it when handling phis that
13390 are not in the loop header.
13392 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
13394 PR tree-optimization/83914
13395 * tree-vect-loop.c (vectorizable_induction): Don't convert
13396 init_expr or apply the peeling adjustment for inductions
13397 that are nested within the vectorized loop.
13399 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13401 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
13404 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13408 * function.h (gimplify_parameters): Add gimple_seq * argument.
13409 * function.c: Include gimple.h and options.h.
13410 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
13411 for the added local temporaries if needed.
13412 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
13413 if there are any parameter cleanups, wrap whole body into a
13414 try/finally with the cleanups.
13416 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
13419 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
13420 Use GET_MODE_CLASS for scalar floating point.
13422 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13426 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
13427 Fix call of call_cgraph_insertion_hooks.
13429 2018-01-18 Martin Sebor <msebor@redhat.com>
13431 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
13433 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
13436 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
13439 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
13442 * common.opt: (-ffile-prefix-map): New option.
13443 * opts.c (common_handle_option): Defer it.
13444 * opts-global.c (handle_common_deferred_options): Handle it.
13445 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
13446 * file-prefix-map.h: New file.
13447 (remap_debug_filename, add_debug_prefix_map): ...here.
13448 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
13449 * final.c (debug_prefix_map, add_debug_prefix_map
13450 remap_debug_filename): Move to...
13451 * file-prefix-map.c: New file.
13452 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
13453 generalize, get rid of alloca(), use strrchr() instead of strchr().
13454 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
13455 Implement in terms of add_prefix_map().
13456 (remap_macro_filename, remap_debug_filename): Implement in term of
13458 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
13459 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
13460 * dbxout.c: Include file-prefix-map.h.
13461 * varasm.c: Likewise.
13462 * vmsdbgout.c: Likewise.
13463 * xcoffout.c: Likewise.
13464 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
13465 * doc/cppopts.texi (-fmacro-prefix-map): Document.
13466 * doc/invoke.texi (-ffile-prefix-map): Document.
13467 (-fdebug-prefix-map): Update description.
13469 2018-01-18 Martin Liska <mliska@suse.cz>
13471 * config/i386/i386.c (indirect_thunk_name): Document that also
13473 (output_indirect_thunk): Document why both instructions
13474 (pause and lfence) are generated.
13476 2018-01-18 Richard Biener <rguenther@suse.de>
13478 PR tree-optimization/83887
13479 * graphite-scop-detection.c
13480 (scop_detection::get_nearest_dom_with_single_entry): Remove.
13481 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
13482 (scop_detection::merge_sese): Re-implement with a flood-fill
13483 algorithm that properly finds a SESE region if it exists.
13485 2018-01-18 Jakub Jelinek <jakub@redhat.com>
13488 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
13489 pointer_diff optimizations use view_convert instead of convert.
13491 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13493 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13494 Generate different code for -mno-speculate-indirect-jumps.
13495 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13496 (*call_indirect_aix<mode>): Disable for
13497 -mno-speculate-indirect-jumps.
13498 (*call_indirect_aix<mode>_nospec): New define_insn.
13499 (*call_value_indirect_aix<mode>): Disable for
13500 -mno-speculate-indirect-jumps.
13501 (*call_value_indirect_aix<mode>_nospec): New define_insn.
13502 (*sibcall_nonlocal_sysv<mode>): Generate different code for
13503 -mno-speculate-indirect-jumps.
13504 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13506 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13508 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
13509 long double type, set the flags for noting the default long double
13510 type, even if we don't pass or return a long double type.
13512 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
13515 * ipa-inline.c (flatten_function): Do not overwrite final inlining
13518 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
13520 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
13521 support for merge[hl].
13522 (fold_mergehl_helper): New helper function.
13523 (tree-vector-builder.h): New #include for tree_vector_builder usage.
13524 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
13525 (altivec_vmrglw_direct): Add xxmrglw insn.
13527 2018-01-17 Andrew Waterman <andrew@sifive.com>
13529 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13530 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13532 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13535 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13536 call the lto_location_cache before reading the
13537 DECL_SOURCE_LOCATION of the types.
13539 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13540 Richard Sandiford <richard.sandiford@linaro.org>
13542 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13543 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13544 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13545 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13546 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13548 * config/aarch64/constraints.md (aarch64_movti_operand):
13550 * config/aarch64/predicates.md (Uti): Add new constraint.
13552 2018-01-17 Carl Love <cel@us.ibm.com>
13554 * config/rs6000/vsx.md (define_expand xl_len_r,
13555 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13556 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13558 (define_expand, define_insn): Move the shift left from the
13559 define_insn to the define_expand for lxvl and stxvl instructions.
13560 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13561 and XL_LEN_R definitions to PURE.
13563 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13565 * config/i386/i386.c (indirect_thunk_name): Declare regno
13566 as unsigned int. Compare regno with INVALID_REGNUM.
13567 (output_indirect_thunk): Ditto.
13568 (output_indirect_thunk_function): Ditto.
13569 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13570 in the call to output_indirect_thunk_function.
13572 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13574 PR middle-end/83884
13575 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13576 rather than the size of inner_type to determine the stack slot size
13577 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13579 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13582 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13585 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13587 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13588 endian Linux systems to optionally enable multilibs for selecting
13589 the long double type if the user configured an explicit type.
13590 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13591 have no long double multilibs if not defined.
13592 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13593 warn if the user used -mabi={ieee,ibm}longdouble and we built
13594 multilibs for long double.
13595 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13596 appropriate multilib option.
13597 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13599 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13600 for building long double multilibs.
13601 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13603 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13605 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13608 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13610 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13613 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13616 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13619 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13621 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13622 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13624 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13626 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13627 different rtl trees depending on TARGET_64BIT.
13628 (rs6000_gen_lvx): Likewise.
13630 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13632 * config/visium/visium.md (nop): Tweak comment.
13633 (hazard_nop): Likewise.
13635 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13637 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13638 -mspeculate-indirect-jumps.
13639 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13640 for -mno-speculate-indirect-jumps.
13641 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13642 (*call_value_indirect_elfv2<mode>): Disable for
13643 -mno-speculate-indirect-jumps.
13644 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13645 (indirect_jump): Emit different RTL for
13646 -mno-speculate-indirect-jumps.
13647 (*indirect_jump<mode>): Disable for
13648 -mno-speculate-indirect-jumps.
13649 (*indirect_jump<mode>_nospec): New define_insn.
13650 (tablejump): Emit different RTL for
13651 -mno-speculate-indirect-jumps.
13652 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13653 (tablejumpsi_nospec): New define_expand.
13654 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13655 (tablejumpdi_nospec): New define_expand.
13656 (*tablejump<mode>_internal1): Disable for
13657 -mno-speculate-indirect-jumps.
13658 (*tablejump<mode>_internal1_nospec): New define_insn.
13659 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13662 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13664 * caller-save.c (insert_save): Drop unnecessary parameter. All
13667 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13668 Richard Biener <rguenth@suse.de>
13671 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13672 return early, inline manually is_gimple_sizepos. Make sure if we
13673 call gimplify_expr we don't end up with a gimple constant.
13674 * tree.c (variably_modified_type_p): Don't return true for
13675 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13676 * gimplify.h (is_gimple_sizepos): Remove.
13678 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13680 PR tree-optimization/83857
13681 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13682 vectorizable_live_operation for pure SLP statements.
13683 (vectorizable_live_operation): Handle PHIs.
13685 2018-01-16 Richard Biener <rguenther@suse.de>
13687 PR tree-optimization/83867
13688 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13689 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13691 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13694 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13695 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13696 If off is not INTEGER_CST, issue a may not be aligned warning
13697 rather than isn't aligned. Use isn%'t rather than isn't.
13698 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13700 <case MULT_EXPR>: Improve the case when bottom and one of the
13701 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13702 operand, in that case check if the other operand is multiple of
13703 bottom divided by the INTEGER_CST operand.
13705 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13708 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13709 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13710 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13711 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13712 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13713 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13714 (pa_function_arg_size): New function.
13716 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13718 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13719 in a separate statement.
13721 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13723 PR tree-optimization/83847
13724 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13725 group gathers and scatters.
13727 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13729 PR rtl-optimization/86620
13730 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13732 PR rtl-optimization/83213
13733 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13734 to last if both are JUMP_INSNs.
13736 PR tree-optimization/83843
13737 * gimple-ssa-store-merging.c
13738 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13739 store_immediate_info for bswap/nop orig_stores.
13741 2018-01-15 Andrew Waterman <andrew@sifive.com>
13743 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13745 <UDIV>: Increase cost if !TARGET_DIV.
13747 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13749 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13750 (define_attr "cr_logical_3op"): New.
13751 (cceq_ior_compare): Adjust.
13752 (cceq_ior_compare_complement): Adjust.
13753 (*cceq_rev_compare): Adjust.
13754 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13755 (is_cracked_insn): Adjust.
13756 (insn_must_be_first_in_group): Adjust.
13757 * config/rs6000/40x.md: Adjust.
13758 * config/rs6000/440.md: Adjust.
13759 * config/rs6000/476.md: Adjust.
13760 * config/rs6000/601.md: Adjust.
13761 * config/rs6000/603.md: Adjust.
13762 * config/rs6000/6xx.md: Adjust.
13763 * config/rs6000/7450.md: Adjust.
13764 * config/rs6000/7xx.md: Adjust.
13765 * config/rs6000/8540.md: Adjust.
13766 * config/rs6000/cell.md: Adjust.
13767 * config/rs6000/e300c2c3.md: Adjust.
13768 * config/rs6000/e500mc.md: Adjust.
13769 * config/rs6000/e500mc64.md: Adjust.
13770 * config/rs6000/e5500.md: Adjust.
13771 * config/rs6000/e6500.md: Adjust.
13772 * config/rs6000/mpc.md: Adjust.
13773 * config/rs6000/power4.md: Adjust.
13774 * config/rs6000/power5.md: Adjust.
13775 * config/rs6000/power6.md: Adjust.
13776 * config/rs6000/power7.md: Adjust.
13777 * config/rs6000/power8.md: Adjust.
13778 * config/rs6000/power9.md: Adjust.
13779 * config/rs6000/rs64.md: Adjust.
13780 * config/rs6000/titan.md: Adjust.
13782 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13784 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13785 ix86_indirect_branch_register logic.
13787 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13789 * config/i386/constraints.md (Bs): Update
13790 ix86_indirect_branch_register check. Don't check
13791 ix86_indirect_branch_register with GOT_memory_operand.
13793 * config/i386/predicates.md (GOT_memory_operand): Don't check
13794 ix86_indirect_branch_register here.
13795 (GOT32_symbol_operand): Likewise.
13797 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13799 * config/i386/predicates.md (constant_call_address_operand):
13800 Rewrite ix86_indirect_branch_register logic.
13801 (sibcall_insn_operand): Likewise.
13803 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13805 * config/i386/constraints.md (Bs): Replace
13806 ix86_indirect_branch_thunk_register with
13807 ix86_indirect_branch_register.
13809 * config/i386/i386.md (indirect_jump): Likewise.
13810 (tablejump): Likewise.
13811 (*sibcall_memory): Likewise.
13812 (*sibcall_value_memory): Likewise.
13813 Peepholes of indirect call and jump via memory: Likewise.
13814 * config/i386/i386.opt: Likewise.
13815 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13816 (GOT_memory_operand): Likewise.
13817 (call_insn_operand): Likewise.
13818 (sibcall_insn_operand): Likewise.
13819 (GOT32_symbol_operand): Likewise.
13821 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13823 PR middle-end/83837
13824 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13825 type rather than type addr's type points to.
13826 (expand_omp_atomic_mutex): Likewise.
13827 (expand_omp_atomic): Likewise.
13829 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13832 * config/i386/i386.c (output_indirect_thunk_function): Use
13833 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13834 for __x86_return_thunk.
13836 2018-01-15 Richard Biener <rguenther@suse.de>
13838 PR middle-end/83850
13839 * expmed.c (extract_bit_field_1): Fix typo.
13841 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13844 * config/arm/iterators.md (VF): New mode iterator.
13845 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13846 Remove integer-related logic from pattern.
13847 (neon_vabd<mode>_3): Likewise.
13849 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13851 PR middle-end/82694
13852 * common.opt (fstrict-overflow): No longer an alias.
13853 (fwrapv-pointer): New option.
13854 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13855 also for pointer types based on flag_wrapv_pointer.
13856 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13857 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13858 opts->x_flag_wrapv got set.
13859 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13860 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13861 POINTER_TYPE_OVERFLOW_UNDEFINED.
13862 * match.pd: Likewise in address comparison pattern.
13863 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13865 2018-01-15 Richard Biener <rguenther@suse.de>
13868 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13869 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13870 Reset type names to their identifier if their TYPE_DECL doesn't
13871 have linkage (and thus is used for ODR and devirt).
13872 (save_debug_info_for_decl): Remove.
13873 (save_debug_info_for_type): Likewise.
13874 (add_tree_to_fld_list): Adjust.
13875 * tree-pretty-print.c (dump_generic_node): Make dumping of
13876 type names more robust.
13878 2018-01-15 Richard Biener <rguenther@suse.de>
13880 * BASE-VER: Bump to 8.0.1.
13882 2018-01-14 Martin Sebor <msebor@redhat.com>
13885 * builtins.c (check_access): Avoid warning when the no-warning bit
13888 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13890 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13891 * ira-color (allocno_hard_regs_compare): Likewise.
13893 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13896 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13897 Use .pushsection/.popsection.
13899 2018-01-14 Martin Sebor <msebor@redhat.com>
13902 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13904 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13906 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13907 entry from extra_headers.
13908 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13909 extra_headers, make the list bitwise identical to the i?86-*-* one.
13911 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13913 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13914 -mcmodel=large with -mindirect-branch=thunk,
13915 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13916 -mfunction-return=thunk-extern.
13917 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13918 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13919 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13921 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13923 * config/i386/i386.c (print_reg): Print the name of the full
13924 integer register without '%'.
13925 (ix86_print_operand): Handle 'V'.
13926 * doc/extend.texi: Document 'V' modifier.
13928 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13930 * config/i386/constraints.md (Bs): Disallow memory operand for
13931 -mindirect-branch-register.
13933 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13934 (GOT_memory_operand): Likewise.
13935 (call_insn_operand): Likewise.
13936 (sibcall_insn_operand): Likewise.
13937 (GOT32_symbol_operand): Likewise.
13938 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13939 for -mindirect-branch-register.
13940 (tablejump): Likewise.
13941 (*sibcall_memory): Likewise.
13942 (*sibcall_value_memory): Likewise.
13943 Disallow peepholes of indirect call and jump via memory for
13944 -mindirect-branch-register.
13945 (*call_pop): Replace m with Bw.
13946 (*call_value_pop): Likewise.
13947 (*sibcall_pop_memory): Replace m with Bs.
13948 * config/i386/i386.opt (mindirect-branch-register): New option.
13949 * doc/invoke.texi: Document -mindirect-branch-register option.
13951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13953 * config/i386/i386-protos.h (ix86_output_function_return): New.
13954 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13955 set function_return_type.
13956 (indirect_thunk_name): Add ret_p to indicate thunk for function
13958 (output_indirect_thunk_function): Pass false to
13959 indirect_thunk_name.
13960 (ix86_output_indirect_branch_via_reg): Likewise.
13961 (ix86_output_indirect_branch_via_push): Likewise.
13962 (output_indirect_thunk_function): Create alias for function
13963 return thunk if regno < 0.
13964 (ix86_output_function_return): New function.
13965 (ix86_handle_fndecl_attribute): Handle function_return.
13966 (ix86_attribute_table): Add function_return.
13967 * config/i386/i386.h (machine_function): Add
13968 function_return_type.
13969 * config/i386/i386.md (simple_return_internal): Use
13970 ix86_output_function_return.
13971 (simple_return_internal_long): Likewise.
13972 * config/i386/i386.opt (mfunction-return=): New option.
13973 (indirect_branch): Mention -mfunction-return=.
13974 * doc/extend.texi: Document function_return function attribute.
13975 * doc/invoke.texi: Document -mfunction-return= option.
13977 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13979 * config/i386/i386-opts.h (indirect_branch): New.
13980 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13981 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13982 with local indirect jump when converting indirect call and jump.
13983 (ix86_set_indirect_branch_type): New.
13984 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13985 (indirectlabelno): New.
13986 (indirect_thunk_needed): Likewise.
13987 (indirect_thunk_bnd_needed): Likewise.
13988 (indirect_thunks_used): Likewise.
13989 (indirect_thunks_bnd_used): Likewise.
13990 (INDIRECT_LABEL): Likewise.
13991 (indirect_thunk_name): Likewise.
13992 (output_indirect_thunk): Likewise.
13993 (output_indirect_thunk_function): Likewise.
13994 (ix86_output_indirect_branch_via_reg): Likewise.
13995 (ix86_output_indirect_branch_via_push): Likewise.
13996 (ix86_output_indirect_branch): Likewise.
13997 (ix86_output_indirect_jmp): Likewise.
13998 (ix86_code_end): Call output_indirect_thunk_function if needed.
13999 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14001 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14002 (ix86_attribute_table): Add indirect_branch.
14003 * config/i386/i386.h (machine_function): Add indirect_branch_type
14004 and has_local_indirect_jump.
14005 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14007 (tablejump): Likewise.
14008 (*indirect_jump): Use ix86_output_indirect_jmp.
14009 (*tablejump_1): Likewise.
14010 (simple_return_indirect_internal): Likewise.
14011 * config/i386/i386.opt (mindirect-branch=): New option.
14012 (indirect_branch): New.
14015 (thunk-inline): Likewise.
14016 (thunk-extern): Likewise.
14017 * doc/extend.texi: Document indirect_branch function attribute.
14018 * doc/invoke.texi: Document -mindirect-branch= option.
14020 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14023 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14025 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14027 * ipa-inline.c (want_inline_small_function_p): Return false if
14028 inlining has already failed with CIF_FINAL_ERROR.
14029 (update_caller_keys): Call want_inline_small_function_p before
14031 (update_callee_keys): Likewise.
14033 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14035 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14037 (rs6000_quadword_masked_address_p): Likewise.
14038 (quad_aligned_load_p): Likewise.
14039 (quad_aligned_store_p): Likewise.
14040 (const_load_sequence_p): Add comment to describe the outer-most loop.
14041 (mimic_memory_attributes_and_flags): New function.
14042 (rs6000_gen_stvx): Likewise.
14043 (replace_swapped_aligned_store): Likewise.
14044 (rs6000_gen_lvx): Likewise.
14045 (replace_swapped_aligned_load): Likewise.
14046 (replace_swapped_load_constant): Capitalize argument name in
14047 comment describing this function.
14048 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14049 and stores that access quad-word aligned addresses and replace
14050 with stvx or lvx instructions when appropriate.
14051 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14052 New function prototype.
14053 (rs6000_quadword_masked_address_p): Likewise.
14054 (rs6000_gen_lvx): Likewise.
14055 (rs6000_gen_stvx): Likewise.
14056 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14057 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14058 when memory address is aligned.
14059 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14060 this split to select lvx instruction when memory address is aligned.
14061 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14062 instruction when memory address is aligned.
14063 (*vsx_le_perm_load_v16qi): Likewise.
14064 (four unnamed splitters): Modify to select the stvx instruction
14065 when memory is aligned.
14067 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14069 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14070 which appears in the queue multiple times.
14072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14073 Alan Hayward <alan.hayward@arm.com>
14074 David Sherwood <david.sherwood@arm.com>
14076 * tree-vectorizer.h (vec_lower_bound): New structure.
14077 (_loop_vec_info): Add check_nonzero and lower_bounds.
14078 (LOOP_VINFO_CHECK_NONZERO): New macro.
14079 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14080 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14081 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14082 fields. Make seg_len the distance travelled, not including the
14084 (dr_direction_indicator): Declare.
14085 (dr_zero_step_indicator): Likewise.
14086 (dr_known_forward_stride_p): Likewise.
14087 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14089 (runtime_alias_check_p): Allow runtime alias checks with
14091 (operator ==): Compare access_size and align.
14092 (prune_runtime_alias_test_list): Rework for new distinction between
14093 the access_size and seg_len.
14094 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14096 (get_segment_min_max): New function.
14097 (create_intersect_range_checks): Use it.
14098 (dr_step_indicator): New function.
14099 (dr_direction_indicator): Likewise.
14100 (dr_zero_step_indicator): Likewise.
14101 (dr_known_forward_stride_p): Likewise.
14102 * tree-loop-distribution.c (data_ref_segment_size): Return
14103 DR_STEP * (niters - 1).
14104 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14106 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14107 (vect_preserves_scalar_order_p): New function, split out from...
14108 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14109 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14110 (vect_vfa_access_size): New function.
14111 (vect_vfa_align): Likewise.
14112 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14113 (dump_lower_bound): New function.
14114 (vect_check_lower_bound): Likewise.
14115 (vect_small_gap_p): Likewise.
14116 (vectorizable_with_step_bound_p): Likewise.
14117 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14118 depencies if the vectorization factor is 1. Convert the checks
14119 for nonzero steps into checks on the bounds of DR_STEP. Try using
14120 a bunds check for variable steps if the minimum required step is
14121 relatively small. Update calls to the dr_with_seg_len
14122 constructor and to vect_compile_time_alias.
14123 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14125 (vect_loop_versioning): Call it.
14126 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14128 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14131 Alan Hayward <alan.hayward@arm.com>
14132 David Sherwood <david.sherwood@arm.com>
14134 * doc/sourcebuild.texi (vect_scatter_store): Document.
14135 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14137 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14139 * genopinit.c (main): Add supports_vec_scatter_store and
14140 supports_vec_scatter_store_cached to target_optabs.
14141 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14142 IFN_MASK_SCATTER_STORE.
14143 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14145 * internal-fn.h (internal_store_fn_p): Declare.
14146 (internal_fn_stored_value_index): Likewise.
14147 * internal-fn.c (scatter_store_direct): New macro.
14148 (expand_scatter_store_optab_fn): New function.
14149 (direct_scatter_store_optab_supported_p): New macro.
14150 (internal_store_fn_p): New function.
14151 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
14152 IFN_MASK_SCATTER_STORE.
14153 (internal_fn_mask_index): Likewise.
14154 (internal_fn_stored_value_index): New function.
14155 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
14156 for scatter stores.
14157 * optabs-query.h (supports_vec_scatter_store_p): Declare.
14158 * optabs-query.c (supports_vec_scatter_store_p): New function.
14159 * tree-vectorizer.h (vect_get_store_rhs): Declare.
14160 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
14161 true for scatter stores.
14162 (vect_gather_scatter_fn_p): Handle scatter stores too.
14163 (vect_check_gather_scatter): Consider using scatter stores if
14164 supports_vec_scatter_store_p.
14165 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
14166 scatter stores too.
14167 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14168 internal_fn_stored_value_index.
14169 (check_load_store_masking): Handle scatter stores too.
14170 (vect_get_store_rhs): Make public.
14171 (vectorizable_call): Use internal_store_fn_p.
14172 (vectorizable_store): Handle scatter store internal functions.
14173 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
14174 when deciding whether the end of the group has been reached.
14175 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
14176 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
14177 (mask_scatter_store<mode>): New insns.
14179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14180 Alan Hayward <alan.hayward@arm.com>
14181 David Sherwood <david.sherwood@arm.com>
14183 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
14184 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
14185 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
14187 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
14188 Use vect_truncate_gather_scatter_offset if we can't treat the
14189 operation as a normal gather load or scatter store.
14190 (get_group_load_store_type): Take the gather_scatter_info
14191 as argument. Try using a gather load or scatter store for
14192 single-element groups.
14193 (get_load_store_type): Update calls to get_group_load_store_type
14194 and vect_use_strided_gather_scatters_p.
14196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14197 Alan Hayward <alan.hayward@arm.com>
14198 David Sherwood <david.sherwood@arm.com>
14200 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
14201 optional tree argument.
14202 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
14204 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
14205 but continue to use the current value as a fallback.
14206 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
14207 to compare the updates.
14208 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
14209 (get_load_store_type): Use it when handling a strided access.
14210 (vect_get_strided_load_store_ops): New function.
14211 (vect_get_data_ptr_increment): Likewise.
14212 (vectorizable_load): Handle strided gather loads. Always pass
14213 a step to vect_create_data_ref_ptr and bump_vector_ptr.
14215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14216 Alan Hayward <alan.hayward@arm.com>
14217 David Sherwood <david.sherwood@arm.com>
14219 * doc/md.texi (gather_load@var{m}): Document.
14220 (mask_gather_load@var{m}): Likewise.
14221 * genopinit.c (main): Add supports_vec_gather_load and
14222 supports_vec_gather_load_cached to target_optabs.
14223 * optabs-tree.c (init_tree_optimization_optabs): Use
14224 ggc_cleared_alloc to allocate target_optabs.
14225 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
14226 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
14228 * internal-fn.h (internal_load_fn_p): Declare.
14229 (internal_gather_scatter_fn_p): Likewise.
14230 (internal_fn_mask_index): Likewise.
14231 (internal_gather_scatter_fn_supported_p): Likewise.
14232 * internal-fn.c (gather_load_direct): New macro.
14233 (expand_gather_load_optab_fn): New function.
14234 (direct_gather_load_optab_supported_p): New macro.
14235 (direct_internal_fn_optab): New function.
14236 (internal_load_fn_p): Likewise.
14237 (internal_gather_scatter_fn_p): Likewise.
14238 (internal_fn_mask_index): Likewise.
14239 (internal_gather_scatter_fn_supported_p): Likewise.
14240 * optabs-query.c (supports_at_least_one_mode_p): New function.
14241 (supports_vec_gather_load_p): Likewise.
14242 * optabs-query.h (supports_vec_gather_load_p): Declare.
14243 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
14244 and memory_type field.
14245 (NUM_PATTERNS): Bump to 15.
14246 * tree-vect-data-refs.c: Include internal-fn.h.
14247 (vect_gather_scatter_fn_p): New function.
14248 (vect_describe_gather_scatter_call): Likewise.
14249 (vect_check_gather_scatter): Try using internal functions for
14250 gather loads. Recognize existing calls to a gather load function.
14251 (vect_analyze_data_refs): Consider using gather loads if
14252 supports_vec_gather_load_p.
14253 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
14254 (vect_get_gather_scatter_offset_type): Likewise.
14255 (vect_convert_mask_for_vectype): Likewise.
14256 (vect_add_conversion_to_patterm): Likewise.
14257 (vect_try_gather_scatter_pattern): Likewise.
14258 (vect_recog_gather_scatter_pattern): New pattern recognizer.
14259 (vect_vect_recog_func_ptrs): Add it.
14260 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
14261 internal_fn_mask_index and internal_gather_scatter_fn_p.
14262 (check_load_store_masking): Take the gather_scatter_info as an
14263 argument and handle gather loads.
14264 (vect_get_gather_scatter_ops): New function.
14265 (vectorizable_call): Check internal_load_fn_p.
14266 (vectorizable_load): Likewise. Handle gather load internal
14268 (vectorizable_store): Update call to check_load_store_masking.
14269 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
14270 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
14271 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
14272 (aarch64_gather_scale_operand_d): New predicates.
14273 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
14274 (mask_gather_load<mode>): New insns.
14276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14277 Alan Hayward <alan.hayward@arm.com>
14278 David Sherwood <david.sherwood@arm.com>
14280 * optabs.def (fold_left_plus_optab): New optab.
14281 * doc/md.texi (fold_left_plus_@var{m}): Document.
14282 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
14283 * internal-fn.c (fold_left_direct): Define.
14284 (expand_fold_left_optab_fn): Likewise.
14285 (direct_fold_left_optab_supported_p): Likewise.
14286 * fold-const-call.c (fold_const_fold_left): New function.
14287 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
14288 * tree-parloops.c (valid_reduction_p): New function.
14289 (gather_scalar_reductions): Use it.
14290 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
14291 (vect_finish_replace_stmt): Declare.
14292 * tree-vect-loop.c (fold_left_reduction_fn): New function.
14293 (needs_fold_left_reduction_p): New function, split out from...
14294 (vect_is_simple_reduction): ...here. Accept reductions that
14295 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
14296 (vect_force_simple_reduction): Also store the reduction type in
14297 the assignment's STMT_VINFO_REDUC_TYPE.
14298 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
14299 (merge_with_identity): New function.
14300 (vect_expand_fold_left): Likewise.
14301 (vectorize_fold_left_reduction): Likewise.
14302 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
14303 scalar phi in place for it. Check for target support and reject
14304 cases that would reassociate the operation. Defer the transform
14305 phase to vectorize_fold_left_reduction.
14306 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
14307 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
14308 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
14310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14312 * tree-if-conv.c (predicate_mem_writes): Remove redundant
14313 call to ifc_temp_var.
14315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14316 Alan Hayward <alan.hayward@arm.com>
14317 David Sherwood <david.sherwood@arm.com>
14319 * target.def (legitimize_address_displacement): Take the original
14320 offset as a poly_int.
14321 * targhooks.h (default_legitimize_address_displacement): Update
14323 * targhooks.c (default_legitimize_address_displacement): Likewise.
14324 * doc/tm.texi: Regenerate.
14325 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
14326 as an argument, moving assert of ad->disp == ad->disp_term to...
14327 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
14328 Try calling targetm.legitimize_address_displacement before expanding
14329 the address rather than afterwards, and adjust for the new interface.
14330 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
14331 Match the new hook interface. Handle SVE addresses.
14332 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
14333 new hook interface.
14335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14337 * Makefile.in (OBJS): Add early-remat.o.
14338 * target.def (select_early_remat_modes): New hook.
14339 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
14340 * doc/tm.texi: Regenerate.
14341 * targhooks.h (default_select_early_remat_modes): Declare.
14342 * targhooks.c (default_select_early_remat_modes): New function.
14343 * timevar.def (TV_EARLY_REMAT): New timevar.
14344 * passes.def (pass_early_remat): New pass.
14345 * tree-pass.h (make_pass_early_remat): Declare.
14346 * early-remat.c: New file.
14347 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
14349 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
14351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14352 Alan Hayward <alan.hayward@arm.com>
14353 David Sherwood <david.sherwood@arm.com>
14355 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
14356 vfm1 with a bound_epilog parameter.
14357 (vect_do_peeling): Update calls accordingly, and move the prologue
14358 call earlier in the function. Treat the base bound_epilog as 0 for
14359 fully-masked loops and retain vf - 1 for other loops. Add 1 to
14360 this base when peeling for gaps.
14361 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
14362 with fully-masked loops.
14363 (vect_estimate_min_profitable_iters): Handle the single peeled
14364 iteration in that case.
14366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14367 Alan Hayward <alan.hayward@arm.com>
14368 David Sherwood <david.sherwood@arm.com>
14370 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
14371 single-element interleaving even if the size is not a power of 2.
14372 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
14373 accesses for single-element interleaving if the group size is
14376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14377 Alan Hayward <alan.hayward@arm.com>
14378 David Sherwood <david.sherwood@arm.com>
14380 * doc/md.texi (fold_extract_last_@var{m}): Document.
14381 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
14382 * optabs.def (fold_extract_last_optab): New optab.
14383 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
14384 * internal-fn.c (fold_extract_direct): New macro.
14385 (expand_fold_extract_optab_fn): Likewise.
14386 (direct_fold_extract_optab_supported_p): Likewise.
14387 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
14388 * tree-vect-loop.c (vect_model_reduction_cost): Handle
14389 EXTRACT_LAST_REDUCTION.
14390 (get_initial_def_for_reduction): Do not create an initial vector
14391 for EXTRACT_LAST_REDUCTION reductions.
14392 (vectorizable_reduction): Leave the scalar phi in place for
14393 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
14394 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
14395 epilogue code for EXTRACT_LAST_REDUCTION and defer the
14396 transform phase to vectorizable_condition.
14397 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
14399 (vect_finish_stmt_generation): ...here.
14400 (vect_finish_replace_stmt): New function.
14401 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
14402 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
14404 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
14406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14407 Alan Hayward <alan.hayward@arm.com>
14408 David Sherwood <david.sherwood@arm.com>
14410 * doc/md.texi (extract_last_@var{m}): Document.
14411 * optabs.def (extract_last_optab): New optab.
14412 * internal-fn.def (EXTRACT_LAST): New internal function.
14413 * internal-fn.c (cond_unary_direct): New macro.
14414 (expand_cond_unary_optab_fn): Likewise.
14415 (direct_cond_unary_optab_supported_p): Likewise.
14416 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
14417 loops using EXTRACT_LAST.
14418 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
14419 (extract_last_<mode>): ...this optab.
14420 (vec_extract<mode><Vel>): Update accordingly.
14422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14423 Alan Hayward <alan.hayward@arm.com>
14424 David Sherwood <david.sherwood@arm.com>
14426 * target.def (empty_mask_is_expensive): New hook.
14427 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
14428 * doc/tm.texi: Regenerate.
14429 * targhooks.h (default_empty_mask_is_expensive): Declare.
14430 * targhooks.c (default_empty_mask_is_expensive): New function.
14431 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
14432 if the target says that empty masks are expensive.
14433 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
14435 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
14437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14438 Alan Hayward <alan.hayward@arm.com>
14439 David Sherwood <david.sherwood@arm.com>
14441 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
14442 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
14443 (vect_use_loop_mask_for_alignment_p): New function.
14444 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
14445 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
14446 niters_skip argument. Make sure that the first niters_skip elements
14447 of the first iteration are inactive.
14448 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
14449 Update call to vect_set_loop_masks_directly.
14450 (get_misalign_in_elems): New function, split out from...
14451 (vect_gen_prolog_loop_niters): ...here.
14452 (vect_update_init_of_dr): Take a code argument that specifies whether
14453 the adjustment should be added or subtracted.
14454 (vect_update_init_of_drs): Likewise.
14455 (vect_prepare_for_masked_peels): New function.
14456 (vect_do_peeling): Skip prologue peeling if we're using a mask
14457 instead. Update call to vect_update_inits_of_drs.
14458 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14460 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
14461 alignment. Do not include the number of peeled iterations in
14462 the minimum threshold in that case.
14463 (vectorizable_induction): Adjust the start value down by
14464 LOOP_VINFO_MASK_SKIP_NITERS iterations.
14465 (vect_transform_loop): Call vect_prepare_for_masked_peels.
14466 Take the number of skipped iterations into account when calculating
14468 * tree-vect-stmts.c (vect_gen_while_not): New function.
14470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14471 Alan Hayward <alan.hayward@arm.com>
14472 David Sherwood <david.sherwood@arm.com>
14474 * doc/sourcebuild.texi (vect_fully_masked): Document.
14475 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
14476 default value to 0.
14477 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
14479 (vect_analyze_loop_2): ...here. Don't check the vectorization
14480 factor against the number of loop iterations if the loop is
14483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14484 Alan Hayward <alan.hayward@arm.com>
14485 David Sherwood <david.sherwood@arm.com>
14487 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
14488 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
14489 (dump_groups): Update accordingly.
14490 (iv_use::mem_type): New member variable.
14491 (address_p): New function.
14492 (record_use): Add a mem_type argument and initialize the new
14494 (record_group_use): Add a mem_type argument. Use address_p.
14495 Remove obsolete null checks of base_object. Update call to record_use.
14496 (find_interesting_uses_op): Update call to record_group_use.
14497 (find_interesting_uses_cond): Likewise.
14498 (find_interesting_uses_address): Likewise.
14499 (get_mem_type_for_internal_fn): New function.
14500 (find_address_like_use): Likewise.
14501 (find_interesting_uses_stmt): Try find_address_like_use before
14502 calling find_interesting_uses_op.
14503 (addr_offset_valid_p): Use the iv mem_type field as the type
14504 of the addressed memory.
14505 (add_autoinc_candidates): Likewise.
14506 (get_address_cost): Likewise.
14507 (split_small_address_groups_p): Use address_p.
14508 (split_address_groups): Likewise.
14509 (add_iv_candidate_for_use): Likewise.
14510 (autoinc_possible_for_pair): Likewise.
14511 (rewrite_groups): Likewise.
14512 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
14513 (determine_group_iv_cost): Update after split of USE_ADDRESS.
14514 (get_alias_ptr_type_for_ptr_address): New function.
14515 (rewrite_use_address): Rewrite address uses in calls that were
14516 identified by find_address_like_use.
14518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14519 Alan Hayward <alan.hayward@arm.com>
14520 David Sherwood <david.sherwood@arm.com>
14522 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
14524 * gimple-expr.h (is_gimple_addressable: Likewise.
14525 * gimple-expr.c (is_gimple_address): Likewise.
14526 * internal-fn.c (expand_call_mem_ref): New function.
14527 (expand_mask_load_optab_fn): Use it.
14528 (expand_mask_store_optab_fn): Likewise.
14530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14531 Alan Hayward <alan.hayward@arm.com>
14532 David Sherwood <david.sherwood@arm.com>
14534 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14535 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14536 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14537 (cond_umax@var{mode}): Document.
14538 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14539 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14540 (cond_umin_optab, cond_umax_optab): New optabs.
14541 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14542 (COND_IOR, COND_XOR): New internal functions.
14543 * internal-fn.h (get_conditional_internal_fn): Declare.
14544 * internal-fn.c (cond_binary_direct): New macro.
14545 (expand_cond_binary_optab_fn): Likewise.
14546 (direct_cond_binary_optab_supported_p): Likewise.
14547 (get_conditional_internal_fn): New function.
14548 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14549 Cope with reduction statements that are vectorized as calls rather
14551 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14552 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14553 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14554 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14555 (UNSPEC_COND_EOR): New unspecs.
14556 (optab): Add mappings for them.
14557 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14558 (sve_int_op, sve_fp_op): New int attributes.
14560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * optabs.def (while_ult_optab): New optab.
14565 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14566 * internal-fn.def (WHILE_ULT): New internal function.
14567 * internal-fn.h (direct_internal_fn_supported_p): New override
14568 that takes two types as argument.
14569 * internal-fn.c (while_direct): New macro.
14570 (expand_while_optab_fn): New function.
14571 (convert_optab_supported_p): Likewise.
14572 (direct_while_optab_supported_p): New macro.
14573 * wide-int.h (wi::udiv_ceil): New function.
14574 * tree-vectorizer.h (rgroup_masks): New structure.
14575 (vec_loop_masks): New typedef.
14576 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14577 and fully_masked_p.
14578 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14579 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14580 (vect_max_vf): New function.
14581 (slpeel_make_loop_iterate_ntimes): Delete.
14582 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14583 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14584 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14585 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14586 internal-fn.h, stor-layout.h and optabs-query.h.
14587 (vect_set_loop_mask): New function.
14588 (add_preheader_seq): Likewise.
14589 (add_header_seq): Likewise.
14590 (interleave_supported_p): Likewise.
14591 (vect_maybe_permute_loop_masks): Likewise.
14592 (vect_set_loop_masks_directly): Likewise.
14593 (vect_set_loop_condition_masked): Likewise.
14594 (vect_set_loop_condition_unmasked): New function, split out from
14595 slpeel_make_loop_iterate_ntimes.
14596 (slpeel_make_loop_iterate_ntimes): Rename to..
14597 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14598 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14599 (vect_do_peeling): Update call accordingly.
14600 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14602 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14603 mask_compare_type, can_fully_mask_p and fully_masked_p.
14604 (release_vec_loop_masks): New function.
14605 (_loop_vec_info): Use it to free the loop masks.
14606 (can_produce_all_loop_masks_p): New function.
14607 (vect_get_max_nscalars_per_iter): Likewise.
14608 (vect_verify_full_masking): Likewise.
14609 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14610 retries, and free the mask rgroups before retrying. Check loop-wide
14611 reasons for disallowing fully-masked loops. Make the final decision
14612 about whether use a fully-masked loop or not.
14613 (vect_estimate_min_profitable_iters): Do not assume that peeling
14614 for the number of iterations will be needed for fully-masked loops.
14615 (vectorizable_reduction): Disable fully-masked loops.
14616 (vectorizable_live_operation): Likewise.
14617 (vect_halve_mask_nunits): New function.
14618 (vect_double_mask_nunits): Likewise.
14619 (vect_record_loop_mask): Likewise.
14620 (vect_get_loop_mask): Likewise.
14621 (vect_transform_loop): Handle the case in which the final loop
14622 iteration might handle a partial vector. Call vect_set_loop_condition
14623 instead of slpeel_make_loop_iterate_ntimes.
14624 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14625 (check_load_store_masking): New function.
14626 (prepare_load_store_mask): Likewise.
14627 (vectorizable_store): Handle fully-masked loops.
14628 (vectorizable_load): Likewise.
14629 (supportable_widening_operation): Use vect_halve_mask_nunits for
14631 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14632 (vect_gen_while): New function.
14633 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14634 (aarch64_uqdec<mode>): New insn.
14636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14637 Alan Hayward <alan.hayward@arm.com>
14638 David Sherwood <david.sherwood@arm.com>
14640 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14641 (reduc_xor_scal_optab): New optabs.
14642 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14643 (reduc_xor_scal_@var{m}): Document.
14644 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14645 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14646 internal functions.
14647 * fold-const-call.c (fold_const_call): Handle them.
14648 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14649 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14650 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14651 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14652 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14653 (UNSPEC_XORV): New unspecs.
14654 (optab): Add entries for them.
14655 (BITWISEV): New int iterator.
14656 (bit_reduc_op): New int attributes.
14658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14659 Alan Hayward <alan.hayward@arm.com>
14660 David Sherwood <david.sherwood@arm.com>
14662 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14663 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14664 * optabs.def (vec_shl_insert_optab): New optab.
14665 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14666 (duplicate_and_interleave): Likewise.
14667 * tree-vect-loop.c: Include internal-fn.h.
14668 (neutral_op_for_slp_reduction): New function, split out from
14669 get_initial_defs_for_reduction.
14670 (get_initial_def_for_reduction): Handle option 2 for variable-length
14671 vectors by loading the neutral value into a vector and then shifting
14672 the initial value into element 0.
14673 (get_initial_defs_for_reduction): Replace the code argument with
14674 the neutral value calculated by neutral_op_for_slp_reduction.
14675 Use gimple_build_vector for constant-length vectors.
14676 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14677 but the first group_size elements have a neutral value.
14678 Use duplicate_and_interleave otherwise.
14679 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14680 Update call to get_initial_defs_for_reduction. Handle SLP
14681 reductions for variable-length vectors by creating one vector
14682 result for each scalar result, with the elements associated
14683 with other scalar results stubbed out with the neutral value.
14684 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14685 Require IFN_VEC_SHL_INSERT for double reductions on
14686 variable-length vectors, or SLP reductions that have
14687 a neutral value. Require can_duplicate_and_interleave_p
14688 support for variable-length unchained SLP reductions if there
14689 is no neutral value, such as for MIN/MAX reductions. Also require
14690 the number of vector elements to be a multiple of the number of
14691 SLP statements when doing variable-length unchained SLP reductions.
14692 Update call to vect_create_epilog_for_reduction.
14693 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14694 and remove initial values.
14695 (duplicate_and_interleave): Make public.
14696 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14697 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14700 Alan Hayward <alan.hayward@arm.com>
14701 David Sherwood <david.sherwood@arm.com>
14703 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14704 (can_duplicate_and_interleave_p): New function.
14705 (vect_get_and_check_slp_defs): Take the vector of statements
14706 rather than just the current one. Remove excess parentheses.
14707 Restriction rejectinon of vect_constant_def and vect_external_def
14708 for variable-length vectors to boolean types, or types for which
14709 can_duplicate_and_interleave_p is false.
14710 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14711 (duplicate_and_interleave): New function.
14712 (vect_get_constant_vectors): Use gimple_build_vector for
14713 constant-length vectors and suitable variable-length constant
14714 vectors. Use duplicate_and_interleave for other variable-length
14715 vectors. Don't defer the update when inserting new statements.
14717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14718 Alan Hayward <alan.hayward@arm.com>
14719 David Sherwood <david.sherwood@arm.com>
14721 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14722 min_profitable_iters doesn't go negative.
14724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14725 Alan Hayward <alan.hayward@arm.com>
14726 David Sherwood <david.sherwood@arm.com>
14728 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14729 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14730 * optabs.def (vec_mask_load_lanes_optab): New optab.
14731 (vec_mask_store_lanes_optab): Likewise.
14732 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14733 (MASK_STORE_LANES): Likewise.
14734 * internal-fn.c (mask_load_lanes_direct): New macro.
14735 (mask_store_lanes_direct): Likewise.
14736 (expand_mask_load_optab_fn): Handle masked operations.
14737 (expand_mask_load_lanes_optab_fn): New macro.
14738 (expand_mask_store_optab_fn): Handle masked operations.
14739 (expand_mask_store_lanes_optab_fn): New macro.
14740 (direct_mask_load_lanes_optab_supported_p): Likewise.
14741 (direct_mask_store_lanes_optab_supported_p): Likewise.
14742 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14744 (vect_load_lanes_supported): Likewise.
14745 * tree-vect-data-refs.c (strip_conversion): New function.
14746 (can_group_stmts_p): Likewise.
14747 (vect_analyze_data_ref_accesses): Use it instead of checking
14748 for a pair of assignments.
14749 (vect_store_lanes_supported): Take a masked_p parameter.
14750 (vect_load_lanes_supported): Likewise.
14751 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14752 vect_store_lanes_supported and vect_load_lanes_supported.
14753 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14754 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14755 parameter. Don't allow gaps for masked accesses.
14756 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14757 and vect_load_lanes_supported.
14758 (get_load_store_type): Take a masked_p parameter and update
14759 call to get_group_load_store_type.
14760 (vectorizable_store): Update call to get_load_store_type.
14761 Handle IFN_MASK_STORE_LANES.
14762 (vectorizable_load): Update call to get_load_store_type.
14763 Handle IFN_MASK_LOAD_LANES.
14765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14766 Alan Hayward <alan.hayward@arm.com>
14767 David Sherwood <david.sherwood@arm.com>
14769 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14771 * config/aarch64/aarch64-protos.h
14772 (aarch64_sve_struct_memory_operand_p): Declare.
14773 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14774 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14775 (VPRED, vpred): Handle SVE structure modes.
14776 * config/aarch64/constraints.md (Utx): New constraint.
14777 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14778 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14779 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14780 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14781 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14782 structure modes. Split into pieces after RA.
14783 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14784 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14786 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14787 SVE structure modes.
14788 (aarch64_classify_address): Likewise.
14789 (sizetochar): Move earlier in file.
14790 (aarch64_print_operand): Handle SVE register lists.
14791 (aarch64_array_mode): New function.
14792 (aarch64_sve_struct_memory_operand_p): Likewise.
14793 (TARGET_ARRAY_MODE): Redefine.
14795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14796 Alan Hayward <alan.hayward@arm.com>
14797 David Sherwood <david.sherwood@arm.com>
14799 * target.def (array_mode): New target hook.
14800 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14801 * doc/tm.texi: Regenerate.
14802 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14803 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14804 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14805 targetm.array_mode.
14806 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14810 Alan Hayward <alan.hayward@arm.com>
14811 David Sherwood <david.sherwood@arm.com>
14813 * fold-const.c (fold_binary_loc): Check the argument types
14814 rather than the result type when testing for a vector operation.
14816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14818 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14819 * doc/tm.texi: Regenerate.
14821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14822 Alan Hayward <alan.hayward@arm.com>
14823 David Sherwood <david.sherwood@arm.com>
14825 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14826 (sve): Document new AArch64 extension.
14827 * doc/md.texi (w): Extend the description of the AArch64
14828 constraint to include SVE vectors.
14829 (Upl, Upa): Document new AArch64 predicate constraints.
14830 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14832 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14833 (msve-vector-bits=): New option.
14834 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14835 SVE when these are disabled.
14836 (sve): New extension.
14837 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14838 modes. Adjust their number of units based on aarch64_sve_vg.
14839 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14840 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14841 aarch64_addr_query_type.
14842 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14843 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14844 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14845 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14846 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14847 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14848 (aarch64_simd_imm_zero_p): Delete.
14849 (aarch64_check_zero_based_sve_index_immediate): Declare.
14850 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14851 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14852 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14853 (aarch64_sve_float_mul_immediate_p): Likewise.
14854 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14855 rather than an rtx.
14856 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14857 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14858 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14859 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14860 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14861 (aarch64_regmode_natural_size): Likewise.
14862 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14863 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14865 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14866 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14867 for VG and the SVE predicate registers.
14868 (V_ALIASES): Add a "z"-prefixed alias.
14869 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14870 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14871 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14872 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14873 (REG_CLASS_NAMES): Add entries for them.
14874 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14875 and the predicate registers.
14876 (aarch64_sve_vg): Declare.
14877 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14878 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14879 (REGMODE_NATURAL_SIZE): Define.
14880 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14882 * config/aarch64/aarch64.c: Include cfgrtl.h.
14883 (simd_immediate_info): Add a constructor for series vectors,
14884 and an associated step field.
14885 (aarch64_sve_vg): New variable.
14886 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14887 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14888 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14889 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14890 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14891 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14892 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14893 (aarch64_get_mask_mode): New functions.
14894 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14895 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14896 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14897 predicate modes and predicate registers. Explicitly restrict
14898 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14899 to store a vector mode if it is recognized by
14900 aarch64_classify_vector_mode.
14901 (aarch64_regmode_natural_size): New function.
14902 (aarch64_hard_regno_caller_save_mode): Return the original mode
14904 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14905 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14906 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14907 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14909 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14910 does not overlap dest if the function is frame-related. Handle
14912 (aarch64_split_add_offset): New function.
14913 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14914 them aarch64_add_offset.
14915 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14916 and update call to aarch64_sub_sp.
14917 (aarch64_add_cfa_expression): New function.
14918 (aarch64_expand_prologue): Pass extra temporary registers to the
14919 functions above. Handle the case in which we need to emit new
14920 DW_CFA_expressions for registers that were originally saved
14921 relative to the stack pointer, but now have to be expressed
14922 relative to the frame pointer.
14923 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14925 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14926 IP0 and IP1 values for SVE frames.
14927 (aarch64_expand_vec_series): New function.
14928 (aarch64_expand_sve_widened_duplicate): Likewise.
14929 (aarch64_expand_sve_const_vector): Likewise.
14930 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14931 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14932 into the register, rather than emitting a SET directly.
14933 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14934 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14935 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14936 (offset_9bit_signed_scaled_p): New functions.
14937 (aarch64_replicate_bitmask_imm): New function.
14938 (aarch64_bitmask_imm): Use it.
14939 (aarch64_cannot_force_const_mem): Reject expressions involving
14940 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14941 (aarch64_classify_index): Handle SVE indices, by requiring
14942 a plain register index with a scale that matches the element size.
14943 (aarch64_classify_address): Handle SVE addresses. Assert that
14944 the mode of the address is VOIDmode or an integer mode.
14945 Update call to aarch64_classify_symbol.
14946 (aarch64_classify_symbolic_expression): Update call to
14947 aarch64_classify_symbol.
14948 (aarch64_const_vec_all_in_range_p): New function.
14949 (aarch64_print_vector_float_operand): Likewise.
14950 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14951 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14952 and the FP immediates 1.0 and 0.5.
14953 (aarch64_print_address_internal): Handle SVE addresses.
14954 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14955 (aarch64_regno_regclass): Handle predicate registers.
14956 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14958 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14959 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14960 (aarch64_convert_sve_vector_bits): New function.
14961 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14962 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14963 rather than an rtx.
14964 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14965 Handle SVE vector and predicate modes. Accept VL-based constants
14966 that need only one temporary register, and VL offsets that require
14967 no temporary registers.
14968 (aarch64_conditional_register_usage): Mark the predicate registers
14969 as fixed if SVE isn't available.
14970 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14971 Return true for SVE vector and predicate modes.
14972 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14973 rather than an unsigned int. Handle SVE modes.
14974 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14976 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14978 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14979 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14980 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14981 (aarch64_sve_float_mul_immediate_p): New functions.
14982 (aarch64_sve_valid_immediate): New function.
14983 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14984 Explicitly reject structure modes. Check for INDEX constants.
14985 Handle PTRUE and PFALSE constants.
14986 (aarch64_check_zero_based_sve_index_immediate): New function.
14987 (aarch64_simd_imm_zero_p): Delete.
14988 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14989 vector modes. Accept constants in the range of CNT[BHWD].
14990 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14991 ask for an Advanced SIMD mode.
14992 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14993 (aarch64_simd_vector_alignment): Handle SVE predicates.
14994 (aarch64_vectorize_preferred_vector_alignment): New function.
14995 (aarch64_simd_vector_alignment_reachable): Use it instead of
14997 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14998 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15000 (MAX_VECT_LEN): Delete.
15001 (expand_vec_perm_d): Add a vec_flags field.
15002 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15003 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15004 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15006 (aarch64_evpc_rev): Rename to...
15007 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15008 (aarch64_evpc_rev_global): New function.
15009 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15010 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15012 (aarch64_evpc_sve_tbl): New function.
15013 (aarch64_expand_vec_perm_const_1): Update after rename of
15014 aarch64_evpc_rev. Handle SVE permutes too, trying
15015 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15016 than aarch64_evpc_tbl.
15017 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15018 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15019 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15020 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15021 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15022 (aarch64_expand_sve_vcond): New functions.
15023 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15024 of aarch64_vector_mode_p.
15025 (aarch64_dwarf_poly_indeterminate_value): New function.
15026 (aarch64_compute_pressure_classes): Likewise.
15027 (aarch64_can_change_mode_class): Likewise.
15028 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15029 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15030 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15031 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15032 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15033 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15034 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15035 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15037 (Dn, Dl, Dr): Accept const as well as const_vector.
15038 (Dz): Likewise. Compare against CONST0_RTX.
15039 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15040 of "vector" where appropriate.
15041 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15042 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15043 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15044 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15045 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15046 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15047 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15048 (v_int_equiv): Extend to SVE modes.
15049 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15051 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15052 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15053 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15054 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15055 (SVE_COND_FP_CMP): New int iterators.
15056 (perm_hilo): Handle the new unpack unspecs.
15057 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15059 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15060 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15061 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15062 (aarch64_equality_operator, aarch64_constant_vector_operand)
15063 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15064 (aarch64_sve_nonimmediate_operand): Likewise.
15065 (aarch64_sve_general_operand): Likewise.
15066 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15067 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15068 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15069 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15070 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15071 (aarch64_sve_float_arith_immediate): Likewise.
15072 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15073 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15074 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15075 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15076 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15077 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15078 (aarch64_sve_float_arith_operand): Likewise.
15079 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15080 (aarch64_sve_float_mul_operand): Likewise.
15081 (aarch64_sve_vec_perm_operand): Likewise.
15082 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15083 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15084 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15085 as well as const_vector.
15086 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15087 in file. Use CONST0_RTX and CONSTM1_RTX.
15088 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15089 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15090 Use aarch64_simd_imm_zero.
15091 * config/aarch64/aarch64-sve.md: New file.
15092 * config/aarch64/aarch64.md: Include it.
15093 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15094 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15095 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15096 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15097 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15098 (sve): New attribute.
15099 (enabled): Disable instructions with the sve attribute unless
15101 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15102 aarch64_expand_mov_immediate.
15103 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15104 CNT[BHSD] immediates.
15105 (movti): Split CONST_POLY_INT moves into two halves.
15106 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15107 Split additions that need a temporary here if the destination
15108 is the stack pointer.
15109 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15110 (*add<mode>3_poly_1): New instruction.
15111 (set_clobber_cc): New expander.
15113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15115 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15116 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15117 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15118 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15119 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15120 Change innermode from fixed_mode_size to machine_mode.
15121 (simplify_subreg): Update call accordingly. Handle a constant-sized
15122 subreg of a variable-length CONST_VECTOR.
15124 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15125 Alan Hayward <alan.hayward@arm.com>
15126 David Sherwood <david.sherwood@arm.com>
15128 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15129 (add_offset_to_base): New function, split out from...
15130 (create_mem_ref): ...here. When handling a scale other than 1,
15131 check first whether the address is valid without the offset.
15132 Add it into the base if so, leaving the index and scale as-is.
15134 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15137 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15138 fold_for_warn before checking if arg2 is INTEGER_CST.
15140 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15142 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15143 (store_multiple_operation): Delete.
15144 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15145 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15146 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15147 guarded by TARGET_STRING.
15148 (rs6000_output_load_multiple): Delete.
15149 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15150 OPTION_MASK_STRING / TARGET_STRING handling.
15151 (print_operand) <'N', 'O'>: Add comment that these are unused now.
15152 (const rs6000_opt_masks) <"string">: Change mask to 0.
15153 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
15154 (MASK_STRING): Delete.
15155 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
15157 (load_multiple): Delete.
15164 (store_multiple): Delete.
15171 (movmemsi_8reg): Delete.
15172 (corresponding unnamed define_insn): Delete.
15173 (movmemsi_6reg): Delete.
15174 (corresponding unnamed define_insn): Delete.
15175 (movmemsi_4reg): Delete.
15176 (corresponding unnamed define_insn): Delete.
15177 (movmemsi_2reg): Delete.
15178 (corresponding unnamed define_insn): Delete.
15179 (movmemsi_1reg): Delete.
15180 (corresponding unnamed define_insn): Delete.
15181 * config/rs6000/rs6000.opt (mno-string): New.
15182 (mstring): Replace by deprecation warning stub.
15183 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
15185 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15187 * regrename.c (regrename_do_replace): If replacing the same
15188 reg multiple times, try to reuse last created gen_raw_REG.
15191 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
15192 main to workaround a bug in GDB.
15194 2018-01-12 Tom de Vries <tom@codesourcery.com>
15197 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
15199 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
15201 PR rtl-optimization/80481
15202 * ira-color.c (get_cap_member): New function.
15203 (allocnos_conflict_by_live_ranges_p): Use it.
15204 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
15205 (setup_slot_coalesced_allocno_live_ranges): Ditto.
15207 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
15210 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
15211 (*saddl_se_1): Ditto.
15212 (*ssubsi_1): Ditto.
15213 (*ssubl_se_1): Ditto.
15215 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15217 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
15218 rather than wi::to_widest for DR_INITs.
15219 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
15220 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
15221 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
15223 (vect_analyze_group_access_1): Note that here.
15225 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15227 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
15228 polynomial type sizes.
15230 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
15232 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
15233 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
15234 (gimple_add_tmp_var): Likewise.
15236 2018-01-12 Martin Liska <mliska@suse.cz>
15238 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
15239 (gimple_alloc_sizes): Likewise.
15240 (dump_gimple_statistics): Use PRIu64 in printf format.
15241 * gimple.h: Change uint64_t to int.
15243 2018-01-12 Martin Liska <mliska@suse.cz>
15245 * tree-core.h: Use uint64_t instead of int.
15246 * tree.c (tree_node_counts): Likewise.
15247 (tree_node_sizes): Likewise.
15248 (dump_tree_statistics): Use PRIu64 in printf format.
15250 2018-01-12 Martin Liska <mliska@suse.cz>
15252 * Makefile.in: As qsort_chk is implemented in vec.c, add
15253 vec.o to linkage of gencfn-macros.
15254 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
15255 passing the info to record_node_allocation_statistics.
15256 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
15258 * ggc-common.c (struct ggc_usage): Add operator== and use
15259 it in operator< and compare function.
15260 * mem-stats.h (struct mem_usage): Likewise.
15261 * vec.c (struct vec_usage): Remove operator< and compare
15262 function. Can be simply inherited.
15264 2018-01-12 Martin Jambor <mjambor@suse.cz>
15267 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
15268 * tree-ssa-math-opts.c: Include domwalk.h.
15269 (convert_mult_to_fma_1): New function.
15270 (fma_transformation_info): New type.
15271 (fma_deferring_state): Likewise.
15272 (cancel_fma_deferring): New function.
15273 (result_of_phi): Likewise.
15274 (last_fma_candidate_feeds_initial_phi): Likewise.
15275 (convert_mult_to_fma): Added deferring logic, split actual
15276 transformation to convert_mult_to_fma_1.
15277 (math_opts_dom_walker): New type.
15278 (math_opts_dom_walker::after_dom_children): New method, body moved
15279 here from pass_optimize_widening_mul::execute, added deferring logic
15281 (pass_optimize_widening_mul::execute): Moved most of code to
15282 math_opts_dom_walker::after_dom_children.
15283 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
15284 * config/i386/i386.c (ix86_option_override_internal): Added
15285 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
15287 2018-01-12 Richard Biener <rguenther@suse.de>
15290 * dwarf2out.c (gen_variable_die): Do not reset old_die for
15291 inline instance vars.
15293 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
15296 * config/rx/rx.c (rx_is_restricted_memory_address):
15297 Handle SUBREG case.
15299 2018-01-12 Richard Biener <rguenther@suse.de>
15301 PR tree-optimization/80846
15302 * target.def (split_reduction): New target hook.
15303 * targhooks.c (default_split_reduction): New function.
15304 * targhooks.h (default_split_reduction): Declare.
15305 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
15306 target requests first reduce vectors by combining low and high
15308 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
15309 (get_vectype_for_scalar_type_and_size): Export.
15310 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
15311 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
15312 * doc/tm.texi: Regenerate.
15313 * config/i386/i386.c (ix86_split_reduction): Implement
15314 TARGET_VECTORIZE_SPLIT_REDUCTION.
15316 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15319 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
15320 in PIC mode except for TARGET_VXWORKS_RTP.
15321 * config/sparc/sparc.c: Include cfgrtl.h.
15322 (TARGET_INIT_PIC_REG): Define.
15323 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
15324 (sparc_pic_register_p): New predicate.
15325 (sparc_legitimate_address_p): Use it.
15326 (sparc_legitimize_pic_address): Likewise.
15327 (sparc_delegitimize_address): Likewise.
15328 (sparc_mode_dependent_address_p): Likewise.
15329 (gen_load_pcrel_sym): Remove 4th parameter.
15330 (load_got_register): Adjust call to above. Remove obsolete stuff.
15331 (sparc_expand_prologue): Do not call load_got_register here.
15332 (sparc_flat_expand_prologue): Likewise.
15333 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
15334 (sparc_use_pseudo_pic_reg): New function.
15335 (sparc_init_pic_reg): Likewise.
15336 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15337 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
15339 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
15341 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
15342 Add item for branch_cost.
15344 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15346 PR rtl-optimization/83565
15347 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
15348 not extend the result to a larger mode for rotate operations.
15349 (num_sign_bit_copies1): Likewise.
15351 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15354 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
15356 Use values-Xc.o for -pedantic.
15357 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
15359 2018-01-12 Martin Liska <mliska@suse.cz>
15362 * ipa-devirt.c (final_warning_record::grow_type_warnings):
15364 (possible_polymorphic_call_targets): Use it.
15365 (ipa_devirt): Likewise.
15367 2018-01-12 Martin Liska <mliska@suse.cz>
15369 * profile-count.h (enum profile_quality): Use 0 as invalid
15370 enum value of profile_quality.
15372 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
15374 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
15375 -mext-string options.
15377 2018-01-12 Richard Biener <rguenther@suse.de>
15379 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
15380 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
15381 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
15383 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
15385 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15387 * configure.ac (--with-long-double-format): Add support for the
15388 configuration option to change the default long double format on
15390 * config.gcc (powerpc*-linux*-*): Likewise.
15391 * configure: Regenerate.
15392 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
15393 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
15394 used without modification.
15396 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15398 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
15399 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
15400 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
15401 MISC_BUILTIN_SPEC_BARRIER.
15402 (rs6000_init_builtins): Likewise.
15403 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
15405 (speculation_barrier): New define_insn.
15406 * doc/extend.texi: Document __builtin_speculation_barrier.
15408 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15411 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
15412 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
15413 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
15415 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
15416 integral modes instead of "ss" and "sd".
15417 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
15418 vectors with 32-bit and 64-bit elements.
15419 (vecdupssescalarmodesuffix): New mode attribute.
15420 (vec_dup<mode>): Use it.
15422 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
15425 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
15426 frame if argument is passed on stack.
15428 2018-01-11 Jakub Jelinek <jakub@redhat.com>
15431 * ree.c (combine_reaching_defs): Optimize also
15432 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
15433 reg2=any_extend(exp); reg1=reg2;, formatting fix.
15435 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15437 PR middle-end/83189
15438 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
15440 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
15442 PR middle-end/83718
15443 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
15444 after they are computed.
15446 2018-01-11 Bin Cheng <bin.cheng@arm.com>
15448 PR tree-optimization/83695
15449 * gimple-loop-linterchange.cc
15450 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
15451 reset cached scev information after interchange.
15452 (pass_linterchange::execute): Remove call to scev_reset_htab.
15454 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15456 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
15457 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
15458 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
15459 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
15460 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
15461 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
15462 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
15463 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
15464 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
15465 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
15466 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
15467 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
15468 (V_lane_reg): Likewise.
15469 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
15471 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
15472 (vfmal_lane_low<mode>_intrinsic,
15473 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
15474 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
15475 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
15476 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
15477 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
15478 vfmsl_lane_high<mode>_intrinsic): New define_insns.
15480 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15482 * config/arm/arm-cpus.in (fp16fml): New feature.
15483 (ALL_SIMD): Add fp16fml.
15484 (armv8.2-a): Add fp16fml as an option.
15485 (armv8.3-a): Likewise.
15486 (armv8.4-a): Add fp16fml as part of fp16.
15487 * config/arm/arm.h (TARGET_FP16FML): Define.
15488 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
15490 * config/arm/arm-modes.def (V2HF): Define.
15491 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
15492 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
15493 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
15494 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
15495 vfmsl_low, vfmsl_high): New set of builtins.
15496 * config/arm/iterators.md (PLUSMINUS): New code iterator.
15497 (vfml_op): New code attribute.
15498 (VFMLHALVES): New int iterator.
15499 (VFML, VFMLSEL): New mode attributes.
15500 (V_reg): Define mapping for V2HF.
15501 (V_hi, V_lo): New mode attributes.
15502 (VF_constraint): Likewise.
15503 (vfml_half, vfml_half_selector): New int attributes.
15504 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
15506 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
15507 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
15509 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
15510 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
15511 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
15512 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
15514 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
15515 Document new effective target and option set.
15517 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15519 * config/arm/arm-cpus.in (armv8_4): New feature.
15520 (ARMv8_4a): New fgroup.
15521 (armv8.4-a): New arch.
15522 * config/arm/arm-tables.opt: Regenerate.
15523 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
15524 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
15525 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15526 Add matching rules for -march=armv8.4-a and extensions.
15527 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15529 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15532 * config/rx/rx.md (BW): New mode attribute.
15533 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15535 2018-01-11 Richard Biener <rguenther@suse.de>
15537 PR tree-optimization/83435
15538 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15539 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15540 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15542 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15543 Alan Hayward <alan.hayward@arm.com>
15544 David Sherwood <david.sherwood@arm.com>
15546 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15548 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15549 (aarch64_print_address_internal): Use it to check for a zero offset.
15551 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15552 Alan Hayward <alan.hayward@arm.com>
15553 David Sherwood <david.sherwood@arm.com>
15555 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15556 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15557 Return a poly_int64 rather than a HOST_WIDE_INT.
15558 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15559 rather than a HOST_WIDE_INT.
15560 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15561 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15562 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15563 final_offset from HOST_WIDE_INT to poly_int64.
15564 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15565 to_constant when getting the number of units in an Advanced SIMD
15567 (aarch64_builtin_vectorized_function): Check for a constant number
15569 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15571 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15572 attribute instead of GET_MODE_NUNITS.
15573 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15574 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15575 GET_MODE_SIZE for fixed-size registers.
15576 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15577 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15578 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15579 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15580 (aarch64_print_operand, aarch64_print_address_internal)
15581 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15582 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15583 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15584 Handle polynomial GET_MODE_SIZE.
15585 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15586 wider than SImode without modification.
15587 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15588 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15589 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15590 passing and returning SVE modes.
15591 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15592 rather than GEN_INT.
15593 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15594 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15595 (aarch64_allocate_and_probe_stack_space): Likewise.
15596 (aarch64_layout_frame): Cope with polynomial offsets.
15597 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15598 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15599 polynomial offsets.
15600 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15601 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15602 poly_int64 rather than a HOST_WIDE_INT.
15603 (aarch64_get_separate_components, aarch64_process_components)
15604 (aarch64_expand_prologue, aarch64_expand_epilogue)
15605 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15606 (aarch64_anchor_offset): New function, split out from...
15607 (aarch64_legitimize_address): ...here.
15608 (aarch64_builtin_vectorization_cost): Handle polynomial
15609 TYPE_VECTOR_SUBPARTS.
15610 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15612 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15613 number of elements from the PARALLEL rather than the mode.
15614 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15615 rather than GET_MODE_BITSIZE.
15616 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15617 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15618 (aarch64_expand_vec_perm_const_1): Handle polynomial
15619 d->perm.length () and d->perm elements.
15620 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15621 Apply to_constant to d->perm elements.
15622 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15623 polynomial CONST_VECTOR_NUNITS.
15624 (aarch64_move_pointer): Take amount as a poly_int64 rather
15626 (aarch64_progress_pointer): Avoid temporary variable.
15627 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15628 the mode attribute instead of GET_MODE.
15630 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15631 Alan Hayward <alan.hayward@arm.com>
15632 David Sherwood <david.sherwood@arm.com>
15634 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15635 x exists before using it.
15636 (aarch64_add_constant_internal): Rename to...
15637 (aarch64_add_offset_1): ...this. Replace regnum with separate
15638 src and dest rtxes. Handle the case in which they're different,
15639 including when the offset is zero. Replace scratchreg with an rtx.
15640 Use 2 additions if there is no spare register into which we can
15641 move a 16-bit constant.
15642 (aarch64_add_constant): Delete.
15643 (aarch64_add_offset): Replace reg with separate src and dest
15644 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15645 Use aarch64_add_offset_1.
15646 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15647 an rtx rather than an int. Take the delta as a poly_int64
15648 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15649 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15650 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15651 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15652 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15653 and aarch64_add_sp.
15654 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15655 aarch64_add_constant.
15657 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15659 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15660 Use scalar_float_mode.
15662 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15664 * config/aarch64/aarch64-simd.md
15665 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15666 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15667 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15668 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15669 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15670 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15671 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15672 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15673 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15674 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15676 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15679 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15680 targ_options->x_arm_arch_string is non NULL.
15682 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15684 * config/aarch64/aarch64.h
15685 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15687 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15690 * expmed.c (emit_store_flag_force): Swap if const op0
15691 and change VOIDmode to mode of op0.
15693 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15695 PR rtl-optimization/83761
15696 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15697 than bytes to mode_for_size.
15699 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15701 PR middle-end/83189
15702 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15703 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15706 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15708 PR middle-end/83575
15709 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15710 when in layout mode.
15711 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15712 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15715 2018-01-10 Michael Collison <michael.collison@arm.com>
15717 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15718 * config/aarch64/aarch64-option-extension.def: Add
15719 AARCH64_OPT_EXTENSION of 'fp16fml'.
15720 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15721 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15722 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15723 * config/aarch64/constraints.md (Ui7): New constraint.
15724 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15725 (VFMLA_SEL_W): Ditto.
15728 (VFMLA16_LOW): New int iterator.
15729 (VFMLA16_HIGH): Ditto.
15730 (UNSPEC_FMLAL): New unspec.
15731 (UNSPEC_FMLSL): Ditto.
15732 (UNSPEC_FMLAL2): Ditto.
15733 (UNSPEC_FMLSL2): Ditto.
15734 (f16mac): New code attribute.
15735 * config/aarch64/aarch64-simd-builtins.def
15736 (aarch64_fmlal_lowv2sf): Ditto.
15737 (aarch64_fmlsl_lowv2sf): Ditto.
15738 (aarch64_fmlalq_lowv4sf): Ditto.
15739 (aarch64_fmlslq_lowv4sf): Ditto.
15740 (aarch64_fmlal_highv2sf): Ditto.
15741 (aarch64_fmlsl_highv2sf): Ditto.
15742 (aarch64_fmlalq_highv4sf): Ditto.
15743 (aarch64_fmlslq_highv4sf): Ditto.
15744 (aarch64_fmlal_lane_lowv2sf): Ditto.
15745 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15746 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15747 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15748 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15749 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15750 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15751 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15752 (aarch64_fmlal_lane_highv2sf): Ditto.
15753 (aarch64_fmlsl_lane_highv2sf): Ditto.
15754 (aarch64_fmlal_laneq_highv2sf): Ditto.
15755 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15756 (aarch64_fmlalq_lane_highv4sf): Ditto.
15757 (aarch64_fmlsl_lane_highv4sf): Ditto.
15758 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15759 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15760 * config/aarch64/aarch64-simd.md:
15761 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15762 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15763 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15764 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15765 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15766 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15767 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15768 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15769 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15770 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15771 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15772 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15773 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15774 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15775 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15776 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15777 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15778 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15779 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15780 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15781 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15782 (vfmlsl_low_u32): Ditto.
15783 (vfmlalq_low_u32): Ditto.
15784 (vfmlslq_low_u32): Ditto.
15785 (vfmlal_high_u32): Ditto.
15786 (vfmlsl_high_u32): Ditto.
15787 (vfmlalq_high_u32): Ditto.
15788 (vfmlslq_high_u32): Ditto.
15789 (vfmlal_lane_low_u32): Ditto.
15790 (vfmlsl_lane_low_u32): Ditto.
15791 (vfmlal_laneq_low_u32): Ditto.
15792 (vfmlsl_laneq_low_u32): Ditto.
15793 (vfmlalq_lane_low_u32): Ditto.
15794 (vfmlslq_lane_low_u32): Ditto.
15795 (vfmlalq_laneq_low_u32): Ditto.
15796 (vfmlslq_laneq_low_u32): Ditto.
15797 (vfmlal_lane_high_u32): Ditto.
15798 (vfmlsl_lane_high_u32): Ditto.
15799 (vfmlal_laneq_high_u32): Ditto.
15800 (vfmlsl_laneq_high_u32): Ditto.
15801 (vfmlalq_lane_high_u32): Ditto.
15802 (vfmlslq_lane_high_u32): Ditto.
15803 (vfmlalq_laneq_high_u32): Ditto.
15804 (vfmlslq_laneq_high_u32): Ditto.
15805 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15806 (AARCH64_FL_FOR_ARCH8_4): New.
15807 (AARCH64_ISA_F16FML): New ISA flag.
15808 (TARGET_F16FML): New feature flag for fp16fml.
15809 (doc/invoke.texi): Document new fp16fml option.
15811 2018-01-10 Michael Collison <michael.collison@arm.com>
15813 * config/aarch64/aarch64-builtins.c:
15814 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15815 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15816 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15817 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15818 (AARCH64_ISA_SHA3): New ISA flag.
15819 (TARGET_SHA3): New feature flag for sha3.
15820 * config/aarch64/iterators.md (sha512_op): New int attribute.
15821 (CRYPTO_SHA512): New int iterator.
15822 (UNSPEC_SHA512H): New unspec.
15823 (UNSPEC_SHA512H2): Ditto.
15824 (UNSPEC_SHA512SU0): Ditto.
15825 (UNSPEC_SHA512SU1): Ditto.
15826 * config/aarch64/aarch64-simd-builtins.def
15827 (aarch64_crypto_sha512hqv2di): New builtin.
15828 (aarch64_crypto_sha512h2qv2di): Ditto.
15829 (aarch64_crypto_sha512su0qv2di): Ditto.
15830 (aarch64_crypto_sha512su1qv2di): Ditto.
15831 (aarch64_eor3qv8hi): Ditto.
15832 (aarch64_rax1qv2di): Ditto.
15833 (aarch64_xarqv2di): Ditto.
15834 (aarch64_bcaxqv8hi): Ditto.
15835 * config/aarch64/aarch64-simd.md:
15836 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15837 (aarch64_crypto_sha512su0qv2di): Ditto.
15838 (aarch64_crypto_sha512su1qv2di): Ditto.
15839 (aarch64_eor3qv8hi): Ditto.
15840 (aarch64_rax1qv2di): Ditto.
15841 (aarch64_xarqv2di): Ditto.
15842 (aarch64_bcaxqv8hi): Ditto.
15843 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15844 (vsha512h2q_u64): Ditto.
15845 (vsha512su0q_u64): Ditto.
15846 (vsha512su1q_u64): Ditto.
15847 (veor3q_u16): Ditto.
15848 (vrax1q_u64): Ditto.
15849 (vxarq_u64): Ditto.
15850 (vbcaxq_u16): Ditto.
15851 * config/arm/types.md (crypto_sha512): New type attribute.
15852 (crypto_sha3): Ditto.
15853 (doc/invoke.texi): Document new sha3 option.
15855 2018-01-10 Michael Collison <michael.collison@arm.com>
15857 * config/aarch64/aarch64-builtins.c:
15858 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15859 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15860 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15861 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15862 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15863 (AARCH64_ISA_SM4): New ISA flag.
15864 (TARGET_SM4): New feature flag for sm4.
15865 * config/aarch64/aarch64-simd-builtins.def
15866 (aarch64_sm3ss1qv4si): Ditto.
15867 (aarch64_sm3tt1aq4si): Ditto.
15868 (aarch64_sm3tt1bq4si): Ditto.
15869 (aarch64_sm3tt2aq4si): Ditto.
15870 (aarch64_sm3tt2bq4si): Ditto.
15871 (aarch64_sm3partw1qv4si): Ditto.
15872 (aarch64_sm3partw2qv4si): Ditto.
15873 (aarch64_sm4eqv4si): Ditto.
15874 (aarch64_sm4ekeyqv4si): Ditto.
15875 * config/aarch64/aarch64-simd.md:
15876 (aarch64_sm3ss1qv4si): Ditto.
15877 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15878 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15879 (aarch64_sm4eqv4si): Ditto.
15880 (aarch64_sm4ekeyqv4si): Ditto.
15881 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15882 (sm3part_op): Ditto.
15883 (CRYPTO_SM3TT): Ditto.
15884 (CRYPTO_SM3PART): Ditto.
15885 (UNSPEC_SM3SS1): New unspec.
15886 (UNSPEC_SM3TT1A): Ditto.
15887 (UNSPEC_SM3TT1B): Ditto.
15888 (UNSPEC_SM3TT2A): Ditto.
15889 (UNSPEC_SM3TT2B): Ditto.
15890 (UNSPEC_SM3PARTW1): Ditto.
15891 (UNSPEC_SM3PARTW2): Ditto.
15892 (UNSPEC_SM4E): Ditto.
15893 (UNSPEC_SM4EKEY): Ditto.
15894 * config/aarch64/constraints.md (Ui2): New constraint.
15895 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15896 * config/arm/types.md (crypto_sm3): New type attribute.
15897 (crypto_sm4): Ditto.
15898 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15899 (vsm3tt1aq_u32): Ditto.
15900 (vsm3tt1bq_u32): Ditto.
15901 (vsm3tt2aq_u32): Ditto.
15902 (vsm3tt2bq_u32): Ditto.
15903 (vsm3partw1q_u32): Ditto.
15904 (vsm3partw2q_u32): Ditto.
15905 (vsm4eq_u32): Ditto.
15906 (vsm4ekeyq_u32): Ditto.
15907 (doc/invoke.texi): Document new sm4 option.
15909 2018-01-10 Michael Collison <michael.collison@arm.com>
15911 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15912 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15913 (AARCH64_FL_FOR_ARCH8_4): New.
15914 (AARCH64_FL_V8_4): New flag.
15915 (doc/invoke.texi): Document new armv8.4-a option.
15917 2018-01-10 Michael Collison <michael.collison@arm.com>
15919 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15920 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15921 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15922 * config/aarch64/aarch64-option-extension.def: Add
15923 AARCH64_OPT_EXTENSION of 'sha2'.
15924 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15925 (crypto): Disable sha2 and aes if crypto disabled.
15926 (crypto): Enable aes and sha2 if enabled.
15927 (simd): Disable sha2 and aes if simd disabled.
15928 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15930 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15931 (TARGET_SHA2): New feature flag for sha2.
15932 (TARGET_AES): New feature flag for aes.
15933 * config/aarch64/aarch64-simd.md:
15934 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15935 conditional on TARGET_AES.
15936 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15937 (aarch64_crypto_sha1hsi): Make pattern conditional
15939 (aarch64_crypto_sha1hv4si): Ditto.
15940 (aarch64_be_crypto_sha1hv4si): Ditto.
15941 (aarch64_crypto_sha1su1v4si): Ditto.
15942 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15943 (aarch64_crypto_sha1su0v4si): Ditto.
15944 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15945 (aarch64_crypto_sha256su0v4si): Ditto.
15946 (aarch64_crypto_sha256su1v4si): Ditto.
15947 (doc/invoke.texi): Document new aes and sha2 options.
15949 2018-01-10 Martin Sebor <msebor@redhat.com>
15951 PR tree-optimization/83781
15952 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15955 2018-01-11 Martin Sebor <msebor@gmail.com>
15956 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15958 PR tree-optimization/83501
15959 PR tree-optimization/81703
15961 * tree-ssa-strlen.c (get_string_cst): Rename...
15962 (get_string_len): ...to this. Handle global constants.
15963 (handle_char_store): Adjust.
15965 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15966 Jim Wilson <jimw@sifive.com>
15968 * config/riscv/riscv-protos.h (riscv_output_return): New.
15969 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15970 (riscv_attribute_table, riscv_output_return),
15971 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15972 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15973 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15975 (riscv_expand_prologue): Add early return for naked function.
15976 (riscv_expand_epilogue): Likewise.
15977 (riscv_function_ok_for_sibcall): Return false for naked function.
15978 (riscv_set_current_function): New.
15979 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15980 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15981 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15982 * doc/extend.texi (RISC-V Function Attributes): New.
15984 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15986 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15987 check for 128-bit long double before checking TCmode.
15988 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15989 128-bit long doubles before checking TFmode or TCmode.
15990 (FLOAT128_IBM_P): Likewise.
15992 2018-01-10 Martin Sebor <msebor@redhat.com>
15994 PR tree-optimization/83671
15995 * builtins.c (c_strlen): Unconditionally return zero for the empty
15997 Use -Warray-bounds for warnings.
15998 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15999 for non-constant array indices with COMPONENT_REF, arrays of
16000 arrays, and pointers to arrays.
16001 (gimple_fold_builtin_strlen): Determine and set length range for
16002 non-constant character arrays.
16004 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16006 PR middle-end/81897
16007 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16010 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16012 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16014 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16017 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16018 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16019 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16020 indexed_or_indirect_operand predicate.
16021 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16022 (*vsx_le_perm_load_v8hi): Likewise.
16023 (*vsx_le_perm_load_v16qi): Likewise.
16024 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16025 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16026 (*vsx_le_perm_store_v8hi): Likewise.
16027 (*vsx_le_perm_store_v16qi): Likewise.
16028 (eight unnamed splitters): Likewise.
16030 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16032 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16033 * config/rs6000/emmintrin.h: Likewise.
16034 * config/rs6000/mmintrin.h: Likewise.
16035 * config/rs6000/xmmintrin.h: Likewise.
16037 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16040 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16042 * tree.c (tree_nop_conversion): Return true for location wrapper
16044 (maybe_wrap_with_location): New function.
16045 (selftest::check_strip_nops): New function.
16046 (selftest::test_location_wrappers): New function.
16047 (selftest::tree_c_tests): Call it.
16048 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16049 (maybe_wrap_with_location): New decl.
16050 (EXPR_LOCATION_WRAPPER_P): New macro.
16051 (location_wrapper_p): New inline function.
16052 (tree_strip_any_location_wrapper): New inline function.
16054 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16057 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16058 stack_realign_offset for the largest alignment of stack slot
16060 (ix86_find_max_used_stack_alignment): New function.
16061 (ix86_finalize_stack_frame_flags): Use it. Set
16062 max_used_stack_alignment if we don't realign stack.
16063 * config/i386/i386.h (machine_function): Add
16064 max_used_stack_alignment.
16066 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16068 * config/arm/arm.opt (-mbranch-cost): New option.
16069 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16072 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16075 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16076 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16078 2018-01-10 Richard Biener <rguenther@suse.de>
16081 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16082 early out so it also covers the case where we have a non-NULL
16085 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16087 PR tree-optimization/83753
16088 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16089 for non-strided grouped accesses if the number of elements is 1.
16091 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16094 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16095 * i386.h (TARGET_USE_GATHER): Define.
16096 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16098 2018-01-10 Martin Liska <mliska@suse.cz>
16101 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16102 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16104 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16105 CLEANUP_NO_PARTITIONING is not set.
16107 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16109 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16110 for vectors, as a partial revert of r254296.
16111 * rtl.h (const_vec_p): Delete.
16112 (const_vec_duplicate_p): Don't test for vector CONSTs.
16113 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16114 * expmed.c (make_tree): Likewise.
16117 * common.md (E, F): Use CONSTANT_P instead of checking for
16119 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16120 checking for CONST_VECTOR.
16122 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16124 PR middle-end/83575
16125 * predict.c (force_edge_cold): Handle in more sane way edges
16126 with no prediction.
16128 2018-01-09 Carl Love <cel@us.ibm.com>
16130 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16132 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16133 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16134 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16135 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16136 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16137 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16138 * config/rs6000/rs6000-protos.h: Add extern defition for
16139 rs6000_generate_float2_double_code.
16140 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16142 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16143 (float2_v2df): Add define_expand.
16145 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16148 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16149 op_mode in the force_to_mode call.
16151 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16153 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
16154 instead of checking each element individually.
16155 (aarch64_evpc_uzp): Likewise.
16156 (aarch64_evpc_zip): Likewise.
16157 (aarch64_evpc_ext): Likewise.
16158 (aarch64_evpc_rev): Likewise.
16159 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
16160 instead of checking each element individually. Return true without
16162 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
16163 whether all selected elements come from the same input, instead of
16164 checking each element individually. Remove calls to gen_rtx_REG,
16165 start_sequence and end_sequence and instead assert that no rtl is
16168 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16170 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
16171 order of HIGH and CONST checks.
16173 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
16175 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
16176 if the destination isn't an SSA_NAME.
16178 2018-01-09 Richard Biener <rguenther@suse.de>
16180 PR tree-optimization/83668
16181 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
16183 (canonicalize_loop_form): ... here, renamed from ...
16184 (canonicalize_loop_closed_ssa_form): ... this and amended to
16185 swap successor edges for loop exit blocks to make us use
16186 the RPO order we need for initial schedule generation.
16188 2018-01-09 Joseph Myers <joseph@codesourcery.com>
16190 PR tree-optimization/64811
16191 * match.pd: When optimizing comparisons with Inf, avoid
16192 introducing or losing exceptions from comparisons with NaN.
16194 2018-01-09 Martin Liska <mliska@suse.cz>
16197 * asan.c (shadow_mem_size): Add gcc_assert.
16199 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
16201 Don't save registers in main().
16204 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
16205 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
16206 * config/avr/avr.c (avr_set_current_function): Don't error if
16207 naked, OS_task or OS_main are specified at the same time.
16208 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
16210 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
16212 * common/config/avr/avr-common.c (avr_option_optimization_table):
16213 Switch on -mmain-is-OS_task for optimizing compilations.
16215 2018-01-09 Richard Biener <rguenther@suse.de>
16217 PR tree-optimization/83572
16218 * graphite.c: Include cfganal.h.
16219 (graphite_transform_loops): Connect infinite loops to exit
16220 and remove fake edges at the end.
16222 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16224 * ipa-inline.c (edge_badness): Revert accidental checkin.
16226 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16229 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
16230 symbols; not inline clones.
16232 2018-01-09 Jakub Jelinek <jakub@redhat.com>
16235 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
16236 hard registers. Formatting fixes.
16238 PR preprocessor/83722
16239 * gcc.c (try_generate_repro): Pass
16240 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
16241 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
16244 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
16245 Kito Cheng <kito.cheng@gmail.com>
16247 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
16248 (riscv_leaf_function_p): Delete.
16249 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
16251 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16253 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
16255 (do_ifelse): New function.
16256 (do_isel): New function.
16257 (do_sub3): New function.
16258 (do_add3): New function.
16259 (do_load_mask_compare): New function.
16260 (do_overlap_load_compare): New function.
16261 (expand_compare_loop): New function.
16262 (expand_block_compare): Call expand_compare_loop() when appropriate.
16263 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
16264 option description.
16265 (-mblock-compare-inline-loop-limit): New option.
16267 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16270 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
16271 Reverse order of second and third operands in first alternative.
16272 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
16273 of first and second elements in UNSPEC_VPERMR vector.
16274 (altivec_expand_vec_perm_le): Likewise.
16276 2018-01-08 Jeff Law <law@redhat.com>
16278 PR rtl-optimizatin/81308
16279 * tree-switch-conversion.c (cfg_altered): New file scoped static.
16280 (process_switch): If group_case_labels makes a change, then set
16282 (pass_convert_switch::execute): If a switch is converted, then
16283 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
16285 PR rtl-optimization/81308
16286 * recog.c (split_all_insns): Conditionally cleanup the CFG after
16289 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
16291 PR target/83663 - Revert r255946
16292 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
16293 generation for cases where splatting a value is not useful.
16294 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
16295 across a vec_duplicate and a paradoxical subreg forming a vector
16296 mode to a vec_concat.
16298 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16300 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
16301 -march=armv8.3-a variants.
16302 * config/arm/t-multilib: Likewise.
16303 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
16305 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16307 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
16309 (cceq_ior_compare_complement): Give it a name so I can use it, and
16310 change boolean_or_operator predicate to boolean_operator so it can
16311 be used to generate a crand.
16312 (eqne): New code iterator.
16313 (bd/bd_neg): New code_attrs.
16314 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
16315 a single define_insn.
16316 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
16317 decrement (bdnzt/bdnzf/bdzt/bdzf).
16318 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
16319 with the new names of the branch decrement patterns, and added the
16320 names of the branch decrement conditional patterns.
16322 2018-01-08 Richard Biener <rguenther@suse.de>
16324 PR tree-optimization/83563
16325 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
16328 2018-01-08 Richard Biener <rguenther@suse.de>
16330 PR middle-end/83713
16331 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
16333 2018-01-08 Richard Biener <rguenther@suse.de>
16335 PR tree-optimization/83685
16336 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
16337 references to abnormals.
16339 2018-01-08 Richard Biener <rguenther@suse.de>
16342 * dwarf2out.c (output_indirect_strings): Handle empty
16343 skeleton_debug_str_hash.
16344 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
16346 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16348 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
16349 (emit_store_direct): Likewise.
16350 (arc_trampoline_adjust_address): Likewise.
16351 (arc_asm_trampoline_template): New function.
16352 (arc_initialize_trampoline): Use asm_trampoline_template.
16353 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
16354 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
16355 * config/arc/arc.md (flush_icache): Delete pattern.
16357 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
16359 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
16360 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
16363 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16366 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
16367 by not USED_FOR_TARGET.
16368 (make_pass_resolve_sw_modes): Likewise.
16370 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
16372 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
16375 2018-01-08 Richard Biener <rguenther@suse.de>
16377 PR middle-end/83580
16378 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
16380 2018-01-08 Richard Biener <rguenther@suse.de>
16382 PR middle-end/83517
16383 * match.pd ((t * 2) / 2) -> t): Add missing :c.
16385 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
16387 PR middle-end/81897
16388 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
16389 basic blocks with a small number of successors.
16390 (convert_control_dep_chain_into_preds): Improve handling of
16392 (dump_predicates): Split apart into...
16393 (dump_pred_chain): ...here...
16394 (dump_pred_info): ...and here.
16395 (can_one_predicate_be_invalidated_p): Add debugging printfs.
16396 (can_chain_union_be_invalidated_p): Improve check for invalidation
16398 (uninit_uses_cannot_happen): Avoid unnecessary if
16399 convert_control_dep_chain_into_preds yielded nothing.
16401 2018-01-06 Martin Sebor <msebor@redhat.com>
16403 PR tree-optimization/83640
16404 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
16405 subtracting negative offset from size.
16406 (builtin_access::overlap): Adjust offset bounds of the access to fall
16407 within the size of the object if possible.
16409 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
16411 PR rtl-optimization/83699
16412 * expmed.c (extract_bit_field_1): Restrict the vector usage of
16413 extract_bit_field_as_subreg to cases in which the extracted
16414 value is also a vector.
16416 * lra-constraints.c (process_alt_operands): Test for the equivalence
16417 substitutions when detecting a possible reload cycle.
16419 2018-01-06 Jakub Jelinek <jakub@redhat.com>
16422 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
16423 by default if flag_selective_schedling{,2}. Formatting fixes.
16425 PR rtl-optimization/83682
16426 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
16427 if it has non-VECTOR_MODE element mode.
16428 (vec_duplicate_p): Likewise.
16430 PR middle-end/83694
16431 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
16432 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
16434 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16437 * config/i386/i386-builtin.def
16438 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
16439 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
16440 Require also OPTION_MASK_ISA_AVX512F in addition to
16441 OPTION_MASK_ISA_GFNI.
16442 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
16443 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
16444 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
16445 to OPTION_MASK_ISA_GFNI.
16446 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
16447 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
16448 OPTION_MASK_ISA_AVX512BW.
16449 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
16450 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
16451 addition to OPTION_MASK_ISA_GFNI.
16452 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
16453 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
16454 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
16455 to OPTION_MASK_ISA_GFNI.
16456 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
16457 a requirement for all ISAs rather than any of them with a few
16459 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
16461 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
16462 bitmasks to be enabled with 3 exceptions, instead of requiring any
16463 enabled ISA with lots of exceptions.
16464 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
16465 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
16466 Change avx512bw in isa attribute to avx512f.
16467 * config/i386/sgxintrin.h: Add license boilerplate.
16468 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
16469 to __AVX512F__ and __AVX512VL to __AVX512VL__.
16470 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
16471 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
16473 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
16474 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
16475 temporarily sse2 rather than sse if not enabled already.
16478 * config/i386/sse.md (VI248_VLBW): Rename to ...
16479 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
16480 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
16481 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
16482 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
16483 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
16484 mode iterator instead of VI248_VLBW.
16486 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
16488 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
16489 (record_modified): Skip clobbers; add debug output.
16490 (param_change_prob): Use sreal frequencies.
16492 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16494 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
16495 punt for user-aligned variables.
16497 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
16499 * tree-chrec.c (chrec_contains_symbols): Return true for
16502 2018-01-05 Sudakshina Das <sudi.das@arm.com>
16505 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
16506 of (x|y) == x for BICS pattern.
16508 2018-01-05 Jakub Jelinek <jakub@redhat.com>
16510 PR tree-optimization/83605
16511 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
16512 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
16515 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
16517 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
16518 * config/epiphany/rtems.h: New file.
16520 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16521 Uros Bizjak <ubizjak@gmail.com>
16524 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
16525 QIreg_operand instead of register_operand predicate.
16526 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16527 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16528 comments instead of -fmitigate[-_]rop.
16530 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16533 * cgraphunit.c (symbol_table::compile): Switch to text_section
16534 before calling assembly_start debug hook.
16535 * run-rtl-passes.c (run_rtl_passes): Likewise.
16538 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16540 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16541 range_int_cst_p rather than !symbolic_range_p before calling
16542 extract_range_from_multiplicative_op_1.
16544 2018-01-04 Jeff Law <law@redhat.com>
16546 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16547 redundant test in assertion.
16549 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16551 * doc/rtl.texi: Document machine_mode wrapper classes.
16553 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16555 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16556 using tree_to_uhwi.
16558 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16560 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16561 the VEC_PERM_EXPR fold to fail.
16563 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16566 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16567 to switched_sections.
16569 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16572 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16573 test for d.testing.
16575 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16578 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16579 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16581 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16584 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16585 is BLKmode and bitpos not zero or mode change is needed.
16587 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16590 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16593 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16596 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16597 instead of MULT rtx. Update all corresponding splitters.
16598 (*saddl_se): Ditto.
16599 (*ssub<modesuffix>): Ditto.
16600 (*ssubl_se): Ditto.
16601 (*cmp_sadd_di): Update split patterns.
16602 (*cmp_sadd_si): Ditto.
16603 (*cmp_sadd_sidi): Ditto.
16604 (*cmp_ssub_di): Ditto.
16605 (*cmp_ssub_si): Ditto.
16606 (*cmp_ssub_sidi): Ditto.
16607 * config/alpha/predicates.md (const23_operand): New predicate.
16608 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16609 Look for ASHIFT, not MULT inner operand.
16610 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16612 2018-01-04 Martin Liska <mliska@suse.cz>
16614 PR gcov-profile/83669
16615 * gcov.c (output_intermediate_file): Add version to intermediate
16617 * doc/gcov.texi: Document new field 'version' in intermediate
16618 file format. Fix location of '-k' option of gcov command.
16620 2018-01-04 Martin Liska <mliska@suse.cz>
16623 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16625 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16627 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16629 2018-01-03 Martin Sebor <msebor@redhat.com>
16631 PR tree-optimization/83655
16632 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16633 checking calls with invalid arguments.
16635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16637 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16638 (vectorizable_mask_load_store): Delete.
16639 (vectorizable_call): Return false for masked loads and stores.
16640 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16641 instead of gimple_assign_rhs1.
16642 (vectorizable_load): Handle IFN_MASK_LOAD.
16643 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16647 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16649 (vectorizable_mask_load_store): ...here.
16650 (vectorizable_load): ...and here.
16652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16654 * tree-vect-stmts.c (vect_build_all_ones_mask)
16655 (vect_build_zero_merge_argument): New functions, split out from...
16656 (vectorizable_load): ...here.
16658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16660 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16662 (vectorizable_mask_load_store): ...here.
16663 (vectorizable_store): ...and here.
16665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16667 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16669 (vectorizable_mask_load_store): ...here.
16671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16673 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16674 (vect_model_store_cost): Take a vec_load_store_type instead of a
16676 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16677 (vect_model_store_cost): Take a vec_load_store_type instead of a
16679 (vectorizable_mask_load_store): Update accordingly.
16680 (vectorizable_store): Likewise.
16681 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16685 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16686 IFN_MASK_LOAD calls here rather than...
16687 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16690 Alan Hayward <alan.hayward@arm.com>
16691 David Sherwood <david.sherwood@arm.com>
16693 * expmed.c (extract_bit_field_1): For vector extracts,
16694 fall back to extract_bit_field_as_subreg if vec_extract
16697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16698 Alan Hayward <alan.hayward@arm.com>
16699 David Sherwood <david.sherwood@arm.com>
16701 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16702 they are variable or constant sized.
16703 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16704 slots for constant-sized data.
16706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16707 Alan Hayward <alan.hayward@arm.com>
16708 David Sherwood <david.sherwood@arm.com>
16710 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16711 handling COND_EXPRs with boolean comparisons, try to find a better
16712 basis for the mask type than the boolean itself.
16714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16716 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16717 is calculated and how it can be overridden.
16718 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16719 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16721 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16725 Alan Hayward <alan.hayward@arm.com>
16726 David Sherwood <david.sherwood@arm.com>
16728 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16729 Remove the mode argument.
16730 (aarch64_simd_valid_immediate): Remove the mode and inverse
16732 * config/aarch64/iterators.md (bitsize): New iterator.
16733 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16734 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16735 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16736 aarch64_simd_valid_immediate.
16737 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16738 (aarch64_reg_or_bic_imm): Likewise.
16739 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16740 with an insn_type enum and msl with a modifier_type enum.
16741 Replace element_width with a scalar_mode. Change the shift
16742 to unsigned int. Add constructors for scalar_float_mode and
16743 scalar_int_mode elements.
16744 (aarch64_vect_float_const_representable_p): Delete.
16745 (aarch64_can_const_movi_rtx_p)
16746 (aarch64_simd_scalar_immediate_valid_for_move)
16747 (aarch64_simd_make_constant): Update call to
16748 aarch64_simd_valid_immediate.
16749 (aarch64_advsimd_valid_immediate_hs): New function.
16750 (aarch64_advsimd_valid_immediate): Likewise.
16751 (aarch64_simd_valid_immediate): Remove mode and inverse
16752 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16753 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16754 and aarch64_float_const_representable_p on the result.
16755 (aarch64_output_simd_mov_immediate): Remove mode argument.
16756 Update call to aarch64_simd_valid_immediate and use of
16757 simd_immediate_info.
16758 (aarch64_output_scalar_simd_mov_immediate): Update call
16761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16762 Alan Hayward <alan.hayward@arm.com>
16763 David Sherwood <david.sherwood@arm.com>
16765 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16766 (mode_nunits): Likewise CONST_MODE_NUNITS.
16767 * machmode.def (ADJUST_NUNITS): Document.
16768 * genmodes.c (mode_data::need_nunits_adj): New field.
16769 (blank_mode): Update accordingly.
16770 (adj_nunits): New variable.
16771 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16773 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16774 listed in adj_nunits.
16775 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16776 listed in adj_nunits. Don't emit case statements for such modes.
16777 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16778 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16779 nothing if adj_nunits is nonnull.
16780 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16781 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16782 (emit_mode_fbit): Update use of print_maybe_const_decl.
16783 (emit_move_size): Likewise. Treat the array as non-const
16785 (emit_mode_adjustments): Handle adj_nunits.
16787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16789 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16790 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16791 (VECTOR_MODES): Use it.
16792 (make_vector_modes): Take the prefix as an argument.
16794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16795 Alan Hayward <alan.hayward@arm.com>
16796 David Sherwood <david.sherwood@arm.com>
16798 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16799 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16800 for MODE_VECTOR_BOOL.
16801 * machmode.def (VECTOR_BOOL_MODE): Document.
16802 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16803 (make_vector_bool_mode): New function.
16804 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16806 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16807 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16809 * stor-layout.c (int_mode_for_mode): Likewise.
16810 * tree.c (build_vector_type_for_mode): Likewise.
16811 * varasm.c (output_constant_pool_2): Likewise.
16812 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16813 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16814 for MODE_VECTOR_BOOL.
16815 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16816 of mode class checks.
16817 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16818 instead of a list of mode class checks.
16819 (expand_vector_scalar_condition): Likewise.
16820 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16823 Alan Hayward <alan.hayward@arm.com>
16824 David Sherwood <david.sherwood@arm.com>
16826 * machmode.h (mode_size): Change from unsigned short to
16828 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16829 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16830 or if measurement_type is not polynomial.
16831 (fixed_size_mode::includes_p): Check for constant-sized modes.
16832 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16833 return a poly_uint16 rather than an unsigned short.
16834 (emit_mode_size): Change the type of mode_size from unsigned short
16835 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16836 (emit_mode_adjustments): Cope with polynomial vector sizes.
16837 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16839 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16841 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16842 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16843 * caller-save.c (setup_save_areas): Likewise.
16844 (replace_reg_with_saved_mem): Likewise.
16845 * calls.c (emit_library_call_value_1): Likewise.
16846 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16847 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16848 (gen_lowpart_for_combine): Likewise.
16849 * convert.c (convert_to_integer_1): Likewise.
16850 * cse.c (equiv_constant, cse_insn): Likewise.
16851 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16852 (cselib_subst_to_values): Likewise.
16853 * dce.c (word_dce_process_block): Likewise.
16854 * df-problems.c (df_word_lr_mark_ref): Likewise.
16855 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16856 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16857 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16858 (rtl_for_decl_location): Likewise.
16859 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16860 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16861 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16862 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16863 (expand_expr_real_1): Likewise.
16864 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16865 (pad_below): Likewise.
16866 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16867 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16868 * ira.c (get_subreg_tracking_sizes): Likewise.
16869 * ira-build.c (ira_create_allocno_objects): Likewise.
16870 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16871 (ira_sort_regnos_for_alter_reg): Likewise.
16872 * ira-costs.c (record_operand_costs): Likewise.
16873 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16874 (resolve_simple_move): Likewise.
16875 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16876 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16877 (lra_constraints): Likewise.
16878 (CONST_POOL_OK_P): Reject variable-sized modes.
16879 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16880 (add_pseudo_to_slot, lra_spill): Likewise.
16881 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16882 * optabs-query.c (get_best_extraction_insn): Likewise.
16883 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16884 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16885 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16886 * recog.c (offsettable_address_addr_space_p): Likewise.
16887 * regcprop.c (maybe_mode_change): Likewise.
16888 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16889 * regrename.c (build_def_use): Likewise.
16890 * regstat.c (dump_reg_info): Likewise.
16891 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16892 (find_reloads, find_reloads_subreg_address): Likewise.
16893 * reload1.c (eliminate_regs_1): Likewise.
16894 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16895 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16896 (simplify_binary_operation_1, simplify_subreg): Likewise.
16897 * targhooks.c (default_function_arg_padding): Likewise.
16898 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16899 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16900 (verify_gimple_assign_ternary): Likewise.
16901 * tree-inline.c (estimate_move_cost): Likewise.
16902 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16903 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16904 (get_address_cost_ainc): Likewise.
16905 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16906 (vect_supportable_dr_alignment): Likewise.
16907 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16908 (vectorizable_reduction): Likewise.
16909 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16910 (vectorizable_operation, vectorizable_load): Likewise.
16911 * tree.c (build_same_sized_truth_vector_type): Likewise.
16912 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16913 * var-tracking.c (emit_note_insn_var_location): Likewise.
16914 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16915 (ADDR_VEC_ALIGN): Likewise.
16917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16918 Alan Hayward <alan.hayward@arm.com>
16919 David Sherwood <david.sherwood@arm.com>
16921 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16923 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16924 or if measurement_type is polynomial.
16925 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16926 * combine.c (make_extraction): Likewise.
16927 * dse.c (find_shift_sequence): Likewise.
16928 * dwarf2out.c (mem_loc_descriptor): Likewise.
16929 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16930 (extract_bit_field, extract_low_bits): Likewise.
16931 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16932 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16933 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16934 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16935 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16936 * reload.c (find_reloads): Likewise.
16937 * reload1.c (alter_reg): Likewise.
16938 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16939 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16940 * tree-if-conv.c (predicate_mem_writes): Likewise.
16941 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16942 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16943 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16944 * valtrack.c (dead_debug_insert_temp): Likewise.
16945 * varasm.c (mergeable_constant_section): Likewise.
16946 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16949 Alan Hayward <alan.hayward@arm.com>
16950 David Sherwood <david.sherwood@arm.com>
16952 * expr.c (expand_assignment): Cope with polynomial mode sizes
16953 when assigning to a CONCAT.
16955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16956 Alan Hayward <alan.hayward@arm.com>
16957 David Sherwood <david.sherwood@arm.com>
16959 * machmode.h (mode_precision): Change from unsigned short to
16961 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16963 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16964 or if measurement_type is not polynomial.
16965 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16966 in which the mode is already known to be a scalar_int_mode.
16967 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16968 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16970 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16971 for GET_MODE_PRECISION.
16972 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16973 for GET_MODE_PRECISION.
16974 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16976 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16977 (expand_field_assignment, make_extraction): Likewise.
16978 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16979 (get_last_value): Likewise.
16980 * convert.c (convert_to_integer_1): Likewise.
16981 * cse.c (cse_insn): Likewise.
16982 * expr.c (expand_expr_real_1): Likewise.
16983 * lra-constraints.c (simplify_operand_subreg): Likewise.
16984 * optabs-query.c (can_atomic_load_p): Likewise.
16985 * optabs.c (expand_atomic_load): Likewise.
16986 (expand_atomic_store): Likewise.
16987 * ree.c (combine_reaching_defs): Likewise.
16988 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16989 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16990 * tree.h (type_has_mode_precision_p): Likewise.
16991 * ubsan.c (instrument_si_overflow): Likewise.
16993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16994 Alan Hayward <alan.hayward@arm.com>
16995 David Sherwood <david.sherwood@arm.com>
16997 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16998 polynomial numbers of units.
16999 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17000 (valid_vector_subparts_p): New function.
17001 (build_vector_type): Remove temporary shim and take the number
17002 of units as a poly_uint64 rather than an int.
17003 (build_opaque_vector_type): Take the number of units as a
17004 poly_uint64 rather than an int.
17005 * tree.c (build_vector_from_ctor): Handle polynomial
17006 TYPE_VECTOR_SUBPARTS.
17007 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17008 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17009 (build_vector_from_val): If the number of units is variable,
17010 use build_vec_duplicate_cst for constant operands and
17011 VEC_DUPLICATE_EXPR otherwise.
17012 (make_vector_type): Remove temporary is_constant ().
17013 (build_vector_type, build_opaque_vector_type): Take the number of
17014 units as a poly_uint64 rather than an int.
17015 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17017 * cfgexpand.c (expand_debug_expr): Likewise.
17018 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17019 (store_constructor, expand_expr_real_1): Likewise.
17020 (const_scalar_mask_from_tree): Likewise.
17021 * fold-const-call.c (fold_const_reduction): Likewise.
17022 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17023 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17024 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17025 (fold_relational_const): Likewise.
17026 (native_interpret_vector): Likewise. Change the size from an
17027 int to an unsigned int.
17028 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17029 TYPE_VECTOR_SUBPARTS.
17030 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17031 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17032 duplicating a non-constant operand into a variable-length vector.
17033 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17034 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17035 * ipa-icf.c (sem_variable::equals): Likewise.
17036 * match.pd: Likewise.
17037 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17038 * print-tree.c (print_node): Likewise.
17039 * stor-layout.c (layout_type): Likewise.
17040 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17041 * tree-cfg.c (verify_gimple_comparison): Likewise.
17042 (verify_gimple_assign_binary): Likewise.
17043 (verify_gimple_assign_ternary): Likewise.
17044 (verify_gimple_assign_single): Likewise.
17045 * tree-pretty-print.c (dump_generic_node): Likewise.
17046 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17047 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17048 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17049 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17050 (vect_shift_permute_load_chain): Likewise.
17051 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17052 (expand_vector_condition, optimize_vector_constructor): Likewise.
17053 (lower_vec_perm, get_compute_type): Likewise.
17054 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17055 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17056 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17057 (vect_recog_mask_conversion_pattern): Likewise.
17058 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17059 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17060 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17061 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17062 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17063 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17064 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17065 (supportable_widening_operation): Likewise.
17066 (supportable_narrowing_operation): Likewise.
17067 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17069 * varasm.c (output_constant): Likewise.
17071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17072 Alan Hayward <alan.hayward@arm.com>
17073 David Sherwood <david.sherwood@arm.com>
17075 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17076 so that both the length == 3 and length != 3 cases set up their
17077 own permute vectors. Add comments explaining why we know the
17078 number of elements is constant.
17079 (vect_permute_load_chain): Likewise.
17081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17082 Alan Hayward <alan.hayward@arm.com>
17083 David Sherwood <david.sherwood@arm.com>
17085 * machmode.h (mode_nunits): Change from unsigned char to
17087 (ONLY_FIXED_SIZE_MODES): New macro.
17088 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17089 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17090 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17092 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17093 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17094 or if measurement_type is not polynomial.
17095 * genmodes.c (ZERO_COEFFS): New macro.
17096 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17098 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17099 Use ZERO_COEFFS when emitting initializers.
17100 * data-streamer.h (bp_pack_poly_value): New function.
17101 (bp_unpack_poly_value): Likewise.
17102 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17103 for GET_MODE_NUNITS.
17104 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17105 for GET_MODE_NUNITS.
17106 * tree.c (make_vector_type): Remove temporary shim and make
17107 the real function take the number of units as a poly_uint64
17108 rather than an int.
17109 (build_vector_type_for_mode): Handle polynomial nunits.
17110 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17111 * emit-rtl.c (const_vec_series_p_1): Likewise.
17112 (gen_rtx_CONST_VECTOR): Likewise.
17113 * fold-const.c (test_vec_duplicate_folding): Likewise.
17114 * genrecog.c (validate_pattern): Likewise.
17115 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17116 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17117 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17118 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17119 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17120 * rtlanal.c (subreg_get_info): Likewise.
17121 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17122 (vect_grouped_load_supported): Likewise.
17123 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17124 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17125 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17126 (simplify_const_unary_operation, simplify_binary_operation_1)
17127 (simplify_const_binary_operation, simplify_ternary_operation)
17128 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17129 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17130 instead of CONST_VECTOR_NUNITS.
17131 * varasm.c (output_constant_pool_2): Likewise.
17132 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17133 explicit-encoded elements in the XVEC for variable-length vectors.
17135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17137 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17140 Alan Hayward <alan.hayward@arm.com>
17141 David Sherwood <david.sherwood@arm.com>
17143 * coretypes.h (fixed_size_mode): Declare.
17144 (fixed_size_mode_pod): New typedef.
17145 * builtins.h (target_builtins::x_apply_args_mode)
17146 (target_builtins::x_apply_result_mode): Change type to
17147 fixed_size_mode_pod.
17148 * builtins.c (apply_args_size, apply_result_size, result_vector)
17149 (expand_builtin_apply_args_1, expand_builtin_apply)
17150 (expand_builtin_return): Update accordingly.
17152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17154 * cse.c (hash_rtx_cb): Hash only the encoded elements.
17155 * cselib.c (cselib_hash_rtx): Likewise.
17156 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
17157 CONST_VECTOR encoding.
17159 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17160 Jeff Law <law@redhat.com>
17163 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
17164 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
17165 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
17166 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
17169 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
17170 explicitly probe *sp in a noreturn function if there were any callee
17171 register saves or frame pointer is needed.
17173 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17176 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
17177 BLKmode for ternary, binary or unary expressions.
17180 * var-tracking.c (delete_vta_debug_insn): New inline function.
17181 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
17182 insns from get_insns () to NULL instead of each bb separately.
17183 Use delete_vta_debug_insn. No longer static.
17184 (vt_debug_insns_local, variable_tracking_main_1): Adjust
17185 delete_vta_debug_insns callers.
17186 * rtl.h (delete_vta_debug_insns): Declare.
17187 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
17188 instead of variable_tracking_main.
17190 2018-01-03 Martin Sebor <msebor@redhat.com>
17192 PR tree-optimization/83603
17193 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
17194 arguments past the endof the argument list in functions declared
17195 without a prototype.
17196 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
17197 Avoid checking when arguments are null.
17199 2018-01-03 Martin Sebor <msebor@redhat.com>
17202 * doc/extend.texi (attribute const): Fix a typo.
17203 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
17204 issuing -Wsuggest-attribute for void functions.
17206 2018-01-03 Martin Sebor <msebor@redhat.com>
17208 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
17209 offset_int::from instead of wide_int::to_shwi.
17210 (maybe_diag_overlap): Remove assertion.
17211 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
17212 * gimple-ssa-sprintf.c (format_directive): Same.
17213 (parse_directive): Same.
17214 (sprintf_dom_walker::compute_format_length): Same.
17215 (try_substitute_return_value): Same.
17217 2018-01-03 Jeff Law <law@redhat.com>
17219 PR middle-end/83654
17220 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
17221 non-constant residual for zero at runtime and avoid probing in
17222 that case. Reorganize code for trailing problem to mirror handling
17225 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17227 PR tree-optimization/83501
17228 * tree-ssa-strlen.c (get_string_cst): New.
17229 (handle_char_store): Call get_string_cst.
17231 2018-01-03 Martin Liska <mliska@suse.cz>
17233 PR tree-optimization/83593
17234 * tree-ssa-strlen.c: Include tree-cfg.h.
17235 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
17236 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
17237 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
17239 (strlen_dom_walker::before_dom_children): Call
17240 gimple_purge_dead_eh_edges. Dump tranformation with details
17242 (strlen_dom_walker::before_dom_children): Update call by adding
17243 new argument cleanup_eh.
17244 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
17246 2018-01-03 Martin Liska <mliska@suse.cz>
17249 * cif-code.def (VARIADIC_THUNK): New enum value.
17250 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
17253 2018-01-03 Jan Beulich <jbeulich@suse.com>
17255 * sse.md (mov<mode>_internal): Tighten condition for when to use
17256 vmovdqu<ssescalarsize> for TI and OI modes.
17258 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17260 Update copyright years.
17262 2018-01-03 Martin Liska <mliska@suse.cz>
17265 * ipa-visibility.c (function_and_variable_visibility): Skip
17266 functions with noipa attribure.
17268 2018-01-03 Jakub Jelinek <jakub@redhat.com>
17270 * gcc.c (process_command): Update copyright notice dates.
17271 * gcov-dump.c (print_version): Ditto.
17272 * gcov.c (print_version): Ditto.
17273 * gcov-tool.c (print_version): Ditto.
17274 * gengtype.c (create_file): Ditto.
17275 * doc/cpp.texi: Bump @copying's copyright year.
17276 * doc/cppinternals.texi: Ditto.
17277 * doc/gcc.texi: Ditto.
17278 * doc/gccint.texi: Ditto.
17279 * doc/gcov.texi: Ditto.
17280 * doc/install.texi: Ditto.
17281 * doc/invoke.texi: Ditto.
17283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17285 * vector-builder.h (vector_builder::m_full_nelts): Change from
17286 unsigned int to poly_uint64.
17287 (vector_builder::full_nelts): Update prototype accordingly.
17288 (vector_builder::new_vector): Likewise.
17289 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
17290 (vector_builder::operator ==): Likewise.
17291 (vector_builder::finalize): Likewise.
17292 * int-vector-builder.h (int_vector_builder::int_vector_builder):
17293 Take the number of elements as a poly_uint64 rather than an
17295 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
17296 from unsigned int to poly_uint64.
17297 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
17298 (vec_perm_indices::new_vector): Likewise.
17299 (vec_perm_indices::length): Likewise.
17300 (vec_perm_indices::nelts_per_input): Likewise.
17301 (vec_perm_indices::input_nelts): Likewise.
17302 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
17303 number of elements per input as a poly_uint64 rather than an
17304 unsigned int. Use the original encoding for variable-length
17305 vectors, rather than clamping each individual element.
17306 For the second and subsequent elements in each pattern,
17307 clamp the step and base before clamping their sum.
17308 (vec_perm_indices::series_p): Handle polynomial element counts.
17309 (vec_perm_indices::all_in_range_p): Likewise.
17310 (vec_perm_indices_to_tree): Likewise.
17311 (vec_perm_indices_to_rtx): Likewise.
17312 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
17313 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
17314 (tree_vector_builder::new_binary_operation): Handle polynomial
17315 element counts. Return false if we need to know the number
17316 of elements at compile time.
17317 * fold-const.c (fold_vec_perm): Punt if the number of elements
17318 isn't known at compile time.
17320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17322 * vec-perm-indices.h (vec_perm_builder): Change element type
17323 from HOST_WIDE_INT to poly_int64.
17324 (vec_perm_indices::element_type): Update accordingly.
17325 (vec_perm_indices::clamp): Handle polynomial element_types.
17326 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17327 (vec_perm_indices::all_in_range_p): Likewise.
17328 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
17330 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
17331 polynomial vec_perm_indices element types.
17332 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
17333 * fold-const.c (fold_vec_perm): Likewise.
17334 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
17335 * tree-vect-generic.c (lower_vec_perm): Likewise.
17336 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17337 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
17338 element type to HOST_WIDE_INT.
17340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17341 Alan Hayward <alan.hayward@arm.com>
17342 David Sherwood <david.sherwood@arm.com>
17344 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
17345 rather than an int. Use plus_constant.
17346 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
17347 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
17349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17350 Alan Hayward <alan.hayward@arm.com>
17351 David Sherwood <david.sherwood@arm.com>
17353 * calls.c (emit_call_1, expand_call): Change struct_value_size from
17354 a HOST_WIDE_INT to a poly_int64.
17356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17357 Alan Hayward <alan.hayward@arm.com>
17358 David Sherwood <david.sherwood@arm.com>
17360 * calls.c (load_register_parameters): Cope with polynomial
17361 mode sizes. Require a constant size for BLKmode parameters
17362 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
17363 forces a parameter to be padded at the lsb end in order to
17364 fill a complete number of words, require the parameter size
17365 to be ordered wrt UNITS_PER_WORD.
17367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17368 Alan Hayward <alan.hayward@arm.com>
17369 David Sherwood <david.sherwood@arm.com>
17371 * reload1.c (spill_stack_slot_width): Change element type
17372 from unsigned int to poly_uint64_pod.
17373 (alter_reg): Treat mode sizes as polynomial.
17375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17376 Alan Hayward <alan.hayward@arm.com>
17377 David Sherwood <david.sherwood@arm.com>
17379 * reload.c (complex_word_subreg_p): New function.
17380 (reload_inner_reg_of_subreg, push_reload): Use it.
17382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17383 Alan Hayward <alan.hayward@arm.com>
17384 David Sherwood <david.sherwood@arm.com>
17386 * lra-constraints.c (process_alt_operands): Reject matched
17387 operands whose sizes aren't ordered.
17388 (match_reload): Refer to this check here.
17390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17391 Alan Hayward <alan.hayward@arm.com>
17392 David Sherwood <david.sherwood@arm.com>
17394 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
17395 that the mode size is in the set {1, 2, 4, 8, 16}.
17397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17398 Alan Hayward <alan.hayward@arm.com>
17399 David Sherwood <david.sherwood@arm.com>
17401 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
17402 Use plus_constant instead of gen_rtx_PLUS.
17404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17405 Alan Hayward <alan.hayward@arm.com>
17406 David Sherwood <david.sherwood@arm.com>
17408 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
17409 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
17410 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
17411 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
17412 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
17413 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
17414 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
17415 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
17416 * config/i386/i386.c (ix86_push_rounding): ...this new function.
17417 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
17419 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
17420 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
17421 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
17422 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
17423 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
17424 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
17425 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
17426 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
17427 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
17428 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
17430 * expr.c (emit_move_resolve_push): Treat the input and result
17431 of PUSH_ROUNDING as a poly_int64.
17432 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
17433 (emit_push_insn): Likewise.
17434 * lra-eliminations.c (mark_not_eliminable): Likewise.
17435 * recog.c (push_operand): Likewise.
17436 * reload1.c (elimination_effects): Likewise.
17437 * rtlanal.c (nonzero_bits1): Likewise.
17438 * calls.c (store_one_arg): Likewise. Require the padding to be
17439 known at compile time.
17441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17442 Alan Hayward <alan.hayward@arm.com>
17443 David Sherwood <david.sherwood@arm.com>
17445 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
17446 Use plus_constant instead of gen_rtx_PLUS.
17448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17449 Alan Hayward <alan.hayward@arm.com>
17450 David Sherwood <david.sherwood@arm.com>
17452 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
17453 rather than an int.
17455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17456 Alan Hayward <alan.hayward@arm.com>
17457 David Sherwood <david.sherwood@arm.com>
17459 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
17460 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
17461 via stack temporaries. Treat the mode size as polynomial too.
17463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17464 Alan Hayward <alan.hayward@arm.com>
17465 David Sherwood <david.sherwood@arm.com>
17467 * expr.c (expand_expr_real_2): When handling conversions involving
17468 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
17469 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
17470 as a poly_uint64 too.
17472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17473 Alan Hayward <alan.hayward@arm.com>
17474 David Sherwood <david.sherwood@arm.com>
17476 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
17478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17479 Alan Hayward <alan.hayward@arm.com>
17480 David Sherwood <david.sherwood@arm.com>
17482 * combine.c (can_change_dest_mode): Handle polynomial
17483 REGMODE_NATURAL_SIZE.
17484 * expmed.c (store_bit_field_1): Likewise.
17485 * expr.c (store_constructor): Likewise.
17486 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
17487 and polynomial REGMODE_NATURAL_SIZE.
17488 (gen_lowpart_common): Likewise.
17489 * reginfo.c (record_subregs_of_mode): Likewise.
17490 * rtlanal.c (read_modify_subreg_p): Likewise.
17492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17493 Alan Hayward <alan.hayward@arm.com>
17494 David Sherwood <david.sherwood@arm.com>
17496 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
17497 numbers of elements.
17499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17500 Alan Hayward <alan.hayward@arm.com>
17501 David Sherwood <david.sherwood@arm.com>
17503 * match.pd: Cope with polynomial numbers of vector elements.
17505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17506 Alan Hayward <alan.hayward@arm.com>
17507 David Sherwood <david.sherwood@arm.com>
17509 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
17510 in a POINTER_PLUS_EXPR.
17512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17513 Alan Hayward <alan.hayward@arm.com>
17514 David Sherwood <david.sherwood@arm.com>
17516 * omp-simd-clone.c (simd_clone_subparts): New function.
17517 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
17518 (ipa_simd_modify_function_body): Likewise.
17520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17521 Alan Hayward <alan.hayward@arm.com>
17522 David Sherwood <david.sherwood@arm.com>
17524 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
17525 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17526 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17527 (expand_vector_condition, vector_element): Likewise.
17528 (subparts_gt): New function.
17529 (get_compute_type): Use subparts_gt.
17530 (count_type_subparts): Delete.
17531 (expand_vector_operations_1): Use subparts_gt instead of
17532 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17539 (vect_compile_time_alias): ...this new function. Do the calculation
17540 on poly_ints rather than trees.
17541 (vect_prune_runtime_alias_test_list): Update call accordingly.
17543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17544 Alan Hayward <alan.hayward@arm.com>
17545 David Sherwood <david.sherwood@arm.com>
17547 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17549 (vect_schedule_slp_instance): Likewise.
17551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17552 Alan Hayward <alan.hayward@arm.com>
17553 David Sherwood <david.sherwood@arm.com>
17555 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17556 constant and extern definitions for variable-length vectors.
17557 (vect_get_constant_vectors): Note that the number of units
17558 is known to be constant.
17560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17561 Alan Hayward <alan.hayward@arm.com>
17562 David Sherwood <david.sherwood@arm.com>
17564 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17565 of units as polynomial. Choose between WIDE and NARROW based
17568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17569 Alan Hayward <alan.hayward@arm.com>
17570 David Sherwood <david.sherwood@arm.com>
17572 * tree-vect-stmts.c (simd_clone_subparts): New function.
17573 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17576 Alan Hayward <alan.hayward@arm.com>
17577 David Sherwood <david.sherwood@arm.com>
17579 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17580 vectors as polynomial. Use build_index_vector for
17581 IFN_GOMP_SIMD_LANE.
17583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17584 Alan Hayward <alan.hayward@arm.com>
17585 David Sherwood <david.sherwood@arm.com>
17587 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17588 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17589 for variable-length vectors.
17590 (vectorizable_mask_load_store): Treat the number of units as
17591 polynomial, asserting that it is constant if the condition has
17592 already been enforced.
17593 (vectorizable_store, vectorizable_load): Likewise.
17595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17596 Alan Hayward <alan.hayward@arm.com>
17597 David Sherwood <david.sherwood@arm.com>
17599 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17600 of units as polynomial. Punt if we can't tell at compile time
17601 which vector contains the final result.
17603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17604 Alan Hayward <alan.hayward@arm.com>
17605 David Sherwood <david.sherwood@arm.com>
17607 * tree-vect-loop.c (vectorizable_induction): Treat the number
17608 of units as polynomial. Punt on SLP inductions. Use an integer
17609 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17610 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
17618 * tree.c (build_index_vector): New function.
17619 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17620 of units as polynomial, forcibly converting it to a constant if
17621 vectorizable_reduction has already enforced the condition.
17622 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17623 to create a {1,2,3,...} vector.
17624 (vectorizable_reduction): Treat the number of units as polynomial.
17625 Choose vectype_in based on the largest scalar element size rather
17626 than the smallest number of units. Enforce the restrictions
17629 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17630 Alan Hayward <alan.hayward@arm.com>
17631 David Sherwood <david.sherwood@arm.com>
17633 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17634 number of units as polynomial.
17636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17637 Alan Hayward <alan.hayward@arm.com>
17638 David Sherwood <david.sherwood@arm.com>
17640 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17641 * target.def (autovectorize_vector_sizes): Return the vector sizes
17642 by pointer, using vector_sizes rather than a bitmask.
17643 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17644 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17645 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17647 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17648 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17649 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17650 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17651 * omp-general.c (omp_max_vf): Likewise.
17652 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17653 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17654 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17655 * tree-vect-slp.c (vect_slp_bb): Likewise.
17656 * doc/tm.texi: Regenerate.
17657 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17659 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17660 the vector size as a poly_uint64 rather than an unsigned int.
17661 (current_vector_size): Change from an unsigned int to a poly_uint64.
17662 (get_vectype_for_scalar_type): Update accordingly.
17663 * tree.h (build_truth_vector_type): Take the size and number of
17664 units as a poly_uint64 rather than an unsigned int.
17665 (build_vector_type): Add a temporary overload that takes
17666 the number of units as a poly_uint64 rather than an unsigned int.
17667 * tree.c (make_vector_type): Likewise.
17668 (build_truth_vector_type): Take the number of units as a poly_uint64
17669 rather than an unsigned int.
17671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17672 Alan Hayward <alan.hayward@arm.com>
17673 David Sherwood <david.sherwood@arm.com>
17675 * target.def (get_mask_mode): Take the number of units and length
17676 as poly_uint64s rather than unsigned ints.
17677 * targhooks.h (default_get_mask_mode): Update accordingly.
17678 * targhooks.c (default_get_mask_mode): Likewise.
17679 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17680 * doc/tm.texi: Regenerate.
17682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17683 Alan Hayward <alan.hayward@arm.com>
17684 David Sherwood <david.sherwood@arm.com>
17686 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17687 * omp-general.c (omp_max_vf): Likewise.
17688 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17689 (expand_omp_simd): Handle polynomial safelen.
17690 * omp-low.c (omplow_simd_context): Add a default constructor.
17691 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17692 (lower_rec_simd_input_clauses): Update accordingly.
17693 (lower_rec_input_clauses): Likewise.
17695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17696 Alan Hayward <alan.hayward@arm.com>
17697 David Sherwood <david.sherwood@arm.com>
17699 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17700 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17701 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17702 (vect_analyze_slp_cost): Likewise.
17703 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17704 (vect_model_load_cost): Likewise.
17706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17707 Alan Hayward <alan.hayward@arm.com>
17708 David Sherwood <david.sherwood@arm.com>
17710 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17711 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17712 from an unsigned int * to a poly_uint64_pod *.
17713 (calculate_unrolling_factor): New function.
17714 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17717 Alan Hayward <alan.hayward@arm.com>
17718 David Sherwood <david.sherwood@arm.com>
17720 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17721 from an unsigned int to a poly_uint64.
17722 (_loop_vec_info::slp_unrolling_factor): Likewise.
17723 (_loop_vec_info::vectorization_factor): Change from an int
17725 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17726 (vect_get_num_vectors): New function.
17727 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17728 (vect_get_num_copies): Use vect_get_num_vectors.
17729 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17730 to an unsigned int *.
17731 (vect_analyze_data_refs): Change min_vf from an int * to a
17733 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17734 than an unsigned HOST_WIDE_INT.
17735 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17736 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17737 to an unsigned int *.
17738 (vect_analyze_data_ref_dependences): Likewise.
17739 (vect_compute_data_ref_alignment): Handle polynomial vf.
17740 (vect_enhance_data_refs_alignment): Likewise.
17741 (vect_prune_runtime_alias_test_list): Likewise.
17742 (vect_shift_permute_load_chain): Likewise.
17743 (vect_supportable_dr_alignment): Likewise.
17744 (dependence_distance_ge_vf): Take the vectorization factor as a
17745 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17746 (vect_analyze_data_refs): Change min_vf from an int * to a
17748 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17749 vfm1 as a poly_uint64 rather than an int. Make the same change
17750 for the returned bound_scalar.
17751 (vect_gen_vector_loop_niters): Handle polynomial vf.
17752 (vect_do_peeling): Likewise. Update call to
17753 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17754 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17756 * tree-vect-loop.c (vect_determine_vectorization_factor)
17757 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17758 (vect_get_known_peeling_cost): Likewise.
17759 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17760 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17761 (vect_transform_loop): Likewise. Use the lowest possible VF when
17762 updating the upper bounds of the loop.
17763 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17764 rather than an int.
17765 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17766 polynomial unroll factors.
17767 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17768 (vect_make_slp_decision): Likewise.
17769 (vect_supported_load_permutation_p): Likewise, and polynomial
17771 (vect_analyze_slp_cost): Handle polynomial vf.
17772 (vect_slp_analyze_node_operations): Likewise.
17773 (vect_slp_analyze_bb_1): Likewise.
17774 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17775 than an unsigned HOST_WIDE_INT.
17776 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17777 (vectorizable_load): Handle polynomial vf.
17778 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17780 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17783 Alan Hayward <alan.hayward@arm.com>
17784 David Sherwood <david.sherwood@arm.com>
17786 * match.pd: Handle bit operations involving three constants
17787 and try to fold one pair.
17789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17791 * tree-vect-loop-manip.c: Include gimple-fold.h.
17792 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17793 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17794 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17795 Add a path that uses a step of VF instead of 1, but disable it
17797 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17798 and niters_no_overflow parameters. Update calls to
17799 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17800 Create a new SSA name if the latter choses to use a ste other
17801 than zero, and return it via niters_vector_mult_vf_var.
17802 * tree-vect-loop.c (vect_transform_loop): Update calls to
17803 vect_do_peeling, vect_gen_vector_loop_niters and
17804 slpeel_make_loop_iterate_ntimes.
17805 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17806 (vect_gen_vector_loop_niters): Update declarations after above changes.
17808 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17810 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17811 128-bit round to integer instructions.
17812 (ceil<mode>2): Likewise.
17813 (btrunc<mode>2): Likewise.
17814 (round<mode>2): Likewise.
17816 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17818 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17819 unaligned VSX load/store on P8/P9.
17820 (expand_block_clear): Allow the use of unaligned VSX
17821 load/store on P8/P9.
17823 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17825 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17827 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17828 swap associated with both a load and a store.
17830 2018-01-02 Andrew Waterman <andrew@sifive.com>
17832 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17833 * config/riscv/riscv.md (clear_cache): Use it.
17835 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17837 * web.c: Remove out-of-date comment.
17839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17841 * expr.c (fixup_args_size_notes): Check that any existing
17842 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17843 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17844 (emit_single_push_insn): ...here.
17846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17848 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17849 (const_vector_encoded_nelts): New function.
17850 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17851 (const_vector_int_elt, const_vector_elt): Declare.
17852 * emit-rtl.c (const_vector_int_elt_1): New function.
17853 (const_vector_elt): Likewise.
17854 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17855 of CONST_VECTOR_ELT.
17857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17859 * expr.c: Include rtx-vector-builder.h.
17860 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17861 directly on the tree encoding.
17862 (const_vector_from_tree): Likewise.
17863 * optabs.c: Include rtx-vector-builder.h.
17864 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17865 sequence of "u" values.
17866 * vec-perm-indices.c: Include rtx-vector-builder.h.
17867 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17868 directly on the vec_perm_indices encoding.
17870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17872 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17873 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17874 * rtx-vector-builder.h: New file.
17875 * rtx-vector-builder.c: Likewise.
17876 * rtl.h (rtx_def::u2): Add a const_vector field.
17877 (CONST_VECTOR_NPATTERNS): New macro.
17878 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17879 (CONST_VECTOR_DUPLICATE_P): Likewise.
17880 (CONST_VECTOR_STEPPED_P): Likewise.
17881 (CONST_VECTOR_ENCODED_ELT): Likewise.
17882 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17883 (unwrap_const_vec_duplicate): Likewise.
17884 (const_vec_series_p): Check for a non-duplicated vector encoding.
17885 Say that the function only returns true for integer vectors.
17886 * emit-rtl.c: Include rtx-vector-builder.h.
17887 (gen_const_vec_duplicate_1): Delete.
17888 (gen_const_vector): Call gen_const_vec_duplicate instead of
17889 gen_const_vec_duplicate_1.
17890 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17891 (gen_const_vec_duplicate): Use rtx_vector_builder.
17892 (gen_const_vec_series): Likewise.
17893 (gen_rtx_CONST_VECTOR): Likewise.
17894 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17895 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17896 Build a new vector rather than modifying a CONST_VECTOR in-place.
17897 (handle_special_swappables): Update call accordingly.
17898 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17899 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17900 Build a new vector rather than modifying a CONST_VECTOR in-place.
17901 (handle_special_swappables): Update call accordingly.
17903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17905 * simplify-rtx.c (simplify_const_binary_operation): Use
17906 CONST_VECTOR_ELT instead of XVECEXP.
17908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17910 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17911 the selector elements to be different from the data elements
17912 if the selector is a VECTOR_CST.
17913 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17914 ssizetype for the selector.
17916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17918 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17919 before testing each element individually.
17920 * tree-vect-generic.c (lower_vec_perm): Likewise.
17922 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17924 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17925 * selftest-run-tests.c (selftest::run_tests): Call it.
17926 * vector-builder.h (vector_builder::operator ==): New function.
17927 (vector_builder::operator !=): Likewise.
17928 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17929 (vec_perm_indices::all_from_input_p): New function.
17930 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17931 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17932 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17933 instead of reading the VECTOR_CST directly. Detect whether both
17934 vector inputs are the same before constructing the vec_perm_indices,
17935 and update the number of inputs argument accordingly. Use the
17936 utility functions added above. Only construct sel2 if we need to.
17938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17940 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17941 the broadcast of the low byte.
17942 (expand_mult_highpart): Use an explicit encoding for the permutes.
17943 * optabs-query.c (can_mult_highpart_p): Likewise.
17944 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17945 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17946 (vectorizable_bswap): Likewise.
17947 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17948 explicit encoding for the power-of-2 permutes.
17949 (vect_permute_store_chain): Likewise.
17950 (vect_grouped_load_supported): Likewise.
17951 (vect_permute_load_chain): Likewise.
17953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17955 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17956 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17957 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17958 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17959 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17960 (vect_gen_perm_mask_any): Likewise.
17962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17964 * int-vector-builder.h: New file.
17965 * vec-perm-indices.h: Include int-vector-builder.h.
17966 (vec_perm_indices): Redefine as an int_vector_builder.
17967 (auto_vec_perm_indices): Delete.
17968 (vec_perm_builder): Redefine as a stand-alone class.
17969 (vec_perm_indices::vec_perm_indices): New function.
17970 (vec_perm_indices::clamp): Likewise.
17971 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17972 (vec_perm_indices::new_vector): New function.
17973 (vec_perm_indices::new_expanded_vector): Update for new
17974 vec_perm_indices class.
17975 (vec_perm_indices::rotate_inputs): New function.
17976 (vec_perm_indices::all_in_range_p): Operate directly on the
17977 encoded form, without computing elided elements.
17978 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17979 encoding. Update for new vec_perm_indices class.
17980 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17981 the given vec_perm_builder.
17982 (expand_vec_perm_var): Update vec_perm_builder constructor.
17983 (expand_mult_highpart): Use vec_perm_builder instead of
17984 auto_vec_perm_indices.
17985 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17986 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17987 or double series encoding as appropriate.
17988 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17989 vec_perm_indices instead of auto_vec_perm_indices.
17990 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17991 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17992 (vect_permute_store_chain): Likewise.
17993 (vect_grouped_load_supported): Likewise.
17994 (vect_permute_load_chain): Likewise.
17995 (vect_shift_permute_load_chain): Likewise.
17996 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17997 (vect_transform_slp_perm_load): Likewise.
17998 (vect_schedule_slp_instance): Likewise.
17999 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18000 (vectorizable_mask_load_store): Likewise.
18001 (vectorizable_bswap): Likewise.
18002 (vectorizable_store): Likewise.
18003 (vectorizable_load): Likewise.
18004 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18005 vec_perm_indices instead of auto_vec_perm_indices. Use
18006 tree_to_vec_perm_builder to read the vector from a tree.
18007 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18008 vec_perm_builder instead of a vec_perm_indices.
18009 (have_whole_vector_shift): Use vec_perm_builder and
18010 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18011 truncation to calc_vec_perm_mask_for_shift.
18012 (vect_create_epilog_for_reduction): Likewise.
18013 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18014 from auto_vec_perm_indices to vec_perm_indices.
18015 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18016 instead of changing individual elements.
18017 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18018 the vector in d.perm.
18019 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18020 from auto_vec_perm_indices to vec_perm_indices.
18021 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18022 instead of changing individual elements.
18023 (arm_vectorize_vec_perm_const): Use new_vector to install
18024 the vector in d.perm.
18025 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18026 Update vec_perm_builder constructor.
18027 (rs6000_expand_interleave): Likewise.
18028 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18029 (rs6000_expand_interleave): Likewise.
18031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18033 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18034 to qimode could truncate the indices.
18035 * optabs.c (expand_vec_perm_var): Likewise.
18037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18039 * Makefile.in (OBJS): Add vec-perm-indices.o.
18040 * vec-perm-indices.h: New file.
18041 * vec-perm-indices.c: Likewise.
18042 * target.h (vec_perm_indices): Replace with a forward class
18044 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18045 * optabs.h: Include vec-perm-indices.h.
18046 (expand_vec_perm): Delete.
18047 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18048 (expand_vec_perm_const): Declare.
18049 * target.def (vec_perm_const_ok): Replace with...
18050 (vec_perm_const): ...this new hook.
18051 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18052 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18053 * doc/tm.texi: Regenerate.
18054 * optabs.def (vec_perm_const): Delete.
18055 * doc/md.texi (vec_perm_const): Likewise.
18056 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18057 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18058 expand_vec_perm for constant permutation vectors. Assert that
18059 the mode of variable permutation vectors is the integer equivalent
18060 of the mode that is being permuted.
18061 * optabs-query.h (selector_fits_mode_p): Declare.
18062 * optabs-query.c: Include vec-perm-indices.h.
18063 (selector_fits_mode_p): New function.
18064 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18065 is defined, instead of checking whether the vec_perm_const_optab
18066 exists. Use targetm.vectorize.vec_perm_const instead of
18067 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18068 fit in the vector mode before using a variable permute.
18069 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18070 vec_perm_indices instead of an rtx.
18071 (expand_vec_perm): Replace with...
18072 (expand_vec_perm_const): ...this new function. Take the selector
18073 as a vec_perm_indices rather than an rtx. Also take the mode of
18074 the selector. Update call to shift_amt_for_vec_perm_mask.
18075 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18076 Use vec_perm_indices::new_expanded_vector to expand the original
18077 selector into bytes. Check whether the indices fit in the vector
18078 mode before using a variable permute.
18079 (expand_vec_perm_var): Make global.
18080 (expand_mult_highpart): Use expand_vec_perm_const.
18081 * fold-const.c: Includes vec-perm-indices.h.
18082 * tree-ssa-forwprop.c: Likewise.
18083 * tree-vect-data-refs.c: Likewise.
18084 * tree-vect-generic.c: Likewise.
18085 * tree-vect-loop.c: Likewise.
18086 * tree-vect-slp.c: Likewise.
18087 * tree-vect-stmts.c: Likewise.
18088 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18090 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18091 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18092 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18093 (aarch64_vectorize_vec_perm_const): ...this new function.
18094 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18095 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18096 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18097 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18098 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18099 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18100 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18102 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18103 check for NEON modes.
18104 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18105 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18106 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18107 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18109 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18110 the old VEC_PERM_CONST conditions.
18111 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18112 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18113 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18114 (ia64_vectorize_vec_perm_const_ok): Merge into...
18115 (ia64_vectorize_vec_perm_const): ...this new function.
18116 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18117 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18118 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18119 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18120 * config/mips/mips.c (mips_expand_vec_perm_const)
18121 (mips_vectorize_vec_perm_const_ok): Merge into...
18122 (mips_vectorize_vec_perm_const): ...this new function.
18123 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18124 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18125 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18126 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18127 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18128 (rs6000_expand_vec_perm_const): Delete.
18129 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18131 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18132 (altivec_expand_vec_perm_const_le): Take each operand individually.
18133 Operate on constant selectors rather than rtxes.
18134 (altivec_expand_vec_perm_const): Likewise. Update call to
18135 altivec_expand_vec_perm_const_le.
18136 (rs6000_expand_vec_perm_const): Delete.
18137 (rs6000_vectorize_vec_perm_const_ok): Delete.
18138 (rs6000_vectorize_vec_perm_const): New function.
18139 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18140 an element count and rtx array.
18141 (rs6000_expand_extract_even): Update call accordingly.
18142 (rs6000_expand_interleave): Likewise.
18143 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18144 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18145 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18146 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18147 (rs6000_expand_vec_perm_const): Delete.
18148 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18149 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18150 (altivec_expand_vec_perm_const_le): Take each operand individually.
18151 Operate on constant selectors rather than rtxes.
18152 (altivec_expand_vec_perm_const): Likewise. Update call to
18153 altivec_expand_vec_perm_const_le.
18154 (rs6000_expand_vec_perm_const): Delete.
18155 (rs6000_vectorize_vec_perm_const_ok): Delete.
18156 (rs6000_vectorize_vec_perm_const): New function. Remove stray
18157 reference to the SPE evmerge intructions.
18158 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18159 an element count and rtx array.
18160 (rs6000_expand_extract_even): Update call accordingly.
18161 (rs6000_expand_interleave): Likewise.
18162 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
18163 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
18165 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18169 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
18170 vector mode and that that mode matches the mode of the data
18172 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
18173 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
18174 directly using expand_vec_perm_1 when forcing selectors into
18176 (expand_vec_perm_var): New function, split out from expand_vec_perm.
18178 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18180 * optabs-query.h (can_vec_perm_p): Delete.
18181 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
18182 * optabs-query.c (can_vec_perm_p): Split into...
18183 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
18184 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
18185 particular selector is valid.
18186 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18187 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18188 (vect_grouped_load_supported): Likewise.
18189 (vect_shift_permute_load_chain): Likewise.
18190 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18191 (vect_transform_slp_perm_load): Likewise.
18192 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18193 (vectorizable_bswap): Likewise.
18194 (vect_gen_perm_mask_checked): Likewise.
18195 * fold-const.c (fold_ternary_loc): Likewise. Don't take
18196 implementations of variable permutation vectors into account
18197 when deciding which selector to use.
18198 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
18199 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
18200 with a false third argument.
18201 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
18202 to test whether the constant selector is valid and can_vec_perm_var_p
18203 to test whether a variable selector is valid.
18205 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18207 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
18208 * optabs-query.c (can_vec_perm_p): Likewise.
18209 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
18210 instead of vec_perm_indices.
18211 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
18212 (vect_gen_perm_mask_checked): Likewise,
18213 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
18214 (vect_gen_perm_mask_checked): Likewise,
18216 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18218 * optabs-query.h (qimode_for_vec_perm): Declare.
18219 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
18220 (qimode_for_vec_perm): ...this new function.
18221 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
18223 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18225 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
18226 does not have a conditional at the top.
18228 2018-01-02 Richard Biener <rguenther@suse.de>
18230 * ipa-inline.c (big_speedup_p): Fix expression.
18232 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18235 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
18238 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
18242 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
18243 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
18244 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
18245 cond_taken_branch_cost 3->4.
18247 2018-01-01 Jakub Jelinek <jakub@redhat.com>
18249 PR tree-optimization/83581
18250 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
18251 TODO_cleanup_cfg if any changes have been made.
18253 PR middle-end/83608
18254 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
18255 convert_modes if target mode has the right side, but different mode
18258 PR middle-end/83609
18259 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
18260 last argument when extracting from CONCAT. If either from_real or
18261 from_imag is NULL, use expansion through memory. If result is not
18262 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
18263 the parts directly to inner mode, if even that fails, use expansion
18266 PR middle-end/83623
18267 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
18268 check for bswap in mode rather than HImode and use that in expand_unop
18271 Copyright (C) 2018 Free Software Foundation, Inc.
18273 Copying and distribution of this file, with or without modification,
18274 are permitted in any medium without royalty provided the copyright
18275 notice and this notice are preserved.