1 2018-10-11 Martin Sebor <msebor@redhat.com>
3 * doc/extend.texi (attribute packed): Correct typos.
5 2018-10-11 Martin Sebor <msebor@redhat.com>
7 * doc/extend.texi (attribute flatten): Mention interaction with
10 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
13 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14 Set new_decl virtual flag to zero.
16 2018-10-11 Martin Sebor <msebor@redhat.com>
19 * doc/extend.texi (attribute format_arg): Discuss using multiple
20 attributes on a single function.
22 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
24 PR tree-optimization/86829
25 * match.pd (sin (atan (x))): New simplification rules.
26 (cos (atan (x))): Likewise.
27 * real.c (build_sinatan_real): New function.
28 * real.h (build_sinatan_real): Prototype.
30 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
32 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
34 (fold_mergeeo_helper): New helper function.
35 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
36 intrinsics. Correct some whitespace indentation issues.
38 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
41 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
42 Use HOST_WIDE_INT_1U for shift.
44 2018-10-11 Doug Rupp <rupp@adacore.com>
45 Olivier Hainque <hainque@adacore.com>
47 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
48 Pass --relax to the linker for RTPs.
49 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
51 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
52 Jan Hubicka <jh@suse.cz>
53 Martin Jambor <mjambor@suse.cz>
55 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
56 the same elements are repeated rather than printing all of them.
57 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
59 * read-rtl-function.c (test_loading_repeat): New function.
60 (read_rtl_function_c_tests): Call test_loading_repeat.
61 * rtl-tests.c (test_dumping_repeat): New function.
62 (rtl_tests_c_tests): Call test_dumping_repeat.
64 2018-10-11 Richard Biener <rguenther@suse.de>
66 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
67 bdver?_cost): Unify to ...
68 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
69 * config/i386/i386.c (processor_cost_table): Adjust.
71 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
74 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
75 the thunk when expanding to GIMPLE.
77 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
79 * varasm.c (mergeable_string_section): Don't try to move zero-length
80 strings to the merge section.
82 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
85 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
87 2018-10-10 Jakub Jelinek <jakub@redhat.com>
90 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
93 2018-10-10 Richard Biener <rguenther@suse.de>
95 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
96 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
97 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
98 and recursing and pattern terminating the recursion on SSE
99 vector width using ix86_expand_reduc.
100 (reduc_sminmax_scal_<mode>): Split into part reducing to half
101 width and recursing and SSE2 vector variant doing the final
102 reduction with ix86_expand_reduc.
103 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
104 with terminating the recursion at AVX level, splitting that
107 2018-10-09 David Malcolm <dmalcolm@redhat.com>
109 * genmatch.c (error_cb): Rename to...
110 (diagnostic_cb): ...this, converting int params to enums.
111 (fatal_at): Update for renaming.
112 (warning_at): Likewise.
114 * input.c (selftest::ebcdic_execution_charset::apply):
115 Update for renaming of...
116 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
118 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
119 converting level and reason to enums.
120 (class selftest::lexer_error_sink): Rename to...
121 (class selftest::lexer_test_options): ...this, renaming field
122 "m_errors" to "m_diagnostics".
123 (selftest::lexer_test_options::apply): Update for renaming of...
124 (selftest::lexer_test_options::on_error): ...this, renaming to...
125 (selftest::lexer_test_options::on_diagnostic): ...this
126 converting level and reason to enums.
127 (selftest::test_lexer_string_locations_raw_string_unterminated):
128 Update for renamings.
129 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
132 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
134 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
135 * config/rs6000/pmmintrin.h: New file.
137 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
139 PR tree-optimization/86659
140 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
142 2018-10-09 Richard Biener <rguenther@suse.de>
144 PR tree-optimization/63155
145 * tree-ssa-structalias.c: Include tree-ssa.h.
146 (get_constraint_for_ssa_var): For undefs return nothing_id.
147 (find_func_aliases): Cleanup PHI handling.
149 2018-10-09 Richard Biener <rguenther@suse.de>
151 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
154 2018-10-09 Martin Liska <mliska@suse.cz>
156 * asan.c (asan_emit_stack_protection): If a stack variable
157 is located in a same file as current function, then emit
158 line info into variable definition string.
160 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
162 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
165 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
167 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
170 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
172 PR tree-optimization/86659
173 * gimple-match.h (struct gimple_match_op): Add reverse field.
174 (gimple_match_op::set_op): New overloaded method.
175 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
176 the REF_REVERSE_STORAGE_ORDER flag on the value.
177 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
178 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
180 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
183 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
184 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
186 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
189 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
190 Defined with __builtin_ia32_vfmaddsubpd512_mask.
192 2018-10-08 Richard Biener <rguenther@suse.de>
194 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
195 cost the same as AVX128 ones.
197 2018-10-08 Paul Koning <ni1d@arrl.net>
199 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
200 (expand_block_move): New function.
201 * config/pdp11/pdp11.c (output_block_move): Remove.
202 (expand_block_move): New function.
203 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
204 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
205 (*movmemhi1): Remove.
207 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
209 * config/s390/2827.md: Increase latencies for some FP instructions.
211 2018-10-08 Richard Biener <rguenther@suse.de>
213 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
215 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
216 * tree-vectorizer.h (dump_stmt_cost): Adjust.
217 (add_stmt_cost): Dump return value of the hook.
219 2018-10-08 Richard Biener <rguenther@suse.de>
221 PR tree-optimization/63155
222 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
223 (ssa_propagation_engine::ssa_propagate): Remove redundant
226 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
228 PR rtl-optimization/86939
229 PR rtl-optimization/87479
230 * ira.h (non_conflicting_reg_copy_p): New prototype.
231 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
232 (make_hard_regno_dead): Don't add conflicts for register
233 ignore_reg_for_conflicts.
234 (make_object_dead): Likewise.
235 (non_conflicting_reg_copy_p): New function.
236 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
237 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
238 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
239 (make_hard_regno_dead): Don't add conflicts for register
240 ignore_reg_for_conflicts. Remove special conflict handling of
241 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
242 check_pic_pseudo_p and update callers.
243 (mark_pseudo_dead): Don't add conflicts for register
244 ignore_reg_for_conflicts.
245 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
247 2018-10-05 Andrew Waterman <andrew@sifive.com>
248 Jim Wilson <jimw@sifive.com>
250 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
251 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
252 new pattern using HONOR_SNANS that emits one extra instruction.
254 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
256 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
257 patterns): Merge SI and DI patterns to a GPR pattern.
258 (unnamed define_insn and define_split for record form of that): Merge
259 to a single define_insn_and_split pattern.
261 2018-10-05 David Malcolm <dmalcolm@redhat.com>
264 * input.c (expand_location_to_spelling_point): Add param "aspect"
265 and use rather than hardcoding LOCATION_ASPECT_CARET.
266 (get_substring_ranges_for_loc): Handle the case of a single token
267 within a macro expansion.
268 * input.h (expand_location_to_spelling_point): Add "aspect" param,
269 defaulting to LOCATION_ASPECT_CARET.
271 2018-10-05 Paul Koning <ni1d@arrl.net>
273 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
274 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
275 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
276 (pdp11_guard_type): New function.
278 2018-10-05 Paul Koning <ni1d@arrl.net>
280 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
281 * config/pdp11/pdp11.opt (mfloat32): Remove.
283 * doc/invoke.texi (pdp11 -mfloat32): Remove:
284 (pdp11 -mfloat64): Remove.
286 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
288 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
289 (*cmp<mode>_cc_i387): Ditto.
290 (*cmpu<mode>_cc_i387): Ditto.
291 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
292 * config/i386/i386.c (ix86_expand_fp_compare): Remove
294 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
295 Emit x86_sahf_1 pattern.
296 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
297 (ix86_expand_carry_flag_compare): Ditto.
299 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
301 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
302 to reg_or_0_operand. Add "C" constraint.
303 (*cmpxf_cc_i387): Ditto.
304 (*cmp<mode>_i387): Change operand 2 predicate
305 to nonimm_or_0_operand. Add "C" constraint.
306 (*cmp<mode>_cc_i387): Ditto.
307 (*cmp<mode>_0_i387): Remove insn pattern.
308 (*cmp<mode>_0_cc_i387): Ditto.
310 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
312 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
313 * config/i386/predicates.md (nonimm_or_0_operand): Rename
314 from vector_move_operand. Update all uses.
316 2018-10-05 Martin Sebor <msebor@redhat.com>
318 PR tree-optimization/87490
319 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
322 2018-10-05 Richard Biener <rguenther@suse.de>
324 PR tree-optimization/63155
325 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
326 vertical space in dumpfiles.
327 * tree-ssa-propagate.h
328 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
329 * tree-ssa-propagate.c (cfg_blocks_back): New global.
330 (ssa_edge_worklist_back): Likewise.
331 (curr_order): Likewise.
332 (cfg_blocks_get): Remove abstraction.
333 (cfg_blocks_add): Likewise.
334 (cfg_blocks_empty_p): Likewise.
335 (add_ssa_edge): Add to current or next worklist based on
337 (add_control_edge): Likewise.
338 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
340 (ssa_propagation_engine::ssa_propagate): ... here. Unify
341 iteration from CFG and SSA edge worklist so we process
342 everything in RPO order, prioritizing forward progress
344 (ssa_prop_init): Allocate new worklists, do not dump
346 (ssa_prop_fini): Free new worklists.
348 2018-10-05 Richard Biener <rguenther@suse.de>
350 * tree-core.h (tree_block::abstract_flag): Remove.
351 (tree_block::block_num): Make full 32bits.
352 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
353 * tree.h (BLOCK_ABSTRACT): Remove.
354 * dwarf2out.c (gen_lexical_block_die): Remove dead code
355 resulting from BLOCK_ABSTRACT being always false.
356 (gen_inlined_subroutine_die): Likewise.
357 (gen_block_die): Likewise.
358 * tree.c (block_ultimate_origin): Likewise.
359 * tree-pretty-print.c (dump_block_node): Remove code dealing
361 * tree-ssa-live.c (dump_scope_block): Likewise.
362 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
363 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
365 2018-10-05 Richard Biener <rguenther@suse.de>
367 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
368 is asked for initialize mode to the component mode of the
371 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
374 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
376 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
378 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
381 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
383 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
384 to be DImode. When using mffscrn, force the operand to a register.
386 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
388 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
389 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
390 X87MODEF mode iterator.
391 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
392 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
393 X87MODEF mode iterator.
395 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
397 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
400 2018-10-04 David Malcolm <dmalcolm@redhat.com>
402 * Makefile.in (OBJS): Add opt-problem.o.
403 * dump-context.h: Include "selftest.h.
404 (selftest::temp_dump_context): New forward decl.
405 (class dump_context): Make friend of class
406 selftest::temp_dump_context.
407 (dump_context::dump_loc_immediate): New decl.
408 (class dump_pretty_printer): Move here from dumpfile.c.
409 (class temp_dump_context): Move to namespace selftest.
410 (temp_dump_context::temp_dump_context): Add param
411 "forcibly_enable_dumping".
412 (selftest::verify_dumped_text):
413 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
414 (selftest::verify_item):
415 (ASSERT_IS_TEXT): Move here from dumpfile.c.
416 (ASSERT_IS_TREE): Likewise.
417 (ASSERT_IS_GIMPLE): Likewise.
418 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
420 (dump_context::dump_loc_immediate): ...this new function.
421 (class dump_pretty_printer): Move to dump-context.h.
422 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
423 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
424 (temp_dump_context::temp_dump_context): Move to "selftest"
425 namespace. Add param "forcibly_enable_dumping", and use it to
426 conditionalize the use of m_pp;
427 (selftest::verify_dumped_text): Make non-static.
428 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
429 (selftest::verify_item): Make non-static.
430 (ASSERT_IS_TEXT): Move to dump-context.h.
431 (ASSERT_IS_TREE): Likewise.
432 (ASSERT_IS_GIMPLE): Likewise.
433 (selftest::test_capture_of_dump_calls): Pass "true" for new
434 param of temp_dump_context.
435 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
436 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
438 * opt-problem.cc: New file.
439 * opt-problem.h: New file.
440 * optinfo-emit-json.cc
441 (selftest::test_building_json_from_dump_calls): Pass "true" for
442 new param of temp_dump_context.
443 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
444 (optinfo::emit_for_opt_problem): New function.
445 (optinfo::emit): Clarity which emit_item is used.
446 * optinfo.h (optinfo::get_dump_location): New accessor.
447 (optinfo::emit_for_opt_problem): New decl.
448 (optinfo::emit): Make const.
449 * selftest-run-tests.c (selftest::run_tests): Call
450 selftest::opt_problem_cc_tests.
451 * selftest.h (selftest::opt_problem_cc_tests): New decl.
452 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
453 bool to opt_result, converting fprintf messages to
454 opt_result::failure_at calls. Add "stmt" param for use by the
456 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
457 (runtime_alias_check_p): Convert return type from bool to
458 opt_result, converting dump_printf calls to
459 opt_result::failure_at, using the statement DDR_A for their
461 (find_data_references_in_stmt): Convert return type from bool to
462 opt_result, converting "return false" to opt_result::failure_at
464 * tree-data-ref.h: Include "opt-problem.h".
465 (dr_analyze_innermost): Convert return type from bool to opt_result,
466 and add a const gimple * param.
467 (find_data_references_in_stmt): Convert return type from bool to
469 (runtime_alias_check_p): Likewise.
470 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
471 dr_analyze_innermost.
472 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
473 Convert return type from bool to opt_result, adding a message for
474 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
475 (vect_analyze_data_ref_dependence): Convert return type from bool
476 to opt_result. Change sense of return type from "false"
477 effectively meaning "no problems" to "false" meaning a problem,
478 so that "return false" becomes "return opt_result::success".
479 Convert "return true" calls to opt_result::failure_at, using
480 the location of statement A rather than vect_location.
481 (vect_analyze_data_ref_dependences): Convert return type from bool
483 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
484 calls to opt_result::failure_at, using the stmt location rather
486 (vect_verify_datarefs_alignment): Convert return type from bool
488 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
489 into multiple more-tightly-scoped copies.
490 (vect_analyze_data_refs_alignment): Convert return type from bool
492 (vect_analyze_data_ref_accesses): Likewise, converting a
493 "return false" to a "return opt_result::failure_at", adding a
495 (vect_prune_runtime_alias_test_list): Convert return type from
496 bool to opt_result, converting dump_printf_loc to
497 opt_result::failure_at. Add a %G to show the pertinent statement,
498 and use the stmt's location rather than vect_location.
499 (vect_find_stmt_data_reference): Convert return type from
500 bool to opt_result, converting dump_printf_loc to
501 opt_result::failure_at, using stmt's location.
502 (vect_analyze_data_refs): Convert return type from bool to
503 opt_result. Convert "return false" to "return
504 opt_result::failure_at", adding messages as needed.
505 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
506 type from bool to opt_result.
507 (vect_determine_vf_for_stmt): Likewise.
508 (vect_determine_vectorization_factor): Likewise, converting
509 dump_printf_loc to opt_result::failure_at, using location of phi
510 rather than vect_location.
511 (vect_analyze_loop_form_1): Convert return type from bool to
512 opt_result, converting dump_printf_loc calls, retaining the use of
514 (vect_analyze_loop_form): Convert return type from loop_vec_info
515 to opt_loop_vec_info.
516 (vect_analyze_loop_operations): Convert return type from bool to
517 opt_result, converting dump_printf_loc calls, using the location
518 of phi/stmt rather than vect_location where available. Convert
519 various "return false" to "return opt_result::failure_at" with
520 "unsupported phi" messages.
521 (vect_get_datarefs_in_loop): Convert return type from bool to
522 opt_result. Add a message for the
523 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
524 (vect_analyze_loop_2): Convert return type from bool to
525 opt_result. Ensure "ok" is set to a opt_result::failure_at before
526 each "goto again;", adding new messages where needed.
527 Add "unsupported grouped {store|load}" messages.
528 (vect_analyze_loop): Convert return type from loop_vec_info to
530 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
532 * tree-vect-stmts.c (process_use): Likewise, converting
533 dump_printf_loc call and using stmt location, rather than
535 (vect_mark_stmts_to_be_vectorized): Likeise.
536 (vect_analyze_stmt): Likewise, adding a %G.
537 (vect_get_vector_types_for_stmt): Convert return type from bool to
538 opt_result, converting dump_printf_loc calls and using stmt
539 location, rather than vect_location.
540 (vect_get_mask_type_for_stmt): Convert return type from tree to
541 opt_tree, converting dump_printf_loc calls and using stmt location.
542 * tree-vectorizer.c: Include "opt-problem.h.
543 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
544 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
545 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
546 enabled, use it to report at the top level "couldn't vectorize
547 loop" followed by the problem.
548 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
549 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
551 (vect_analyze_stmt): Likewise.
552 (vect_get_vector_types_for_stmt): Likewise.
553 (tree vect_get_mask_type_for_stmt): Likewise.
554 (vect_analyze_data_ref_dependences): Likewise.
555 (vect_enhance_data_refs_alignment): Likewise.
556 (vect_analyze_data_refs_alignment): Likewise.
557 (vect_verify_datarefs_alignment): Likewise.
558 (vect_analyze_data_ref_accesses): Likewise.
559 (vect_prune_runtime_alias_test_list): Likewise.
560 (vect_find_stmt_data_reference): Likewise.
561 (vect_analyze_data_refs): Likewise.
562 (vect_analyze_loop): Convert return type from loop_vec_info to
564 (vect_analyze_loop_form): Likewise.
565 (vect_analyze_slp): Convert return type from bool to opt_result.
567 2018-10-04 David Malcolm <dmalcolm@redhat.com>
569 * doc/invoke.texi (-fopt-info): Document new "internals"
571 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
572 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
574 (optinfo_verbosity_options): Add "internals".
575 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
576 (dump_context::apply_dump_filter_p): New member function.
577 (dump_context::dump_loc): Use apply_dump_filter_p rather than
578 explicitly masking the dump_kind.
579 (dump_context::begin_scope): Increment the scope depth first. Use
580 apply_dump_filter_p rather than explicitly masking the dump_kind.
581 (dump_context::emit_item): Use apply_dump_filter_p rather than
582 explicitly masking the dump_kind.
583 (dump_dec): Likewise.
584 (dump_hex): Likewise.
585 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
586 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
587 (opt_info_switch_p): Update handling of default
588 MSG_OPTIMIZED_LOCATIONS to cope with default of
589 MSG_PRIORITY_USER_FACING.
590 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
591 masking the dump_kind.
592 (selftest::test_capture_of_dump_calls): Update test_dump_context
593 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
594 than MSG_ALL. Generalize scope test to be run at all four
595 combinations of with/without MSG_PRIORITY_USER_FACING and
596 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
597 for each of the two values.
598 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
599 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
600 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
601 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
602 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
604 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
605 dump messages as MSG_PRIORITY_USER_FACING.
606 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
607 about the interaction with MSG_PRIORITY_*.
609 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
611 * varasm.c (output_constant): Add new parameter merge_strings.
612 Make strings properly zero terminated in merge string sections.
613 (mergeable_string_section): Don't fail if the last char is non-zero.
614 (assemble_variable_contents): Handle merge string sections.
615 (assemble_variable): Likewise.
616 (assemble_constant_contents): Likewise.
617 (output_constant_def_contents): Likewise.
618 (output_constructor_array_range,
619 output_constructor_regular_field): Adjust call to output_constant.
620 (output_object_block): Adjust call to assemble_constant_contents
621 and assemble_variable_contents.
623 2018-10-04 Martin Liska <mliska@suse.cz>
626 * cgraphunit.c (process_function_and_variable_attributes):
627 Warn about a function with alias attribute and a body.
629 2018-10-04 Martin Liska <mliska@suse.cz>
632 * multiple_target.c (redirect_to_specific_clone): New function.
633 (ipa_target_clone): Use it.
634 * tree-inline.c: Fix comment.
636 2018-10-04 David Malcolm <dmalcolm@redhat.com>
638 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
640 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
641 (gcc::dump_manager::register_pass): New member function, adapted
642 from loop body in gcc::pass_manager::register_pass, adding a
643 call to update_dfi_for_opt_info.
644 (gcc::dump_manager::opt_info_enable_passes): Store the
645 -fopt-info options into the new fields. Move the loop
647 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
649 * dumpfile.h (struct opt_pass): New forward decl.
650 (gcc::dump_manager::register_pass): New decl.
651 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
652 (class gcc::dump_manager): Add fields "m_optgroup_flags",
653 "m_optinfo_flags", and "m_optinfo_filename".
654 * passes.c (gcc::pass_manager::register_pass): Move all of the
655 dump-handling code to gcc::dump_manager::register_pass.
657 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
659 PR rtl-optimization/87466
660 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
661 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
662 * doc/tm.texi: Regenerate.
663 * ira-lives.c (process_bb_node_lives): Use the new target hook.
664 * lra-lives.c (process_bb_lives): Likewise.
665 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
668 2018-10-04 Tamar Christina <tamar.christina@arm.com>
670 * params.c (add_params): Fix initialization.
672 2018-10-04 Martin Liska <mliska@suse.cz>
674 PR gcov-profile/84107
675 * tree-profile.c (init_ic_make_global_vars):
676 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
677 Come up with new ic_tuple* variables. Emit
678 __gcov_indirect_call{,_topn} variables.
679 (gimple_gen_ic_profiler): Access the variable
681 (gimple_gen_ic_func_profiler): Access
682 __gcov_indirect_call.callee field.
683 (gimple_init_gcov_profiler): Use ptr_type_node.
684 * value-prof.c (gimple_ic): Use ptr_type_node.
686 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
688 PR tree-optimization/85787
689 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
690 into this function and add support for detecting multiple phis.
691 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
693 2018-10-04 Martin Liska <mliska@suse.cz>
696 * ipa-inline.c (inline_to_all_callers_1):
697 Call ultimate_alias_target for node being inlined.
699 2018-10-03 Jeff Law <law@redhat.com>
701 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
703 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
704 * tree.h (get_typenode_from_name): Prototype.
706 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
708 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
709 Change operand 2 predicate to nonimmediate_operand.
710 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
712 2018-10-03 Martin Sebor <msebor@redhat.com>
713 Jeff Law <law@redhat.com>
715 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
717 (get_string_length): Detect unterminated arrays.
718 (format_string): Same.
719 (format_directive): Warn about unterminated arrays.
720 (handle_gimple_call): Mark statements with no_warning as needed.
722 2018-10-03 Jim Wilson <jimw@sifive.com>
724 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
725 also define __riscv_abi_rve. Delete trailing white space.
727 2018-10-03 Paul Koning <ni1d@arrl.net>
729 Enable LRA register allocator for PDP11.
730 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
733 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
734 * config/pdp11/pdp11.opt (-mlra): New option.
735 * doc/invoke.texi (PDP-11 Options): Document -mlra.
737 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
739 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
740 (*<absneg:code>extend<mode>xf2): Ditto.
742 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
744 PR tree-optimization/87415
745 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
748 2018-10-02 Jeff Law <law@redhat.com>
750 * gimple-fold.c (get_range_strlen): Only set *nonstr when
751 an unterminated string is discovered. Bubble up range
752 even for unterminated strings.
753 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
754 indicates the string was not terminated via NONSTR.
756 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
758 * tree-vrp.c (extract_range_from_unary_expr): Special case all
760 Do not do anything special for anti-ranges.
762 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
764 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
765 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
767 2018-10-03 Martin Liska <mliska@suse.cz>
769 PR gcov-profile/86109
770 * coverage.c (coverage_begin_function): Do not
771 mark lambdas as artificial.
772 * tree-core.h (struct GTY): Remove tm_clone_flag
773 and introduce new lambda_function.
774 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
776 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
779 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
780 P8_VECTOR and VSX are enabled.
782 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
784 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
785 0x3907 as CPU model number.
787 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
789 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
790 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
791 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
792 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
793 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
794 * config/s390/s390.md: Likewise. Rename also the cpu attribute
795 value from arch12 to z14.
797 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
799 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
801 (isinf<mode>2): Ditto.
803 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
805 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
806 before emitting fxam. Perform calculations in XFmode.
808 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
810 * match.pd (((X /[ex] A) +- B) * A): New transformation.
812 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
815 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
816 * tree.c (signed_or_unsigned_type_for): Handle complex.
818 2018-10-02 Jeff Law <law@redhat.com>
820 * gimple-fold.c (get_range_strlen): Remove dead code.
822 2018-10-02 Martin Sebor <msebor@redhat.com>
823 Jeff Law <law@redhat.com>
825 * builtins.c (unterminated_array): Add new arguments.
826 If argument is not terminated, bubble up size and exact
828 (expand_builtin_strnlen): Detect, avoid expanding
829 and diagnose unterminated arrays.
830 (c_strlen): Fill in offset of start of unterminated strings.
831 * builtins.h (unterminated_array): Update prototype.
833 2018-10-02 Richard Biener <rguenther@suse.de>
835 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
836 of haddv4df, first reduce to SSE width and exploit the fact
837 that we only need element zero with the reduction result.
838 (reduc_plus_scal_v2df): Likewise.
840 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
842 * dojump.h (do_jump): Delete.
843 (do_jump_1): Likewise.
844 (split_comparison): Move around.
845 * dojump.c (do_jump): Make static.
846 (do_jump_1): Likewise.
847 (jumpifnot): Move around.
848 (jumpifnot_1): Likewise.
850 (jumpif_1): Likewise.
851 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
853 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
855 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
856 insns in the delay slot and add_insn_after for the jump insn.
858 2018-10-02 Richard Biener <rguenther@suse.de>
860 * tree-inline.c (expand_call_inline): Use the location of
861 the callee declaration for the inline-entry marker.
862 * final.c (notice_source_line): Remove special-casing of
863 NOTE_INSN_INLINE_ENTRY.
865 2018-10-01 Carl Love <cel@us.ibm.com>
868 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
869 (__builtin_mtfsb0): New.
870 (__builtin_mtfsb1): New.
871 ( __builtin_set_fpscr_rn): New.
872 (__builtin_set_fpscr_drn): New.
873 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
874 (rs6000_expand_set_fpscr_rn_builtin): Add.
875 (rs6000_expand_set_fpscr_drn_builtin): Add.
876 (rs6000_expand_builtin): Add case statement entries for
877 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
878 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
879 RS6000_BUILTIN_MFFSL.
880 (rs6000_init_builtins): Add ftype initialization and def_builtin
881 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
882 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
883 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
884 rs6000_mffscdrn): Add define_insn.
885 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
886 * doc/extend.texi: Add documentation for the builtins.
888 2018-10-01 Richard Biener <rguenther@suse.de>
890 PR tree-optimization/87465
891 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
892 causing branch miscounts.
894 2018-10-01 Tamar Christina <tamar.christina@arm.com>
896 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
897 aarch64_option_default_param): New.
899 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
900 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
901 stack-clash protection validation code.
903 2018-10-01 Tamar Christina <tamar.christina@arm.com>
905 * params.c (validate_param): New.
906 (add_params): Use it.
907 (set_param_value): Refactor param validation into validate_param.
908 (diagnostic.h): Include.
909 * diagnostic.h (diagnostic_ready_p): New.
911 2018-10-01 Tamar Christina <tamar.christina@arm.com>
913 * params.c (set_param_value):
914 Add index of parameter being validated.
915 * common/common-target.def (option_validate_param): New.
916 * common/common-targhooks.h (default_option_validate_param): New.
917 * common/common-targhooks.c (default_option_validate_param): New.
918 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
919 * doc/tm.texi: Regenerate.
921 2018-10-01 Tamar Christina <tamar.christina@arm.com>
924 * config/aarch64/aarch64.c (aarch64_override_options_internal):
925 Add validation for stack-clash parameters and set defaults.
927 2018-10-01 Tamar Christina <tamar.christina@arm.com>
930 * configure.ac: Add stack-clash-protection-guard-size.
931 * doc/install.texi: Document it.
932 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
933 * params.def: Update comment for guard-size.
934 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
935 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
936 * configure: Regenerate.
938 2018-10-01 Tamar Christina <tamar.christina@arm.com>
941 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
942 STACK_DYNAMIC_OFFSET): New.
943 * config/aarch64/aarch64.c (aarch64_layout_frame):
944 Update outgoing args size.
945 (aarch64_stack_clash_protection_alloca_probe_range,
946 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
948 2018-10-01 Tamar Christina <tamar.christina@arm.com>
951 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
953 * target.def (stack_clash_protection_alloca_probe_range): New.
954 (stack_clash_protection_final_dynamic_probe): Remove.
955 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
956 (default_stack_clash_protection_final_dynamic_probe): Remove.
957 * targhooks.c: Likewise.
958 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
959 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
960 * doc/tm.texi: Regenerate.
962 2018-10-01 Tamar Christina <tamar.christina@arm.com>
965 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
966 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
967 aarch64_clamp_to_uimm12_shift): New.
968 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
969 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
971 2018-10-01 Tamar Christina <tamar.christina@arm.com>
974 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
976 2018-10-01 Jeff Law <law@redhat.com>
977 Richard Sandiford <richard.sandiford@linaro.org>
978 Tamar Christina <tamar.christina@arm.com>
981 * config/aarch64/aarch64.md
982 (probe_stack_range): Add k (SP) constraint.
983 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
984 STACK_CLASH_MAX_UNROLL_PAGES): New.
985 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
986 stack probes for stack clash.
987 (aarch64_allocate_and_probe_stack_space): New.
988 (aarch64_expand_prologue): Use it.
989 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
990 (aarch64_sub_sp): Add emit_move_imm optional param.
992 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
994 PR tree-optimization/87261
995 * match.pd: Remove trailing whitespace.
996 Add (x & y) | ~(x | y) -> ~(x ^ y),
997 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
999 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1001 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1003 (ashlsi3_insn): Update instruction constraints.
1004 (ashrsi3_insn): Likewise.
1005 (rotrsi3): Likewise.
1006 (add_shift): Likewise.
1007 * config/arc/constraints.md (Csz): New 32 bit constraint. It
1008 avoids placing in the limm field small constants which, otherwise,
1009 could end into a small instruction.
1011 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1013 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1014 destination register is not odd-even.
1015 (umaddsidi4_split): Likewise.
1017 2018-10-01 Richard Biener <rguenther@suse.de>
1019 * tree-inline.c (expand_call_inline): Store origin of fn
1020 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1021 * tree.c (block_ultimate_origin): Simplify and do some
1024 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
1026 * config/i386/mmx.md (EMMS): New int iterator.
1027 (emms): New int attribute.
1028 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1029 EMMS int iterator. Explicitly declare clobbers.
1030 (mmx_emms): Remove expander.
1032 * config/i386/predicates.md (emms_operation): Remove predicate.
1033 (vzeroall_pattern): New predicate.
1034 (vzeroupper_pattern): Rename from vzeroupper_operation.
1035 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1036 vzeroupper_pattern and vzeroall_pattern predicates.
1038 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
1040 PR rtl-optimization/86939
1041 * ira-lives.c (make_hard_regno_born): Rename from this...
1042 (make_hard_regno_live): ... to this. Remove update to conflict
1043 information. Update function comment.
1044 (make_hard_regno_dead): Add conflict information update. Update
1046 (make_object_born): Rename from this...
1047 (make_object_live): ... to this. Remove update to conflict information.
1048 Update function comment.
1049 (make_object_dead): Add conflict information update. Update function
1051 (mark_pseudo_regno_live): Call make_object_live.
1052 (mark_pseudo_regno_subword_live): Likewise.
1053 (mark_hard_reg_dead): Update function comment.
1054 (mark_hard_reg_live): Call make_hard_regno_live.
1055 (process_bb_node_lives): Likewise.
1056 * lra-lives.c (make_hard_regno_born): Rename from this...
1057 (make_hard_regno_live): ... to this. Remove update to conflict
1058 information. Remove now uneeded check_pic_pseudo_p argument.
1059 Update function comment.
1060 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1061 to conflict information. Update function comment.
1062 (mark_pseudo_live): Remove update to conflict information. Update
1064 (mark_pseudo_dead): Add conflict information update.
1065 (mark_regno_live): Call make_hard_regno_live.
1066 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1067 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1069 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
1072 * config/i386/i386.c (construct_container): Use TImode for
1073 BLKmode values in 2 integer registers.
1075 2018-09-29 Jeff Law <law@redhat.com>
1077 * builtins.c (unterminated_array): Pass in c_strlen_data * to
1078 c_strlen rather than just a tree *.
1079 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1080 Update recursive calls appropriately. If caller did not provide a
1081 suitable data pointer, create a local one. When a non-terminated
1082 string is discovered, bubble up information about the string via the
1083 c_strlen_data object.
1084 * builtins.h (c_strlen): Update prototype.
1085 (c_strlen_data): New structure.
1086 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1087 For a type 2 call, if c_strlen indicates a non-terminated string
1088 use the length of the non-terminated string.
1089 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1091 2018-09-29 Jakub Jelinek <jakub@redhat.com>
1094 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1095 __m512d type for __A argument rather than __m512.
1097 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
1099 * match.pd (simple_comparison): Don't optimize if either operand is
1100 a function pointer when target needs function pointer canonicalization.
1102 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1104 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1105 power5 .. power9 to remove indirection.
1106 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1107 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1108 ASM_CPU_476_SPEC): Delete.
1109 (ASM_CPU_SPEC): Adjust.
1110 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1111 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1113 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1115 * config.in: Delete HAVE_AS_DCI.
1116 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1117 * config/rs6000/rs6000.h: Ditto.
1118 * configure.ac: Delete HAVE_AS_DCI.
1119 * configure: Regenerate.
1121 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1123 * config.in (HAVE_AS_LWSYNC): Delete.
1124 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1125 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1127 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1128 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1130 * configure.ac: Delete HAVE_AS_LWSYNC.
1131 * configure: Regenerate.
1133 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1134 Pierre-Marie de Rodat <derodat@adacore.com>
1136 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1137 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1138 (cgraph_node::create_thunk): Add indirect_offset parameter.
1139 (thunk_adjust): Likewise.
1140 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1141 and initialize the corresponding field with it.
1142 (cgraph_node::dump): Dump indirect_offset field.
1143 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1144 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1145 (thunk_adjust): Add indirect_offset parameter and deal with it.
1146 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1147 pass it to thunk_adjust. Do not call the target hook if it's non-zero
1148 or if the thunk is external or local. Fix formatting. Do not chain
1149 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
1150 if any, in the GIMPLE representation.
1151 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1152 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1153 (input_node): Read indirect_offset field.
1154 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1155 call to thunk_adjust.
1156 * tree-nested.c (struct nesting_info): Add thunk_p field.
1157 (create_nesting_tree): Set it.
1158 (convert_all_function_calls): Copy static chain from targets to thunks.
1159 (finalize_nesting_tree_1): Return early for thunks.
1160 (unnest_nesting_tree_1): Do not finalize thunks.
1161 (gimplify_all_functions): Do not gimplify thunks.
1163 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1165 * opt-suggestions.c (option_proposer::build_option_suggestions):
1166 Release "option_values".
1168 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1170 * coverage.c (get_coverage_counts): Convert problem-reporting dump
1171 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1172 * dumpfile.c (kind_as_string): New function.
1173 (dump_loc): Rather than a hardcoded prefix of "note: ", use
1174 kind_as_string to vary the prefix based on dump_kind.
1175 (selftest::test_capture_of_dump_calls): Update for above.
1177 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1179 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1180 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
1182 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1184 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1185 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1186 INVALID_REGNUM instead of FPSR_REG.
1187 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1188 * config/i386/i386.md: Update comment of FP compares.
1189 (fldenv): Do not clobber FPSR_REG.
1191 2018-09-28 Richard Biener <rguenther@suse.de>
1193 * tree.h (BLOCK_ORIGIN): New.
1194 * omp-expand.c (grid_expand_target_grid_body): Assign
1195 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1196 * tree-inline.c (remap_block): Likewise.
1197 * auto-profile.c (get_function_decl_from_block): Simplify
1198 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1199 * langhooks.c (lhd_print_error_function): Likewise.
1200 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1202 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1203 * tree.c (block_nonartificial_location): Likewise.
1204 (block_ultimate_origin): Likewise.
1205 * tree-pretty-print.c (percent_K_format): Likewise. Remove
1206 no longer needed LTO case.
1208 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
1209 Jan Hubicka <jh@suse.cz>
1210 Martin Jambor <mjambor@suse.cz>
1212 * simplify-rtx.c (simplify_merge_mask): New function.
1213 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1214 same masks are used in op1 or op2.
1215 (test_vec_merge): New function.
1216 (test_vector_ops): Call test_vec_merge.
1218 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1220 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1221 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1222 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1223 (cypress_costs): Set it.
1224 (supersparc_costs): Likewise.
1225 (hypersparc_costs): Likewise.
1226 (leon_cost): Likewise.
1227 (leon3_costs): Likewise.
1228 (sparclet_costs): Likewise.
1229 (ultrasparc_costs): Likewise.
1230 (ultrasparc_costs): Likewise.
1231 (niagara_costs): Likewise.
1232 (niagara2_costs): Likewise.
1233 (niagara3_costs): Likewise.
1234 (niagara4_costs): Likewise.
1235 (niagara7_costs): Likewise.
1236 (m8_costs): Likewise.
1237 (TARGET_CAN_FOLLOW_JUMP): Define.
1238 (pass_work_around_errata::gate): Minor tweak.
1239 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1240 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1241 Automaitcally clear MASK_FSMULD mask for V7 processors.
1242 (sparc_can_follow_jump): New static function.
1243 (output_ubranch): Deal with CROSSING_JUMP_P.
1244 (sparc_use_sched_lookahead): Rewrite using switch statement.
1245 (sparc_issue_rate): Reorder.
1246 (sparc_branch_cost): New function.
1248 2018-09-27 Martin Sebor <msebor@redhat.com>
1250 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1251 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1252 (int_fits_type_p): Same.
1254 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
1256 * config/i386/i386.md (FPCR_REG): Remove.
1257 (UNSPEC_FLDCW): Remove.
1258 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1259 (x86_fldcw_1): Remove insn pattern.
1260 (fnstenv): Do not clobber FPCR_REG.
1262 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1263 (CALL_USED_REGISTERS): Ditto.
1264 (REG_ALLOC_ORDER): Ditto.
1265 (REG_CLASS_CONTENTS): Ditto.
1266 (HI_REGISTER_NAMES): Ditto.
1267 (ADDITIONAL_REGISTER_NAMES): Use defines instead
1268 of numerical constants.
1269 * config/i386/i386.c (regclass_map): Remove fpsr register.
1270 (dbx_register_map): Ditto.
1271 (dbx64_register_map): Ditto.
1272 (svr4_dbx_register_map): Ditto.
1273 (print_reg): Do not handle FPCR_REG.
1275 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
1278 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1279 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1280 Delete, always treat as true.
1281 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1282 Ditto. Simplify remaining code.
1283 * config/powerpcspe/powerpcspe.h: Ditto.
1284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1285 Simplify remaining code.
1286 (rs6000_expand_builtin): Ditto.
1287 * config/rs6000/rs6000.h: Ditto.
1288 * configure.ac: Ditto.
1289 * configure: Regenerate.
1291 2018-09-27 Martin Liska <mliska@suse.cz>
1293 * coverage.c (get_coverage_counts): Revert the formatting
1294 of missing profile opt info.
1296 2018-09-27 Richard Biener <rguenther@suse.de>
1300 * dwarf2out.c (set_block_origin_self): Do not mark outermost
1301 block as we do not output that.
1302 (gen_inlined_subroutine_die): Elide the originally outermost
1303 block, matching what we do for concrete instances.
1304 (decls_for_scope): Add parameter specifying whether to recurse
1307 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
1308 Tom de Vries <tom@codesourcery.com>
1312 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1313 STORE_FLAG_VALUE == 1.
1315 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
1317 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1318 constant definitions.
1319 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1320 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1321 ("speculation_barrier"): New expander definition.
1323 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
1325 PR gcov-profile/86957
1326 * common.opt: New warning option -Wmissing-profile.
1327 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1328 * doc/invoke.texi: Document -Wmissing-profile.
1330 2018-09-26 Jim Wilson <jimw@sifive.com>
1332 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1333 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1335 2018-09-26 Martin Sebor <msebor@redhat.com>
1337 * tree.c (zerop): Change return type to bool.
1338 (integer_zerop, integer_onep, integer_each_onep): Same.
1339 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1340 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1341 (real_onep, real_minus_onep, chain_index): Same.
1342 (print_type_hash_statistics, type_list_equal): Same.
1343 * tree.h (zerop): Same.
1344 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1345 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1346 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1347 (real_onep, real_minus_onep, chain_index): Same.
1348 (print_type_hash_statistics, type_list_equal): Same.
1350 2018-09-26 Jim Wilson <jimw@sifive.com>
1352 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1354 2018-09-26 Jakub Jelinek <jakub@redhat.com>
1357 * config/i386/i386.c: Include debug.h and dwarf2out.h.
1358 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1361 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
1363 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1365 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1367 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1368 and FP_SECOND_SSE_REGS.
1369 (REG_CLASS_NAMES): Ditto.
1370 (REG_CLASS_CONTENTS): Ditto.
1371 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1372 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1373 (ix86_preferred_output_reload_class): Ditto.
1374 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1375 clobber constraint to "=&f".
1376 (fix_truncdi_i387): Ditto.
1377 (lrintxfdi2): Ditto.
1378 (fistdi2_<rounding>): Ditto.
1379 (fpremxf4_i387): Change "=u" constraint to "=f".
1380 (fprem1xf4_i387): Ditto.
1382 (fptanxf4_i387): Ditto.
1383 (fxtractxf3_i387): Ditto.
1384 (fscalexf4_i387): Ditto.
1385 (atan2xf3): Change "u" constraint to "f".
1386 (fyl2xxf3_i387): Ditto.
1387 (fyl2xp1xf3_i387): Ditto.
1389 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1392 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1393 for removed I387_MASK_PM entity.
1396 2018-09-26 Jeff Law <law@redhat.com>
1398 2018-09-26 Alexey Neyman <stilor@att.net>
1400 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1401 headers are no longer pulled in by <isl/val.h>.
1403 2018-09-26 Richard Biener <rguenther@suse.de>
1406 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1407 or concrete instance DIE to the tree. Create abstract origin
1408 attributes also for concrete instances.
1410 2018-09-26 Alexey Neyman <stilor@att.net>
1412 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1413 headers are no longer pulled in by <isl/val.h>.
1415 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
1417 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1418 Use new helper functions.
1419 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1420 Use new helper functions.
1421 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1422 aarch_mm_needs_release): New declarations.
1423 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1424 aarch_mm_needs_release): New.
1426 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
1428 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1429 (arm32_output_mi_thunk): Deal with long calls.
1431 2018-09-26 Richard Biener <rguenther@suse.de>
1435 * tree-inline.c (expand_call_inline): When the location
1436 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1437 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1438 the inserted BLOCK to make inlined_function_outer_scope_p
1440 * dwarf2out.c (add_call_src_coords_attributes): Do not add
1441 coords for reserved locations.
1443 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1445 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1446 (*call_indirect_nonlocal_sysv<mode>): Ditto.
1447 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1448 (*sibcall_nonlocal_sysv<mode>): Ditto.
1449 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1450 (<bd>_<mode>): Ditto.
1451 (<bd>tf_<mode>): Ditto.
1453 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1455 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1456 control string as a list of templates instead of as C code.
1457 (*altivec_movti): Ditto.
1458 * config/rs6000/darwin.md (movdf_low_di): Ditto.
1460 2018-09-25 Jim Wilson <jimw@sifive.com>
1462 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1463 when target symbol is weak.
1465 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
1468 * builtins.c (unterminated_array): Simplify.
1469 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
1470 where pointer arithmetic is safe.
1472 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1475 * config/rs6000/altivec.md (altivec_vspltb): Use
1476 const_0_to_15_operand instead of u5bit_cint_operand.
1477 (*altivec_vspltb_internal): Ditto.
1478 (altivec_vspltb_direct): Ditto.
1479 (altivec_vsplth): Use const_0_to_7_operand instead of
1481 (*altivec_vsplth_internal): Ditto.
1482 (altivec_vsplth_direct): Ditto.
1483 (altivec_vspltw): Use const_0_to_3_operand instead of
1485 (*altivec_vspltw_internal): Ditto.
1486 (altivec_vspltw_direct): Ditto.
1487 (altivec_vspltsf): Ditto.
1488 (*altivec_vspltsf_internal): Ditto.
1489 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1490 various splats with the proper size immediate. Reorder the various
1491 cases by ascending size of immediate, and put all such together.
1493 2018-09-25 Richard Biener <rguenther@suse.de>
1496 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1498 (maybe_create_die_with_external_ref): Do not create
1499 DW_TAG_imported_unit here.
1500 (add_abstract_origin_attribute): Handle external BLOCK refs.
1501 (dwarf2out_abstract_function): Simplify LTO case.
1502 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1503 rather than using maybe_create_die_with_external_ref.
1505 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
1508 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1509 (frndintxf2_mask_pm_i387): Ditto.
1510 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1511 Enable for !flag_trapping_math.
1512 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1513 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1514 Change operand 1 predicate to nonimmediate_operand.
1515 (attr "i387_cw"): Remove mask_pm.
1516 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1517 (enum ix86_entity): Remove I387_MASK_PM.
1518 * config/i386/i386.c (ix86_i387_mode_needed): Do not
1519 handle I387_MASK_PM.
1520 (ix86_mode_needed): Ditto.
1521 (ix86_mode_after): Ditto.
1522 (ix86_mode_entry): Ditto.
1523 (ix86_mode_exit): Ditto.
1524 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1526 2018-09-25 Jakub Jelinek <jakub@redhat.com>
1528 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1529 to_update_switch_stmts to vNULL instead of calling create on them
1532 2018-09-25 Richard Biener <rguenther@suse.de>
1534 PR tree-optimization/87402
1535 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1536 (visit_phi): Re-instantiate handling of supposed to be VARYING
1537 but non-VARYING backedge value.
1539 2018-09-25 Richard Biener <rguenther@suse.de>
1542 * dwarf2out.c (struct sym_off_pair): New.
1543 (external_die_map): New global.
1544 (lookup_decl_die): When in LTO create DIEs lazily from the
1546 (lookup_block_die): New function, create DIEs lazily in LTO.
1547 (equate_block_to_die): New function.
1548 (dwarf2out_die_ref_for_decl): During WPA get the association
1549 from the external DIE map.
1550 (dwarf2out_register_external_die): Record mapping into the
1552 (maybe_create_die_with_external_ref): New function split out from
1553 DIE generation part of old dwarf2out_register_external_die.
1554 (add_abstract_origin_attribute): Do not return the DIE. When
1555 in LTO reference externals directly.
1556 (dwarf2out_abstract_function): When in LTO ignore calls for
1557 decls with external DIEs (already present abstract instances).
1558 (gen_call_site_die): Adjust.
1559 (add_high_low_attributes): Likewise.
1560 (gen_lexical_block_die): Likewise.
1561 (gen_inlined_subroutine_die): Likewie.
1562 (gen_block_die): Likewise.
1563 (dwarf2out_inline_entry): Likewise.
1564 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1567 2018-09-25 Martin Liska <mliska@suse.cz>
1569 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1570 integers and not by a float value.
1572 2018-09-25 Martin Liska <mliska@suse.cz>
1575 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1577 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1578 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1579 * godump.c (go_format_type): Remove extra parenthesis.
1581 2018-09-25 Martin Liska <mliska@suse.cz>
1583 * alias.c (set_dest_equal_p): Remove unused function.
1584 * config/i386/i386.c (def_builtin_pure2): Likewise.
1585 * diagnostic-show-locus.c (class layout): Remove
1587 (layout::layout): Likewise here.
1588 * dump-context.h (class temp_dump_context): Likewise.
1589 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1590 (add_AT_loclistsptr): Likewise.
1591 (add_AT_offset): Likewise.
1592 (get_AT_hi_pc): Likewise.
1593 (is_comdat_die): Likewise.
1594 (type_is_enum): Likewise.
1595 (ceiling): Likewise.
1596 (add_AT_vms_delta): Likewise.
1597 (is_class_die): Likewise.
1598 * edit-context.c (class line_event): Remove unused field.
1599 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1601 * ipa-cp.c (ipa_get_vr_lat): Likewise.
1602 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1603 (ok_for_base_p_nonstrict): Likewise.
1604 * tree-chrec.c (is_not_constant_evolution): Likewise.
1605 (chrec_fold_poly_cst): Likewise.
1606 * tree-if-conv.c (has_pred_critical_p): Likewise.
1607 * tree-ssa-coalesce.c (print_exprs): Likewise.
1608 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1609 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1610 * tree-vrp.c (value_ranges_intersect_p): Likewise.
1611 (value_range_nonnegative_p): Likewise.
1613 2018-09-25 Martin Liska <mliska@suse.cz>
1615 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1616 Do not handle "GNU Pascal".
1617 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1619 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1620 from documentation. Likewise.
1621 * dbxout.c (dbxout_range_type): Likewise.
1622 * doc/cpp.texi: Likewise.
1623 * doc/extend.texi: Likewise.
1624 * doc/frontends.texi: Likewise.
1625 * doc/invoke.texi: Remove Pascal entry.
1626 * tree.def (CLEANUP_POINT_EXPR): Likewise.
1627 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1629 2018-09-25 Martin Liska <mliska@suse.cz>
1632 * doc/invoke.texi: Document all parameters and remove default
1635 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
1638 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1639 contains HOST_WIDE_INTs when computing its size.
1641 2018-09-24 Jim Wilson <jimw@sifive.com>
1644 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1646 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1648 2018-09-24 Andrew Pinski <apinski@marvell.com>
1650 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1651 access prev before checking it for NULLness in the
1652 AARCH64_FUSE_CMP_BRANCH case.
1654 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
1657 * config/i386/i386.c (rest_of_insert_endbranch): Set
1658 endbr_queued_at_entrance to true and don't insert ENDBR if
1659 x86_function_profiler will be called.
1660 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1662 * config/i386/i386.h (machine_function): Add
1663 endbr_queued_at_entrance.
1665 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1667 * genattrtab.c (mk_attr_alt): Use alternative_mask.
1668 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1670 (check_attr_test): Use alternative_mask.
1671 (get_attr_value): Likewise.
1672 (compute_alternative_mask): Use alternative_mask and XWINT.
1673 (make_alternative_compare): Use alternative_mask.
1674 (attr_alt_subset_p): Use XWINT.
1675 (attr_alt_subset_of_compl_p): Likewise.
1676 (attr_alt_intersection): Use alternative_mask and XWINT.
1677 (attr_alt_union): Likewise.
1678 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1679 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1680 (simplify_test_exp): Use alternative_mask and XWINT.
1681 (write_test_expr): Use alternative_mask and XWINT, adjust bit
1682 number calculation to support 64 bits. Generate code that
1683 checks 64-bit masks.
1684 (main): Use alternative_mask.
1685 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1687 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1690 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1691 RETURN+USE when returning via %r14.
1693 2018-09-24 Martin Liska <mliska@suse.cz>
1695 * gcov.c (output_lines): Print colorization legend
1696 for both flag_use_colors and flag_use_hotness_colors.
1699 2018-09-24 Martin Liska <mliska@suse.cz>
1701 * coverage.c (get_coverage_counts): Use warning_at
1702 with current_function_decl location. Use %qD in warning
1705 2018-09-24 Martin Liska <mliska@suse.cz>
1707 * memory-block.h (memory_block_pool::release): Annotate with
1708 valgrind that the memory is not accessible.
1710 2018-09-24 Martin Liska <mliska@suse.cz>
1713 * asan.c: Make asan_handled_variables extern.
1715 * cfgexpand.c (expand_stack_vars): Make sure
1716 a representative is unpoison if another
1717 variable in the partition is handled by
1718 use-after-scope sanitization.
1720 2018-09-24 Richard Biener <rguenther@suse.de>
1722 PR tree-optimization/63155
1723 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1724 the worklist when the edge of the respective argument isn't
1727 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1729 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1730 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1731 (MASK_CLASS_P): Update for rename.
1732 (MAYBE_MASK_CLASS_P): Ditto.
1733 (REG_CLASS_NAMES): Update.
1734 (REG_CLASS_CONTENT): Update.
1735 * config/i386/i386.c (regclass_map): Update for MASK_REG
1736 and ALL_MASK_REGS rename.
1737 * config/i386/constraints.md (Yk): Update for rename.
1740 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1742 * config/i386/i386.h (enum reg_class): Remove
1743 EVEX_SSE_REGS and MOD4_SSE_REGS.
1744 (REG_CLASS_NAMES): Update.
1745 (REG_CLASS_CONTENT): Update.
1746 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1747 registers as ALL_SSE_REGS.
1748 (ix86_additional_allocno_class_p): Remove.
1749 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1750 (ix86_register_priority): Lower priority of EVEX SSE registers.
1751 Use IN_RANGE macro where appropriate.
1752 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1753 AVX-5124VNNIW checks.
1754 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1755 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1756 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1757 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1758 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1759 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1760 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1761 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1762 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1763 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1764 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1765 * config/i386/constraints.md (Yh): Remove.
1767 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1769 * config/i386/i386.c (regclass_map): Declare integer REX registers
1772 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1774 * doc/service.texi (Service): Switch the fsf.org link to https.
1776 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1779 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1780 Define to speculation_safe_value_not_needed.
1782 2018-09-21 Florian Weimer <fweimer@redhat.com>
1785 * doc/extend.texi (Common Function Attributes): Mention that
1786 noreturn suppresses tail call optimization.
1788 2018-09-21 Jeff Law <law@redhat.com>
1790 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1791 vr_values::cleanup_edges_and_switches.
1792 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1794 (identify_jump_threads): Remove EDGE_IGNORE handling.
1795 (execute_vrp): Move handling of to_remove_edges and
1796 to_update_switch_stmts into vr_values class member functions.
1797 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1798 (to_update_switch_stmts): Likewise.
1799 * vr-values.c: Include cfghooks.h.
1800 (vr_values::vr_values): Initialize to_remove_edges and
1801 to_update_switch_stmts.
1802 (vr_values::~vr_values): Verify to_remove_edges and
1803 to_update_switch_stmts are empty.
1804 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1805 (vr_values::cleanup_edges_and_switches): New member function.
1806 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1807 function. Add new data members.
1809 2018-09-21 David Malcolm <dmalcolm@redhat.com>
1811 PR tree-optimization/87309
1812 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1813 calls with pflags and alt_flags.
1814 (selftest::test_capture_of_dump_calls): Add test of interaction of
1815 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1817 2018-09-21 Olivier Hainque <hainque@adacore.com>
1819 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1821 2018-09-21 Olivier Hainque <hainque@adacore.com>
1823 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1825 2018-09-21 Olivier Hainque <hainque@adacore.com>
1827 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1828 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1829 (VXWORKS_LIBS_RTP): Minor reordering.
1831 2018-09-21 Olivier Hainque <hainque@adacore.com>
1833 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1834 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1836 2018-09-21 Olivier Hainque <hainque@adacore.com>
1838 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1839 (PTRDIFF_TYPE): Likewise.
1841 2018-09-21 Olivier Hainque <hainque@adacore.com>
1843 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1844 triplet, similar to support for VxWorks7.
1845 * config/vxworks-dummy.h: Provide a default definition
1846 of TARGET_VXWORKS64 to 0.
1848 2018-09-21 Olivier Hainque <hainque@adacore.com>
1850 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1851 * config/vxworks-dummy.h: here.
1853 2018-09-21 Olivier Hainque <hainque@adacore.com>
1855 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1857 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
1858 Bo Zhou <zbo.zhou@hisilicon.com>
1860 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1861 * config/aarch64/aarch64-tune.md: Regenerated.
1862 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1863 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1864 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1866 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
1867 Julian Brown <julian@codesourcery.com>
1869 * builtins.c (get_builtin_sync_mem): Handle address spaces.
1871 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1873 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1874 if the call takes a static chain.
1876 2018-09-21 Martin Liska <mliska@suse.cz>
1878 * auto-profile.c (autofdo_source_profile::read): Do not
1880 (read_profile): Do not add working sets.
1881 (read_autofdo_file): Remove sum_all.
1882 (afdo_callsite_hot_enough_for_early_inline): Remove const
1884 * coverage.c (struct counts_entry): Remove gcov_summary.
1885 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1886 do not support GCOV_TAG_PROGRAM_SUMMARY.
1887 (get_coverage_counts): Remove summary and expected
1889 * coverage.h (get_coverage_counts): Likewise.
1890 * doc/gcov-dump.texi: Remove -w option.
1891 * gcov-dump.c (dump_working_sets): Remove.
1892 (main): Do not support '-w' option.
1893 (print_usage): Likewise.
1894 (tag_summary): Likewise.
1895 * gcov-io.c (gcov_write_summary): Do not dump
1897 (gcov_read_summary): Likewise.
1898 (gcov_histo_index): Remove.
1899 (gcov_histogram_merge): Likewise.
1900 (compute_working_sets): Likewise.
1901 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1903 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1904 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1905 (GCOV_HISTOGRAM_SIZE): Remove.
1906 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1907 (struct gcov_summary): Simplify rapidly just
1908 to runs and sum_max fields.
1909 (gcov_histo_index): Remove.
1910 (NUM_GCOV_WORKING_SETS): Likewise.
1911 (compute_working_sets): Likewise.
1912 * gcov-tool.c (print_overlap_usage_message): Remove
1913 trailing empty line.
1914 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1915 (output_lines): Remove program related line.
1916 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1917 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1919 (input_profile_summary): Do not read it.
1920 (merge_profile_summaries): And do not merge it.
1921 (input_symtab): Do not call removed function.
1922 * modulo-sched.c (sms_schedule): Do not print sum_max.
1923 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1924 removed when histogram method was invented.
1925 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1927 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1929 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1930 and dump selected value.
1931 * profile.c (add_working_set): Remove.
1932 (get_working_sets): Likewise.
1933 (find_working_set): Likewise.
1934 (get_exec_counts): Do not work with working sets.
1935 (read_profile_edge_counts): Do not inform as sum_max is removed.
1936 (compute_branch_probabilities): Likewise.
1937 (compute_value_histograms): Remove argument for call of
1938 get_coverage_counts.
1939 * profile.h: Do not make gcov_summary const.
1941 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
1943 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1945 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1947 PR tree-optimization/86990
1948 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1949 Check that the entire merged store group is made of constants only for
1952 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
1954 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1955 (VTABLE_VERIFICATION_SPEC): Likewise.
1956 (SANITIZER_EARLY_SPEC): Likewise.
1957 (SANITIZER_SPEC): Likewise.
1958 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1959 * doc/invoke.texi (Link Options): Document -r.
1961 2018-09-20 Richard Biener <rguenther@suse.de>
1964 * gimplify.c (gimplify_expr): Retain alignment of
1965 addressable lvalue in dereference.
1967 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
1970 * configure.ac: Check for .loc is_stmt support.
1971 * configure, config.in: Rebuilt.
1972 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1975 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1977 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1978 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1981 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1983 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1985 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1988 2018-09-20 Marek Polacek <polacek@redhat.com>
1990 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1992 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1994 PR tree-optimization/87288
1995 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1996 into account when determining PEELING_FOR_NITERS.
1998 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2000 PR tree-optimization/86877
2001 * tree-vect-loop.c (vect_analyze_loop_2): Call
2002 vect_verify_datarefs_alignment.
2004 2018-09-19 Marek Polacek <polacek@redhat.com>
2006 * doc/invoke.texi: Document -Wclass-conversion.
2008 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2010 * config/pa/pa.c (pa_adjust_priority): Delete.
2011 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2013 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2014 (atomic_storehi): Likewise.
2015 (atomic_storesi): Likewise.
2016 (atomic_loaddi): Restore compare and swap exchange loop code.
2018 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
2020 PR rtl-optimization/86902
2021 * combine.c (try_combine): When changing the CC mode used, don't change
2022 an unrelated mode in other_insn to that new CC mode.
2024 2018-09-19 David Malcolm <dmalcolm@redhat.com>
2026 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2027 with %T in place of calls to dump_generic_expr.
2028 (prune_runtime_alias_test_list): Likewise.
2029 (create_runtime_alias_checks): Likewise.
2030 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2031 (vect_analyze_data_ref_dependence): Likewise.
2032 (vect_slp_analyze_data_ref_dependence): Likewise.
2033 (vect_record_base_alignment): Likewise. Use %G in place of call
2034 to dump_gimple_stmt.
2035 (vect_compute_data_ref_alignment): Likewise.
2036 (verify_data_ref_alignment): Likewise.
2037 (vect_find_same_alignment_drs): Likewise.
2038 (vect_analyze_group_access_1): Likewise.
2039 (vect_analyze_data_ref_accesses): Likewise.
2040 (dependence_distance_ge_vf): Likewise.
2041 (dump_lower_bound): Likewise.
2042 (vect_prune_runtime_alias_test_list): Likewise.
2043 (vect_find_stmt_data_reference): Likewise.
2044 (vect_analyze_data_refs): Likewise.
2045 (vect_create_addr_base_for_vector_ref): Likewise.
2046 (vect_create_data_ref_ptr): Likewise.
2047 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2048 (vect_can_advance_ivs_p): Likewise.
2049 (vect_update_ivs_after_vectorizer): Likewise.
2050 (vect_gen_prolog_loop_niters): Likewise.
2051 (vect_prepare_for_masked_peels): Likewise.
2052 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2053 (vect_determine_vectorization_factor): Likewise.
2054 (vect_is_simple_iv_evolution): Likewise.
2055 (vect_analyze_scalar_cycles_1): Likewise.
2056 (vect_analyze_loop_operations): Likewise.
2057 (report_vect_op): Likewise.
2058 (vect_is_slp_reduction): Likewise.
2059 (check_reduction_path): Likewise.
2060 (vect_is_simple_reduction): Likewise.
2061 (vect_create_epilog_for_reduction): Likewise.
2062 (vect_finalize_reduction:): Likewise.
2063 (vectorizable_induction): Likewise.
2064 (vect_transform_loop_stmt): Likewise.
2065 (vect_transform_loop): Likewise.
2066 (optimize_mask_stores): Likewise.
2067 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2068 (vect_split_statement): Likewise.
2069 (vect_recog_over_widening_pattern): Likewise.
2070 (vect_recog_average_pattern): Likewise.
2071 (vect_determine_min_output_precision_1): Likewise.
2072 (vect_determine_precisions_from_range): Likewise.
2073 (vect_determine_precisions_from_users): Likewise.
2074 (vect_mark_pattern_stmts): Likewise.
2075 (vect_pattern_recog_1): Likewise.
2076 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2077 (vect_record_max_nunits): Likewise.
2078 (vect_build_slp_tree_1): Likewise.
2079 (vect_build_slp_tree_2): Likewise.
2080 (vect_print_slp_tree): Likewise.
2081 (vect_analyze_slp_instance): Likewise.
2082 (vect_detect_hybrid_slp_stmts): Likewise.
2083 (vect_detect_hybrid_slp_1): Likewise.
2084 (vect_slp_analyze_operations): Likewise.
2085 (vect_slp_analyze_bb_1): Likewise.
2086 (vect_transform_slp_perm_load): Likewise.
2087 (vect_schedule_slp_instance): Likewise.
2088 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2089 (vect_mark_stmts_to_be_vectorized): Likewise.
2090 (vect_init_vector_1): Likewise.
2091 (vect_get_vec_def_for_operand): Likewise.
2092 (vect_finish_stmt_generation_1): Likewise.
2093 (vect_check_load_store_mask): Likewise.
2094 (vectorizable_call): Likewise.
2095 (vectorizable_conversion): Likewise.
2096 (vectorizable_operation): Likewise.
2097 (vectorizable_load): Likewise.
2098 (vect_analyze_stmt): Likewise.
2099 (vect_is_simple_use): Likewise.
2100 (vect_get_vector_types_for_stmt): Likewise.
2101 (vect_get_mask_type_for_stmt): Likewise.
2102 * tree-vectorizer.c (increase_alignment): Likewise.
2104 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
2106 * doc/rtl.texi: Adjust vec_select description.
2107 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2108 non-constant selectors.
2110 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
2112 * config/aarch64/aarch64-protos.h
2113 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2114 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2115 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2116 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2117 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2118 (AARCH64_FL_PROFILE): Move index so flags are ordered.
2119 (AARCH64_ISA_RCPC8_4): New flag.
2120 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2121 to aarch64_offset_9bit_signed_unscaled_p.
2122 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2124 * config/aarch64/constraints.md (Ust): New constraint.
2125 * config/aarch64/predicates.md.
2126 (aarch64_9bit_offset_memory_operand): New predicate.
2127 (aarch64_rcpc_memory_operand): New predicate.
2129 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
2131 PR rtl-optimization/87361
2132 * rtlanal.c (nonzero_bits1): Revert accidental change.
2134 2018-09-19 Richard Biener <rguenther@suse.de>
2136 PR tree-optimization/87349
2137 PR tree-optimization/87342
2138 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2140 2018-09-18 Marek Polacek <polacek@redhat.com>
2142 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2143 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2145 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2147 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2149 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2151 PR rtl-optimization/86882
2152 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2154 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
2156 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2157 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2159 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
2162 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2164 2018-09-18 Richard Biener <rguenther@suse.de>
2167 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2168 (compute_samebase_partition_bases): Likewise.
2169 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2170 (gimple_can_coalesce_p): Simplify.
2172 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
2174 Handle a library implementation of ffs calling __builtin_ffs.
2175 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2176 (mmix_init_libfuncs): New function: make __builtin_ffs expand
2179 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2181 * diagnostic-show-locus.c (class layout_range): Add field
2183 (layout_range::layout_range): Add "original_idx" param and use it
2184 to initialize new field.
2185 (make_range): Use 0 for original_idx.
2186 (layout::layout): Pass in index to calls to
2187 maybe_add_location_range.
2188 (layout::maybe_add_location_range): Add param "original_idx" and
2189 pass it on to layout_range.
2190 (layout::print_any_labels): Pass on range->m_original_idx to
2192 (gcc_rich_location::add_location_if_nearby): Use 0 for
2194 * gcc-rich-location.h (text_range_label::get_text): Update for new
2196 (range_label_for_type_mismatch::get_text): Likewise.
2198 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2200 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2202 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2204 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2205 format_string_diagnostic_t.
2206 (fmtwarn_n): Likewise.
2207 * substring-locations.c
2208 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2209 (format_warning_n_va): Convert to...
2210 (format_string_diagnostic_t::emit_warning_n_va): ...this.
2211 (format_warning_va): Convert to...
2212 (format_string_diagnostic_t::emit_warning_va): ...this.
2213 (format_warning_at_substring): Convert to...
2214 (format_string_diagnostic_t::emit_warning): ...this.
2215 (format_warning_at_substring_n): Convert to...
2216 (format_string_diagnostic_t::emit_warning_n): ...this.
2217 * substring-locations.h (class format_string_diagnostic_t): New
2219 (format_warning_va): Convert to
2220 format_string_diagnostic_t::emit_warning_va.
2221 (format_warning_n_va): Convert to
2222 format_string_diagnostic_t::emit_warning_n_va.
2223 (format_warning_at_substring): Convert to
2224 format_string_diagnostic_t::emit_warning.
2225 (format_warning_at_substring_n): Convert to
2226 format_string_diagnostic_t::emit_warning_n.
2228 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
2229 Bernd Schmidt <bernds_cb1@t-online.de>
2231 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2234 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2236 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2237 operand 0 predicate to nonimmediate operand.
2238 (rint<mode>2_frndint): Remove insn pattern.
2239 (rint<mode>2): Change operand 1 predicate to general_operand.
2240 Extend operand 1 to XFmode and generate rintxf2 insn.
2241 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2242 Do not use X87MODEF mode macro.
2243 (frndintxf2_<rounding>_i387): Rename from
2244 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
2245 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2246 to XFmode and generate significandxf3 insn.
2248 2018-09-17 Richard Biener <rguenther@suse.de>
2250 PR tree-optimization/87328
2251 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2252 visiting unexecutable backedges when not iterating.
2253 (do_rpo_vn): Mark all edges not executable even when not
2256 2018-09-17 Martin Jambor <mjambor@suse.cz>
2259 * doc/invoke.texi (Warning Options): Likewise.
2261 2018-09-17 Richard Biener <rguenther@suse.de>
2263 PR tree-optimization/87301
2264 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2265 clean EH info from leftover copy assignments.
2267 2018-09-17 Martin Liska <mliska@suse.cz>
2269 PR gcov-profile/85871
2270 * gcov.c (output_intermediate_file): Fix out of bounds
2273 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
2275 * config/arc/arc.c: Object attributes for core4 not reflected
2277 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2280 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2282 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2284 2018-09-17 Martin Liska <mliska@suse.cz>
2286 * doc/gcov.texi: Document new option --use-hotness-colors.
2287 * gcov.c (struct source_info): Declare new field.
2288 (source_info::source_info): Set default for maximum_count.
2289 (print_usage): Add new -q option.
2290 (process_args): Process it.
2291 (accumulate_line_info): Save src->maximum_count.
2292 (output_line_beginning): Make color line number if
2293 flag_use_hotness_colors is set.
2294 (output_line_details): Pass default argument value.
2295 (output_lines): Pass src->maximum_count.
2297 2018-09-17 Martin Liska <mliska@suse.cz>
2299 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2300 Use processor_names table.
2301 * config/i386/i386.c (ix86_default_align): Use
2302 processor_cost_table for alignment values.
2303 (ix86_option_override_internal): Use processor_names.
2304 (ix86_function_specific_print): Likewise.
2305 * config/i386/i386.h (struct processor_costs):
2306 Add alignment values.
2307 (struct ptt): Remove and replace with const char *.
2308 * config/i386/x86-tune-costs.h (struct processor_costs):
2309 Declare default alignments for all costs.
2311 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
2313 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2314 symbolics or VR_VARYING ranges for ABS_EXPR.
2315 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2316 when range will wrap.
2318 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
2321 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2322 before and after a JUMP_TABLE_DATA.
2324 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2327 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2328 when one operand is a function pointer. Use POINTER_TYPE_P and
2329 FUNC_OR_METHOD_TYPE_P.
2330 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2331 * fold-const.c (build_range_check): Likewise.
2332 * match.pd (simple_comparison): Likewise.
2334 2018-09-14 David Malcolm <dmalcolm@redhat.com>
2337 * spellcheck.c (get_edit_distance_cutoff): New function.
2338 (selftest::test_edit_distance_unit_test_oneway): Rename to...
2339 (selftest::test_get_edit_distance_one_way): ...this.
2340 (selftest::test_get_edit_distance_unit): Rename to...
2341 (selftest::test_get_edit_distance_both_ways): ...this.
2342 (selftest::test_edit_distances): Move tests to this new function,
2343 and test some more pairs of strings. Update for above renaming.
2344 (selftest::get_old_cutoff): New function.
2345 (selftest::test_get_edit_distance_cutoff): New function.
2346 (selftest::assert_suggested_for): New function.
2347 (ASSERT_SUGGESTED_FOR): New macro.
2348 (selftest::assert_not_suggested_for): New function.
2349 (ASSERT_NOT_SUGGESTED_FOR): New macro.
2350 (selftest::test_suggestions): New function.
2351 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2352 tests to selftest::test_edit_distances and call it. Add calls to
2353 selftest::test_get_edit_distance_cutoff and
2354 selftest::test_suggestions.
2355 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2356 (best_match::consider): Replace hard-coded cutoff calculation with
2358 (best_match::get_cutoff): New declaration.
2359 (best_match::get_best_meaningful_candidate): Likewise.
2361 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2363 * builtins.c (fold_builtin_strlen): Remove TODO comment.
2365 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2368 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2370 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2371 terminated string literal.
2373 2018-09-14 Martin Sebor <msebor@redhat.com>
2375 * builtins.c (unterminated_array): Handle ARRAY_REF.
2376 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2377 * builtins.h (unterminated_array): Declare extern.
2378 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2380 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2383 2018-09-14 Martin Sebor <msebor@redhat.com>
2384 Jeff Law <law@redhat.com>
2386 * builtins.c (unterminated_array): New.
2387 (expand_builtin_strcpy): Adjust.
2388 (expand_builtin_strcpy_args): Detect unterminated arrays.
2389 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
2390 unterminated arrays.
2391 * gimple-fold.h (get_maxval_strlen): Add argument.
2392 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2394 * gimple-fold.c (get_range_strlen): Add argument.
2395 (get_maxval_strlen): Adjust.
2396 * gimple-fold.h (get_range_strlen): Add argument.
2398 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
2400 * config/i386/movdirintrin.h: Fix copyright year.
2402 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
2404 * reg-stack.c: Include regs.h.
2405 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2406 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2407 FIRST_STACK_REG, not DFmode.
2408 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2409 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
2410 (change stack): Default register mode to the reg_raw_mode of
2411 FIRST_STACK_REG, not DFmode.
2412 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2413 (*swapxf): Rename from swapxf.
2415 2018-09-14 Carl Love <cel@us.ibm.com>
2417 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2418 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2420 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2423 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2426 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
2429 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2431 2018-09-14 Jason Merrill <jason@redhat.com>
2433 Fix --enable-gather-detailed-mem-stats.
2434 * hash-table.c (hash_table_usage): Change from variable to function.
2435 * hash-table.h: Adjust.
2436 * Makefile.in: Add missing dependencies on hash-table.h.
2438 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2440 PR tree-optimization/87259
2442 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2443 execute_cse_reciprocals_1 has tried transforming.
2445 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
2447 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2448 VR_VARYING for PLUS/MINUS_EXPR.
2450 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
2452 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2455 2018-09-14 Richard Biener <rguenther@suse.de>
2458 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2459 bits for the merged partition.
2461 2018-09-13 Martin Sebor <msebor@redhat.com>
2462 Bernd Edlinger <bernd.edlinger@hotmail.de>
2464 * builtins.h (c_srlen): Add argument.
2465 * builtins.c (warn_string_no_nul): New function.
2466 (c_strlen): Add argument and use it. Update recursive calls.
2467 Pass DECL argument to string_constant to get info on non
2468 terminated strings. Update *NONSTR as needed.
2469 (fold_builtin_strlen): Add argument to calls to c_strlen.
2470 Warn for unterminated arrays.
2471 (warn_string_no_null): Add prototype.
2472 * expr.c (string_constant): Update arguments. Update recursive
2473 calls appropriately. Detect missing NUL terminator and outermost
2474 declaration its missing in.
2475 Improve checks for arrays with nonzero lower bound or elements
2476 that are not a single byte. Simplify offset computation.
2477 Simplify checks for non-NUL terminated strings.
2478 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2479 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2481 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
2483 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2485 * fold-const.c (c_getstr): Fix function comment. Remove unused third
2486 argument. Fix range checks.
2487 * fold-const.h (c_getstr): Adjust protoype.
2488 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2489 string is constant but contains no NUL byte.
2491 * expr.c (string_constant): Adjust function comment.
2492 Remove bogus check for zero termination.
2494 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2496 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2497 (get_constant_size): Don't make STRING_CSTs larger than they are.
2498 (check_string_literal): New check function for STRING_CSTs.
2499 (output_constant): Use it.
2501 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
2504 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2506 2018-09-13 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/87263
2509 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2510 (struct unwind_state): Add max_rpo field.
2511 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2512 Compute max_rpo, the max RPO number a block can be backwards reached
2513 from. Re-write non-iterating mode to a RPO ordered worklist approach,
2514 separating it from the iterating mode.
2516 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2518 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2519 (rfs_decision): New scheduling decision.
2521 2018-09-13 Richard Biener <rguenther@suse.de>
2524 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2525 (vn_nary_op_insert_pieces_predicated): Do not write useless
2526 valid_dominated_by_p entry outside of the allocated storage.
2528 2018-09-13 Omar Sandoval <osandov@osandov.com>
2529 Tom de Vries <tdevries@suse.de>
2532 * dwarf2out.c (is_c): New function.
2533 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2535 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
2538 * config/aarch64/aarch64.md (*aarch64_bfxil):
2540 * config/aarch64/constraints.md (Ulc): Define.
2541 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2543 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2546 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2548 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2549 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2550 aarch64_layout_frame call.
2551 (aarch64_expand_epilogue): Likewise.
2552 (aarch64_initial_elimination_offset): Likewise.
2553 (aarch64_get_separate_components): Likewise.
2554 (aarch64_use_return_insn_p): Likewise.
2555 (aarch64_layout_frame): Remove unneeded check.
2557 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2559 * configure.ac: Only append
2560 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2561 gcc_config_arguments if it was never reconfigured or last reconfigure
2562 was with different arguments.
2563 * configure: Regenerated.
2565 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2566 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2569 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2570 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2572 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2574 PR tree-optimization/87287
2575 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2576 X % C == 0 to X % (unsigned) C == 0 optimization to ...
2577 * match.pd (X % C == 0): ... here. New optimization.
2579 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2582 * expr.h (maybe_optimize_mod_cmp): Declare.
2583 * expr.c (mod_inv): New function.
2584 (maybe_optimize_mod_cmp): New function.
2585 (do_store_flag): Use it.
2586 * cfgexpand.c (expand_gimple_cond): Likewise.
2588 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
2589 Julian Brown <julian@codesourcery.com>
2592 * gimplify.c (gimplify_scan_omp_clauses): Set
2593 target_firstprivatize_array_bases in OpenACC parallel and kernels
2594 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2595 OpenACC data regions.
2597 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
2599 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2600 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2602 2018-09-12 Richard Biener <rguenther@suse.de>
2604 PR tree-optimization/87280
2605 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2606 edge but unreachable target.
2607 (do_rpo_vn): For conservatively handling a PHI only mark
2608 the backedge executable but not the block reachable.
2610 2018-09-12 Richard Biener <rguenther@suse.de>
2612 PR tree-optimization/87266
2613 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2616 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2618 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2620 ("trunc<BFP:mode><DFP_ALL:mode>2")
2621 ("trunc<DFP_ALL:mode><BFP:mode>2")
2622 ("extend<BFP:mode><DFP_ALL:mode>2")
2623 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2624 according to the target operand type.
2626 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2627 Andreas Krebbel <krebbel@linux.ibm.com>
2629 PR tree-optimization/86844
2630 * gimple-ssa-store-merging.c
2631 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2632 there are any overlapping stores in between them, make sure they are
2633 also coalesced or we give up completely.
2635 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2638 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2639 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2641 2018-09-12 Tom de Vries <tdevries@suse.de>
2643 * common.opt (gdescribe-dies): Add option.
2644 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2645 attribute for artifical and nameless decls.
2646 (dwarf2out_register_external_die): Add description attribute to
2647 external reference die.
2648 (add_desc_attribute): New functions.
2649 (gen_subprogram_die): Add description attribute to
2650 DW_TAG_call_site_parameter.
2651 * tree-pretty-print.c (print_generic_expr_to_str): New function.
2652 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2653 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2655 (@item -gdescribe-dies): Add.
2657 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2659 * tree-vrp.c (vrp_shift_undefined_p): Remove.
2660 (extract_range_from_binary_expr_1: Call
2661 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2662 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2665 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2667 * gimple-ssa-warn-alloca.c
2668 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2669 field for ALLOCA_BOUND_*_LARGE.
2671 2018-09-11 Nathan Sidwell <nathan@acm.org>
2673 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2675 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2677 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2678 for clobbers. Remove obsolete comment.
2680 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2682 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2683 mpxchk, mpxld and mpxst types.
2684 (define_attr length_immediate): Remove all processing of mpx types.
2685 (define_attr prefix_0f): Ditto.
2686 (define_attr memory): Ditto.
2688 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2690 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2691 (log<mode>2): Change operand 1 predicate to general_operand.
2692 Extend operand 1 to XFmode and generate logxf3 insn.
2693 (log10<mode>2): Change operand 1 predicate to general_operand.
2694 Extend operand 1 to XFmode and generate log10xf3 insn.
2695 (log2<mode>2): Change operand 1 predicate to general_operand.
2696 Extend operand 1 to XFmode and generate log2xf3 insn.
2697 (fyl2xp1_extend<mode>xf3_i387): Remove.
2698 (log1p<mode>2): Change operand 1 predicate to general_operand.
2699 Extend operand 1 to XFmode and generate log1pxf3 insn.
2700 (fxtract_extend<mode>xf3_i387): Remove.
2701 (logb<mode>2): Change operand 1 predicate to general_operand.
2702 Extend operand 1 to XFmode and generate logbxf3 insn.
2703 (ilogb<mode>2): Change operand 1 predicate to general_operand.
2704 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2705 (significand<mode>2): Change operand 1 predicate to general_operand.
2706 Extend operand 1 to XFmode and generate significandxf3 insn.
2708 2018-09-11 Nathan Sidwell <nathan@acm.org>
2710 * gcc.c (perror_with_name, pfatal_with_name): Delete.
2711 (load_specs): Use fatal_error.
2712 (DELETE_IF_ORDINARY, process_command): Use error.
2713 (execute, run_attempt): Use fatal_error.
2715 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
2717 * diagnostic-core.h (sorry_at): New prototype.
2718 * diagnostic.c (sorry_at): New function.
2720 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
2722 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2723 by zero as VR_UNDEFINED.
2725 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2727 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2728 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2729 (<sincos>mode2): New expander.
2730 (sincos_extend<mode>xf3_i387): Remove insn pattern.
2731 (sincos -> sin, cos splitters): Remove splitter patterns.
2732 (sincos<mode>3): Change operand 2 predicate to general_operand.
2733 Extend operand 2 to XFmode and generate sincosxf3 insn.
2734 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2735 Change operand 3 predicate to const1_operand.
2736 (fptan_extend<mode>xf4_i387): Remove insn pattern.
2737 (tanxf2): Update operands in the call to fptanxf4_i387.
2738 (tan<mode>2): Change operand 1 predicate to general_operand.
2739 Extend operand 1 to XFmode and generate tanxf3 insn.
2740 (atan2xf3): Rename from *fpatanxf3_i387.
2741 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2742 (atan2xf3): Remove expander.
2743 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
2744 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2745 (atan<mode>2): Change operand 1 predicate to general_operand.
2746 Extend operand 1 to XFmode and generate atanxf3 insn.
2748 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2750 * config/i386/i386.md (x87/SSE constant load splitter): Use
2751 memory_operand instead of nonimmediate_operand for input operand
2754 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
2756 * config/i386/i386.md (float partial SSE register stall splitter): Move
2757 splitter near its instruction pattern.
2758 (float_extend partial SSE register stall splitter): Ditto.
2759 (float_truncate partial SSE register stall splitter): Ditto.
2761 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
2764 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2765 to speculation_safe_value_not_needed.
2768 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2769 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2770 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2771 leaf_function_p, instead use has_hard_reg_initial_val.
2773 2018-09-09 Nathan Sidwell <nathan@acm.org>
2775 * gcc.h (pfatal_with_name): Don't declare here.
2776 * gcc.c (pfatal_with_name): Make static.
2778 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2780 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2783 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2785 PR rtl-optimization/85458
2786 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2787 priority hook to reduce the priority of EXPR.
2789 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2791 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2792 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2793 with X87_ENABLE_FLOAT. Remove preparation code.
2794 (*float<SWI48:mode><MODEF:mode>2): Rename from
2795 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2796 math using "enabled" attribute.
2797 (*floatdi<MODEF:mode>2_i387): Rename from
2798 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2799 enable for 32bit targets only.
2800 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2802 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2803 as operand 1 predicate. Rewrite as define_insn_and_split.
2804 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2806 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2808 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2809 to fallthru to FLOAT case.
2811 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2814 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2815 around folding of vec_sl to handle out of range shift values.
2817 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2819 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2820 Update callers to gen_fix_trunc<mode>_i387_fisttp
2821 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2822 nonimmediate_operand.
2823 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2824 and corresponding splitters.
2825 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2826 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2827 (fix_truncdi_i387_with_temp): Remove insn pattern
2828 and corresponding splitters.
2829 (fix_trunc<mode>_i387): Change operand 0 predicate to
2830 nonimmediate_operand.
2831 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2832 and corresponding splitters.
2833 (*fistdi2_1): Remove.
2835 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2836 (lrintxfdi2): Remove expander. Reimplement as define_insn.
2837 (*fist<mode>2_1): Remove.
2838 (fist<mode>2): Ditto.
2839 (fist<mode>2_with_temp): Remove insn pattern and corresponding
2841 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
2842 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2843 (fistdi2_<rounding>): Change operand 0 predicate to
2844 nonimmediate_operand.
2845 (fistdi2_<rounding>_with_temp): Remove insn pattern
2846 and corresponding splitters.
2847 (fist<mode>2_<rounding>): Change operand 0 predicate to
2848 nonimmediate_operand.
2849 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2850 and corresponding splitters.
2852 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2854 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
2856 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2859 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2861 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2862 early gimple folding of vec_splat().
2863 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2864 * gimple-fold.h: Add an extern define for tree_vec_extract().
2866 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2868 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2869 wrappers around TREE_TYPE comparisons.
2871 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2874 * config/s390/predicates.md: Add nonsym_memory_operand.
2875 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2876 contains a SYMBOL_REF, load it into an intermediate pseudo.
2877 (s390_emit_compare_and_swap): Legitimize operand.
2878 * config/s390/s390.md: Use the new nonsym_memory_operand
2879 with UNSPECV_CAS patterns.
2881 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2884 * config/s390/s390-passes.def: New file.
2885 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2887 (make_pass_s390_early_mach): Add declaration.
2888 * config/s390/s390.c (make_pass_s390_early_mach):
2889 (s390_option_override): Remove dynamic registration.
2890 * config/s390/t-s390: Add s390-passes.def.
2892 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2894 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2895 Remove UNSPEC_LTREL_BASE check.
2896 (annotate_constant_pool_refs): Likewise.
2897 (find_constant_pool_ref): Likewise.
2898 (find_ltrel_base): Removed.
2899 (replace_ltrel_base): Removed.
2900 (s390_mainpool_finish): Remove replace_ltrel_base call.
2901 (s390_chunkify_start): Remove pending LTREL_BASE logic.
2902 (s390_chunkify_finish): Remove replace_ltrel_base call.
2903 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2905 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
2908 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2909 to speculation_safe_value_not_needed.
2911 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
2912 Bernd Schmidt <bernds_cb1@t-online.de>
2914 * config/nvptx/nvptx-opts.h: New file.
2915 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2916 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2918 (TARGET_SM35): New macro.
2919 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2921 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2923 (misa=): New option.
2924 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2926 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
2928 * config/i386/i386.md (truncdfsf2): Remove expander.
2929 (truncdfsf2_with_temp): Ditto.
2930 (truncxf<mode>2): Ditto.
2931 (*truncdfsf_fast_mixed): Remove insn pattern.
2932 (*truncdfsf_fast_i387): Ditto.
2933 (*truncdfsf_mixed): Ditto.
2934 (*truncdfsf_i387): Ditto.
2935 (*truncdfsf2_i387_1): Ditto.
2936 (*truncxfsf2_mixed): Ditto.
2937 (*truncxfdf2_mixed): Ditto.
2938 (*truncxf<mode>2_i387_noop): Ditto. Update callers
2939 to call gen_truncxf<mode>2 instead.
2940 (*truncxf<mode>2_i387): Remove.
2941 (reg->reg splitters): Remove splitter pattern.
2942 (reg->mem splitters): Ditto.
2944 (truncdfsf2): New insn pattern.
2945 (truncxf<mode>2): Ditto.
2947 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2949 * tree-ssa-math-opts.c (is_mult_by): New function.
2950 (is_square_of): Use the above.
2951 (optimize_recip_sqrt): New function.
2952 (pass_cse_reciprocals::execute): Use the above.
2954 2018-09-05 Richard Biener <rguenther@suse.de>
2957 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2958 to zero-init the emplaced vec.
2960 2018-09-05 Martin Liska <mliska@suse.cz>
2962 PR tree-optimization/87205
2963 * tree-switch-conversion.c (pass_lower_switch::execute):
2964 Group cases for switch statements.
2966 2018-09-05 Richard Biener <rguenther@suse.de>
2968 PR tree-optimization/87217
2969 * tree-ssa-sccvn.c (vuse_valueize): New.
2970 (vn_reference_lookup_pieces): Use it.
2971 (vn_reference_lookup): Likewise.
2973 2018-09-05 Nathan Sidwell <nathan@acm.org>
2976 * stor-layout.c (place_field): Scan forwards to check last
2977 bitfield when ms_bitfield_placement is in effect.
2979 2018-09-05 Richard Biener <rguenther@suse.de>
2982 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2985 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
2986 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2988 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2989 * config.gcc (extra_objs): Build it.
2990 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2992 * config/aarch64/aarch64-passes.def
2993 (pass_tag_collision_avoidance): New pass.
2994 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2995 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2996 (aarch64_classify_address): Remove static qualifier.
2997 (aarch64_address_info, aarch64_address_type): Move to...
2998 * config/aarch64/aarch64-protos.h: ... here.
2999 (make_pass_tag_collision_avoidance): New function.
3000 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3003 2018-09-05 Martin Liska <mliska@suse.cz>
3005 * doc/gcov.texi: Update documentation of humar
3007 * gcov.c (format_count): Print one decimal place, it provides
3008 more fine number of situations like '1G' vs. '1.4G'.
3010 2018-09-05 Martin Liska <mliska@suse.cz>
3013 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3014 * optc-gen.awk: Allow 'Var' for Deprecated options in order
3015 to generate a MASK value.
3017 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
3020 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3021 if frame pointer isn't used.
3022 (compute_frame_pointer_to_fb_displacement): Likewise.
3024 2018-09-04 Jakub Jelinek <jakub@redhat.com>
3027 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3028 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3029 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3030 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3031 and OPTION_MASK_ISA_XSAVEC_UNSET.
3033 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3035 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3036 XOR operations in NAND case.
3038 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3040 * wide-int-range.cc (wide_int_range_convert): New.
3041 * wide-int-range.h (wide_int_range_convert): New.
3042 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3043 code into wide_int_range_convert.
3044 (extract_range_into_wide_ints): Do not munge anti range constants
3045 into the entire domain. Just return the range back.
3047 2018-09-04 Martin Liska <mliska@suse.cz>
3049 * genmatch.c (output_line_directive): Add new argument
3051 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3053 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3055 * doc/invoke.texi (Option Summary): Add whitespace.
3057 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3059 2018-09-04 Richard Biener <rguenther@suse.de>
3061 PR tree-optimization/87211
3062 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3063 backedge value we're supposed to treat as VARYING also number
3064 the PHI to VARYING in case it got a different value-number already.
3066 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3068 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3069 (extract_range_from_binary_expr_1): Do not call
3070 vrp_can_optimize_bit_op.
3071 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3073 (wide_int_range_get_mask_and_bounds): New.
3074 (wide_int_range_optimize_bit_op): New.
3075 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3076 (wide_int_range_bit_and): Same.
3077 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3078 (wide_int_range_optimize_bit_op): New.
3079 (wide_int_range_get_mask_and_bounds): New.
3081 2018-09-04 Richard Biener <rguenther@suse.de>
3083 PR tree-optimization/87176
3084 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3085 variable. When value-numbering a virtual PHI node make sure
3086 to not value-number to the backedge value.
3088 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3090 * doc/extend.texi (Long Long, Hex Floats): Document support for
3091 long long and hex floats in more recent versions of ISO C++.
3093 2018-09-03 Richard Biener <rguenther@suse.de>
3095 PR tree-optimization/87177
3096 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3099 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3101 * bb-reorder.c (edge_order): Convert to C-qsort-style
3102 tri-state comparator.
3103 (reorder_basic_blocks_simple): Change std::stable_sort to
3106 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3108 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3109 tri-state comparator.
3110 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3112 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3114 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
3115 (mergesort): ... here as maximum count for using netsort.
3116 (gcc_qsort): Set nlim to 3 if stable sort is requested.
3117 (gcc_stablesort): New.
3118 * system.h (gcc_stablesort): Declare.
3120 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3122 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3123 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
3124 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
3126 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3128 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3129 lxsdx and stxsdx alternatives.
3130 (*mov<mode>_hardfloat64): Ditto.
3131 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3133 2018-09-03 Richard Biener <rguenther@suse.de>
3135 PR tree-optimization/87200
3136 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3139 2018-09-03 Martin Liska <mliska@suse.cz>
3141 PR tree-optimization/87201
3142 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3143 Fix parenthesis in an expression.
3145 2018-09-03 Richard Biener <rguenther@suse.de>
3147 PR tree-optimization/87197
3148 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3149 visited. CSE the VN_INFO hashtable lookup.
3151 PR tree-optimization/87169
3152 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3153 iterating make sure there's no extra backedges from irreducible
3154 regions feeding the header. Mark the destination block
3157 2018-09-03 Martin Liska <mliska@suse.cz>
3160 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3161 * common/common-targhooks.c (default_get_valid_option_values):
3163 * common/common-targhooks.h (default_get_valid_option_values):
3165 * common/config/i386/i386-common.c: Move processor_target_table
3167 (ix86_get_valid_option_values): New function.
3168 (TARGET_GET_VALID_OPTION_VALUES): New macro.
3169 * config/i386/i386.c (struct ptt): Move to i386-common.c.
3170 (PTA_*): Move all defined masks into i386-common.c.
3171 (ix86_function_specific_restore): Use new processor_cost_table.
3172 * config/i386/i386.h (struct ptt): Moved from i386.c.
3173 (struct pta): Likewise.
3174 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3175 * doc/tm.texi.in: Likewise.
3176 * opt-suggestions.c (option_proposer::suggest_option):
3177 Pass prefix to build_option_suggestions.
3178 (option_proposer::get_completions): Likewise.
3179 (option_proposer::build_option_suggestions): Use the new target
3181 * opts.c (struct option_help_tuple): New struct.
3182 (print_filtered_help): Use the new target hook.
3184 2018-09-03 Martin Liska <mliska@suse.cz>
3187 * predict.c (set_even_probabilities): Add likely_edges
3188 argument and handle cases where we have precisely one
3190 (combine_predictions_for_bb): Catch also likely_edges.
3191 (tree_predict_by_opcode): Handle gswitch statements.
3192 * tree-cfg.h (find_case_label_for_value): New declaration.
3193 (find_taken_edge_switch_expr): Likewise.
3194 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3195 Find pivot in decision tree based on probabily, not by number of
3198 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
3200 * doc/standards.texi (Standards): Update Objective-C reference.
3202 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3204 * doc/install.texi (Prerequisites): Update link for MPC.
3206 2018-09-01 Michael Matz <matz@suse.de>
3208 PR tree-optimization/87074
3209 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3210 PHIs for outer-loop uses.
3212 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3214 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3215 * doc/invoke.texi (C Dialect Options): Ditto.
3217 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3219 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3221 2018-08-31 Richard Biener <rguenther@suse.de>
3223 PR tree-optimization/87168
3224 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3225 (rpo_elim::eliminate_avail): When OP was not visited it must
3228 2018-08-31 David Malcolm <dmalcolm@redhat.com>
3230 * tree-vrp.c (copy_value_range): Convert param "from" from
3231 "value_range *" to "const value_range *".
3232 (range_is_null): Likewise for param "vr".
3233 (range_int_cst_p): Likewise.
3234 (range_int_cst_singleton_p): Likewise.
3235 (symbolic_range_p): Likewise.
3236 (value_ranges_intersect_p): Likewise for both params.
3237 (value_range_nonnegative_p): Likewise for param "vr".
3238 (value_range_constant_singleton): Likewise.
3239 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3240 (extract_range_into_wide_ints): Likewise for param "vr".
3241 (extract_range_from_multiplicative_op): Likewise for params "vr0"
3243 (vrp_can_optimize_bit_op): Likewise.
3244 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3246 (extract_range_from_unary_expr): Likewise.
3247 (debug_value_range): Likewise for param "vr".
3248 (value_range::dump): Add "const" qualifier.
3249 (vrp_prop::check_array_ref): Convert local "vr" from
3250 "value_range *" to "const value_range *".
3251 (vrp_prop::check_mem_ref): Likewise.
3252 (vrp_prop::visit_stmt): Likewise for local "old_vr".
3253 (vrp_intersect_ranges_1): Likewise for param "vr_1".
3254 (vrp_intersect_ranges): Likewise.
3255 (simplify_stmt_for_jump_threading): Likewise for local "vr".
3256 (vrp_prop::vrp_finalize): Likewise.
3257 * tree-vrp.h (value_range::dump): Add "const" qualifier.
3258 (vrp_intersect_ranges): Add "const" qualifier to params as above.
3259 (extract_range_from_unary_expr): Likewise.
3260 (value_range_constant_singleton): Likewise.
3261 (symbolic_range_p): Likewise.
3262 (copy_value_range): Likewise.
3263 (extract_range_from_binary_expr_1): Likewise.
3264 (range_int_cst_p): Likewise.
3265 (vrp_set_zero_nonzero_bits): Likewise.
3266 (range_int_cst_singleton_p): Likewise.
3268 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
3270 * config/aarch64/arm_neon.h (vabsd_s64): New.
3271 (vnegd_s64): Likewise.
3273 2018-08-31 Martin Jambor <mjambor@suse.cz>
3275 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3277 2018-08-31 Martin Liska <mliska@suse.cz>
3279 * ipa-icf.c (sem_item::add_type): Use
3280 sem_item::m_type_hash_cache.
3281 * ipa-icf.h: Move the cache from sem_item_optimizer
3284 2018-08-31 Nathan Sidwell <nathan@acm.org>
3286 * doc/extend.texi (Backwards Compatibility): Remove implicit
3287 extern C leeway of () being (...).
3289 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3291 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3293 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
3297 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3299 2018-08-31 Jakub Jelinek <jakub@redhat.com>
3302 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3303 gen_int_mode. Formatting fixes.
3305 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
3307 * target.def (custom_function_descriptors): Improve documentation.
3308 * doc/tm.texi.in (Trampolines): Expand discussion of function
3309 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3310 beginning of the section.
3311 * doc/tm.texi: Regenerated.
3313 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
3315 * cfg.h (class auto_edge_flag): Spell out the template-id of the
3316 base class in the initializer list. This is a workaround for
3317 building with older GCC.
3318 (class auto_bb_flag): Likewise.
3320 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
3322 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3323 (altivec_vcmpequ<VI_char>_p): Remove star.
3324 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3326 (expand_strncmp_vec_sequence): New function.
3327 (emit_final_str_compare_vec): New function.
3328 (expand_strn_compare): Add support for vector strncmp.
3329 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3330 length specification to bytes.
3331 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3332 (vcmpnezb_p): New pattern.
3333 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3334 for option -mstring-compare-inline-limit.
3336 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
3338 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3339 (PTA_SKYLAKE): Add PTA_AES.
3340 (PTA_GOLDMONT): Likewise.
3342 2018-08-29 Jan Hubicka <jh@suse.cz>
3345 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3346 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3348 2018-08-29 Jan Hubicka <jh@suse.cz>
3350 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3352 (hash_tree): Do not visit TYPE_STUB_DECL.
3353 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3354 stream TYPE_STUB_DECL.
3355 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3356 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3357 after free_lang_data.
3358 (type_in_anonymous_namespace_p): Likewise.
3360 2018-08-29 Jan Hubicka <jh@suse.cz>
3362 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3363 comment that it has to be even number.
3364 (class sreal): Change m_sig type to int32_t.
3365 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3366 int64_t for temporary calculations.
3367 (sreal_verify_basics): Drop one bit from minimum and maximum.
3369 2018-08-30 Richard Biener <rguenther@suse.de>
3371 PR tree-optimization/87147
3372 * tree-ssa-sccvn.c (SSA_VISITED): New function.
3373 (visit_phi): When the degenerate result is from the backedge and
3374 we didn't visit its definition yet drop to VARYING.
3375 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3377 2018-08-29 Jan Hubicka <jh@suse.cz>
3379 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3381 (hash_tree): Likewise.
3383 2018-08-29 Jan Hubicka <jh@suse.cz>
3385 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3386 and TYPE_NEXT_REF_TO.
3388 2018-08-29 Jan Hubicka <jh@suse.cz>
3390 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3391 comment that it has to be even number.
3392 (class sreal): Change m_sig type to int32_t.
3393 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3394 int64_t for temporary calculations.
3395 (sreal_verify_basics): Drop one bit from minimum and maximum.
3397 2018-08-30 Tamar Christina <tamar.christina@arm.com>
3399 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3401 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
3404 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3405 if to_add is negative.
3407 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3410 * builtins.c (c_strlen): Improve range checks.
3412 2018-08-29 Martin Sebor <msebor@redhat.com>
3413 Jeff Law <law@redhat.com>
3415 PR tree-optimization/86714
3416 PR tree-optimization/86711
3417 * builtins.c (c_strlen): Add arguments to call to string_constant.
3418 * expr.c (string_constant): Add argument. Detect missing nul
3419 terminator and outermost declaration it's missing in.
3420 * expr.h (string_constant): Add argument.
3421 * fold-const.c (read_from_constant_string): Add arguments to call to
3423 (c_getstr): Likewise.
3424 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3426 * tree-ssa-strlen.c (get_stridx): Likewise.
3428 2018-08-29 Jan Hubicka <jh@suse.cz>
3430 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3431 Do not stream DECL_VINDEX.
3432 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3433 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3434 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3436 2018-08-29 Richard Biener <rguenther@suse.de>
3438 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3439 virtual operands that are not default defs to honor region
3441 (rpo_vn_valueize): Remove ineffective code here.
3443 2018-08-29 Richard Biener <rguenther@suse.de>
3445 PR tree-optimization/87132
3446 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3447 when skipping defs reachable over backedges.
3449 2018-08-29 Richard Biener <rguenther@suse.de>
3451 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3452 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3453 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3454 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3455 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3456 (vn_reference_lookup_3): Remove use of const_parms.
3457 (free_rpo_vn): Do not free const_parms.
3458 (do_rpo_vn): Do not call init_const_parms.
3459 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3460 SSA_NAME_POINTS_TO_READONLY_MEMORY.
3461 (call_may_clobber_ref_p_1): Likewise.
3463 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
3466 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3468 (-ftree-scev-cprop): Document.
3470 2018-08-29 Jan Hubicka <jh@suse.cz>
3472 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3474 (sreal constructor): Update.
3475 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3476 sreal:operator/): Update.
3478 2018-08-29 Martin Liska <mliska@suse.cz>
3480 * tree-switch-conversion.c (switch_conversion::expand):
3481 Strenghten assumption about gswitch statements.
3483 2018-08-29 Richard Biener <rguenther@suse.de>
3485 PR tree-optimization/87117
3486 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3487 re-value-number released SSA VDEFs.
3489 2018-08-29 Richard Biener <rguenther@suse.de>
3491 PR tree-optimization/87126
3492 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3494 2018-08-28 Jim Wilson <jimw@sifive.com>
3496 * config/riscv/pic.md: Rewrite.
3497 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3499 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3500 (SOFTF, default_load, softload, softstore): New.
3502 2018-08-28 Jeff Law <law@redhat.com>
3504 * fold-const.c (fold_binary_loc): Remove recently added assert.
3506 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
3508 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3509 to OP parmeter of generated function.
3511 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
3513 PR tree-optimization/87009
3514 * match.pd: Add boolean optimizations.
3516 2018-08-28 Martin Sebor <msebor@redhat.com>
3519 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3520 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3521 (pass_walloca::gate): Use it.
3522 (alloca_call_type): Same.
3523 (pass_walloca::execute): Same.
3524 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3526 2018-08-28 David Malcolm <dmalcolm@redhat.com>
3528 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3529 GCC_VERSION for usage of "__gcc_dump_printf__" format from
3532 2018-08-28 Richard Biener <rguenther@suse.de>
3534 PR tree-optimization/87124
3535 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3536 constants before looking up avail.
3538 2018-08-28 Jakub Jelinek <jakub@redhat.com>
3541 * calls.c (maybe_warn_nonstring_arg): Punt early if
3542 warn_stringop_overflow is zero. Don't call get_range_strlen
3543 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3544 Swap comparison operands to have constants on rhs. Only use
3545 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
3546 increment lenrng[0].
3548 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
3550 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3551 use of tree_to_shwi. Remove duplicated test for the size being
3552 a whole number of bytes.
3554 2018-08-28 Richard Biener <rguenther@suse.de>
3556 PR tree-optimization/87117
3557 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3558 Handle removed stmt without LHS (GIMPLE_NOP).
3560 2018-08-28 Richard Biener <rguenther@suse.de>
3562 PR tree-optimization/87117
3563 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3564 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3566 2018-08-28 Richard Biener <rguenther@suse.de>
3568 PR tree-optimization/87117
3569 * tree-ssa-pre.c (compute_avail): Do not make expressions
3570 with predicated values available.
3571 (get_expr_value_id): Assert we do not run into predicated value
3574 2018-08-28 Richard Biener <rguenther@suse.de>
3576 PR tree-optimization/87117
3577 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3578 get virtual operands.
3579 (get_expr_operands): Handle STRING_CST like other decls.
3581 2018-08-28 Martin Liska <mliska@suse.cz>
3583 * tree.h: Update documentation of fndecl_built_in_p
3587 2018-08-27 Jeff Law <law@redhat.com>
3588 PR tree-optimization/87110
3589 * tree-ssa-dse.c (compute_trims): Handle non-constant
3592 2018-08-27 Martin Sebor <msebor@redhat.com>
3594 PR tree-optimization/86914
3595 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3597 2018-08-27 Martin Sebor <msebor@redhat.com>
3599 PR tree-optimization/87112
3600 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3601 the type of the bound argument.
3603 2018-08-27 Jeff Law <law@redhat.com>
3605 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3606 type does not have a TYPE_SIZE_UNIT.
3608 2018-08-27 Steve Ellcey <sellcey@cavium.com>
3610 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3611 with include of backend.h.
3613 2018-08-27 Richard Biener <rguenther@suse.de>
3615 PR tree-optimization/86927
3616 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3617 use const cond reduction code.
3619 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
3621 PR tree-optimization/85758
3622 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3624 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3627 * diagnostic-show-locus.c (class layout_range): Update for
3628 conversion of show_caret_p to a tri-state.
3629 (layout_range::layout_range): Likewise.
3630 (make_range): Likewise.
3631 (layout::maybe_add_location_range): Likewise.
3632 (layout::should_print_annotation_line_p): Don't show annotation
3633 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3634 (layout::get_state_at_point): Update for conversion of
3635 show_caret_p to a tri-state. Bail out early for
3636 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3637 underlining or source colorization.
3638 (gcc_rich_location::add_location_if_nearby): Update for conversion
3639 of show_caret_p to a tri-state.
3640 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3641 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3643 (selftest::test_one_liner_labels): Likewise.
3644 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3645 conversion of show_caret_p to a tri-state.
3646 * pretty-print.c (text_info::set_location): Likewise.
3647 * pretty-print.h (text_info::set_location): Likewise.
3648 * substring-locations.c (format_warning_n_va): Likewise.
3649 * tree-diagnostic.c (default_tree_printer): Likewise.
3650 * tree-pretty-print.c (newline_and_indent): Likewise.
3652 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3655 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3656 line above for line-insertion fix-it hints.
3657 (selftest::test_fixit_insert_containing_newline): Update the
3658 expected results, and add a test with line-numbering enabled.
3660 2018-08-27 Martin Liska <mliska@suse.cz>
3663 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3664 params with DECL_HAS_VALUE_EXPR_P.
3666 2018-08-27 Martin Liska <mliska@suse.cz>
3668 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3669 selected expansion strategy.
3671 2018-08-27 Martin Liska <mliska@suse.cz>
3673 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3674 * builtins.c (is_builtin_fn): Likewise.
3675 * attribs.c (diag_attr_exclusions): Use new function
3676 fndecl_built_in_p and remove check for FUNCTION_DECL if
3678 (builtin_mathfn_code): Likewise.
3679 (fold_builtin_expect): Likewise.
3680 (fold_call_expr): Likewise.
3681 (fold_builtin_call_array): Likewise.
3682 (fold_call_stmt): Likewise.
3683 (set_builtin_user_assembler_name): Likewise.
3684 (is_simple_builtin): Likewise.
3685 * calls.c (gimple_alloca_call_p): Likewise.
3686 (maybe_warn_nonstring_arg): Likewise.
3687 * cfgexpand.c (expand_call_stmt): Likewise.
3688 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3689 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3690 (cgraph_node::verify_node): Likewise.
3691 * cgraphclones.c (build_function_decl_skip_args): Likewise.
3692 (cgraph_node::create_clone): Likewise.
3693 * config/arm/arm.c (arm_insert_attributes): Likewise.
3694 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3695 * dse.c (scan_insn): Likewise.
3696 * expr.c (expand_expr_real_1): Likewise.
3697 * fold-const.c (operand_equal_p): Likewise.
3698 (fold_binary_loc): Likewise.
3699 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3700 * gimple-low.c (lower_stmt): Likewise.
3701 * gimple-pretty-print.c (dump_gimple_call): Likewise.
3702 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3704 * gimple.c (gimple_build_call_from_tree): Likewise.
3705 (gimple_call_builtin_p): Likewise.
3706 (gimple_call_combined_fn): Likewise.
3707 * gimplify.c (gimplify_call_expr): Likewise.
3708 (gimple_boolify): Likewise.
3709 (gimplify_modify_expr): Likewise.
3710 (gimplify_addr_expr): Likewise.
3711 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3712 * ipa-cp.c (determine_versionability): Likewise.
3713 * ipa-fnsummary.c (compute_fn_summary): Likewise.
3714 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3715 * ipa-split.c (visit_bb): Likewise.
3716 (split_function): Likewise.
3717 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3718 * lto-cgraph.c (input_node): Likewise.
3719 * lto-streamer-out.c (write_symbol): Likewise.
3720 * omp-low.c (setjmp_or_longjmp_p): Likewise.
3721 (lower_omp_1): Likewise.
3722 * predict.c (strip_predict_hints): Likewise.
3723 * print-tree.c (print_node): Likewise.
3724 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3725 * trans-mem.c (is_tm_irrevocable): Likewise.
3726 (is_tm_load): Likewise.
3727 (is_tm_simple_load): Likewise.
3728 (is_tm_store): Likewise.
3729 (is_tm_simple_store): Likewise.
3730 (is_tm_abort): Likewise.
3731 (tm_region_init_1): Likewise.
3732 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3733 * tree-cfg.c (verify_gimple_call): Likewise.
3734 (move_stmt_r): Likewise.
3735 (stmt_can_terminate_bb_p): Likewise.
3736 * tree-eh.c (lower_eh_constructs_2): Likewise.
3737 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3738 * tree-inline.c (remap_gimple_stmt): Likewise.
3739 (copy_bb): Likewise.
3740 (estimate_num_insns): Likewise.
3741 (fold_marked_statements): Likewise.
3742 * tree-sra.c (scan_function): Likewise.
3743 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3744 (optimize_stack_restore): Likewise.
3745 (pass_fold_builtins::execute): Likewise.
3746 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3747 (mark_all_reaching_defs_necessary_1): Likewise.
3748 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3749 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3750 (pass_forwprop::execute): Likewise.
3751 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3752 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3753 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3754 * tree-ssa-strlen.c (get_string_length): Likewise.
3755 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3756 (find_func_aliases_for_call): Likewise.
3757 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3758 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3759 * tree-tailcall.c (find_tail_calls): Likewise.
3760 * tree.c (need_assembler_name_p): Likewise.
3761 (free_lang_data_in_decl): Likewise.
3762 (get_call_combined_fn): Likewise.
3763 * ubsan.c (is_ubsan_builtin_p): Likewise.
3764 * varasm.c (incorporeal_function_p): Likewise.
3765 * tree.h (DECL_BUILT_IN): Remove and replace with
3767 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3768 (fndecl_built_in_p): New.
3770 2018-08-27 Martin Liska <mliska@suse.cz>
3772 PR tree-optimization/86847
3773 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3774 Dump also subtree probability.
3775 (switch_decision_tree::do_jump_if_equal): New function.
3776 (switch_decision_tree::emit_case_nodes): Handle special
3777 situations in balanced tree that can be emitted much simpler.
3778 Fix calculation of probabilities that happen in tree expansion.
3779 * tree-switch-conversion.h (struct cluster): Add
3781 (struct simple_cluster): Likewise.
3782 (struct case_tree_node): Add new function has_child.
3783 (do_jump_if_equal): New.
3785 2018-08-27 Martin Liska <mliska@suse.cz>
3787 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3788 Add new argument to bit_test_cluster constructor.
3789 (bit_test_cluster::emit): Set bits really number of values
3791 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3792 probability argument.
3793 * tree-switch-conversion.h (struct bit_test_cluster):
3794 Add m_handles_entire_switch.
3796 2018-08-27 Martin Liska <mliska@suse.cz>
3798 PR tree-optimization/86702
3799 * tree-switch-conversion.c (jump_table_cluster::emit):
3800 Make probabilities even for values in jump table
3801 according to number of cases handled.
3802 (switch_decision_tree::compute_cases_per_edge): Pass
3803 argument to reset_out_edges_aux function.
3804 (switch_decision_tree::analyze_switch_statement): Likewise.
3805 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3808 2018-08-27 Martin Liska <mliska@suse.cz>
3810 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3811 cfun argument explicitly.
3812 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3813 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3814 function gimple_switch_default_bb.
3815 (convert_switch_statements):
3817 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3818 * stmt.c (label_to_block_fn): Use label_to_block and pass
3819 cfun argument explicitly and use gimple_switch_label_bb.
3820 (expand_case): Likewise.
3821 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3822 cfun argument explicitly. Likewise.
3823 (make_edges_bb): Likewise.
3824 (make_cond_expr_edges): Likewise.
3825 (get_cases_for_edge): Likewise.
3826 (make_gimple_switch_edges): Likewise.
3827 (label_to_block_fn): Likewise.
3828 (label_to_block): Likewise.
3829 (make_goto_expr_edges): Likewise.
3830 (make_gimple_asm_edges): Likewise.
3831 (main_block_label): Likewise.
3832 (group_case_labels_stmt): Likewise.
3833 (find_taken_edge_computed_goto): Likewise.
3834 (find_taken_edge_switch_expr): Likewise.
3835 (gimple_verify_flow_info): Likewise.
3836 (gimple_redirect_edge_and_branch): Likewise.
3837 (gimple_switch_label_bb): New function.
3838 (gimple_switch_default_bb): Likewise.
3839 (gimple_switch_edge): Likewise.
3840 (gimple_switch_default_edge): Likewise.
3841 * tree-cfg.h (label_to_block_fn): Remove and replace ...
3842 (label_to_block): ... with this.
3843 (gimple_switch_label_bb): New.
3844 (gimple_switch_default_bb): Likewise.
3845 (gimple_switch_edge): Likewise.
3846 (gimple_switch_default_edge): Likewise.
3847 * tree-cfgcleanup.c (convert_single_case_switch): Use
3848 new gimple functions and pass new argument to label_to_block.
3849 (cleanup_control_flow_bb):
3850 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3851 cfun argument explicitly.
3852 (make_eh_edges): Likewise.
3853 (redirect_eh_dispatch_edge): Likewise.
3854 (lower_resx): Likewise.
3855 (lower_eh_dispatch): Likewise.
3856 (maybe_remove_unreachable_handlers): Likewise.
3857 (unsplit_eh): Likewise.
3858 (cleanup_empty_eh): Likewise.
3859 (verify_eh_edges): Likewise.
3860 (verify_eh_dispatch_edge): Likewise.
3861 * tree-ssa-dom.c (record_edge_info): Likewise.
3862 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3863 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3864 (thread_through_normal_block): Likewise.
3865 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3866 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3867 * tree-switch-conversion.c (switch_conversion::collect): Use new
3869 (switch_conversion::check_final_bb): Likewise.
3870 (switch_conversion::gather_default_values): Pass new argument
3872 (switch_conversion::build_constructors): Likewise.
3873 (switch_decision_tree::compute_cases_per_edge): Use new
3874 gimple_switch_edge function.
3875 (switch_decision_tree::analyze_switch_statement): Pass new argument
3877 (switch_decision_tree::try_switch_expansion): Use
3878 gimple_switch_default_edge.
3879 * tree-vrp.c (find_switch_asserts): Pass new argument
3881 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3882 (vr_values::simplify_switch_using_ranges): Likewise.
3884 2018-08-27 Richard Biener <rguenther@suse.de>
3886 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3887 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3889 * tree-ssa-sccvn.h (struct vn_pval): New structure.
3890 (struct vn_nary_op_s): Add unwind_to member. Add
3891 predicated_values flag and put result into a union together
3892 with a linked list of vn_pval.
3893 (struct vn_ssa_aux): Add name member to make maintaining
3894 a map of SSA name to vn_ssa_aux possible. Remove no longer
3895 needed info, dfsnum, low, visited, on_sccstack, use_processed
3896 and range_info_anti_range_p members.
3897 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3898 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3900 (vn_valueize): New global.
3901 (vn_context_bb): Likewise.
3902 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3903 VN_INFO_PTR_INFO): Remove.
3904 * tree-ssa-sccvn.c: ... (rewrite)
3905 (pass_fre::execute): For -O2+ initialize loops and run
3906 RPO VN in optimistic mode (iterating). For -O1 and -Og
3907 run RPO VN in non-optimistic mode.
3908 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3909 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3910 * doc/invoke.texi (sccvn-max-scc-size): Remove.
3911 (rpo-vn-max-loop-depth): Document.
3912 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3913 when valuezing the VUSE signals we walked out of the region.
3914 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3915 (phi_translate): Set VN context block to use for availability
3917 (compute_avail): Likewise.
3918 (pre_valueize): New function.
3919 (pass_pre::execute): Adjust to the RPO VN API.
3921 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3922 (propagate_constants_for_unrolling): Remove.
3923 (tree_unroll_loops_completely): Perform value-numbering
3924 on the unrolled bodies loop parent.
3926 2018-08-27 Richard Biener <rguenther@suse.de>
3928 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3929 for partial antic compute.
3931 2018-08-27 Jakub Jelinek <jakub@redhat.com>
3933 PR rtl-optimization/87065
3934 * combine.c (simplify_if_then_else): Formatting fix.
3935 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3937 (known_cond): Don't return const_true_rtx for vector modes. Use
3938 CONST0_RTX instead of const0_rtx. Formatting fixes.
3940 2018-08-27 Martin Liska <mliska@suse.cz>
3942 PR gcov-profile/87069
3943 * gcov.c (process_file): Record files already processed
3944 and warn about a file being processed multiple times.
3946 2018-08-27 Martin Liska <mliska@suse.cz>
3949 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3950 Set default values for x_aarch64_*_string strings.
3951 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3952 prefix. For -mabi do not print '=ABI' in help and use
3953 <option_value> format for -msve-vector-bits and -moverride
3956 2018-08-26 Jeff Law <law@redhat.com>
3958 * config/mips/frame-header-opt.c: Include "backend.h" rather than
3961 2018-08-26 Marek Polacek <polacek@redhat.com>
3963 PR c++/87029, Implement -Wredundant-move.
3964 * doc/invoke.texi: Document -Wredundant-move.
3966 2018-08-25 Martin Sebor <msebor@redhat.com>
3968 PR tree-optimization/87059
3969 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3970 to the same type as the other.
3971 * fold-const.c (fold_binary_loc): Assert expectation.
3973 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
3975 * config/darwin.c (machopic_legitimize_pic_address): Clean up
3976 extraneous parentheses, dead code section and formatting.
3978 2018-08-24 David Malcolm <dmalcolm@redhat.com>
3981 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3982 wide enough for jumps in the line-numbering to be visible.
3983 (layout::print_gap_in_line_numbering): New member function.
3984 (layout::calculate_line_spans): When using line numbering, merge
3985 line spans that are only 1 line apart.
3986 (diagnostic_show_locus): When printing line numbers, show gaps in
3987 line numbering directly, rather than printing headers.
3988 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3989 line-numbering with multiple line spans.
3990 (selftest::test_fixit_insert_containing_newline_2): Add test of
3991 line-numbering, in which the spans are close enough to be merged.
3993 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3995 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3996 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
3997 * tree-vrp.c (range_is_nonnull): Remove.
3998 (range_includes_zero_p): Accept value_range instead of min/max.
3999 (extract_range_from_binary_expr_1): Do not early bail on
4001 Use range_includes_zero_p instead of range_is_nonnull.
4002 (extract_range_from_unary_expr): Use range_includes_zero_p instead
4003 of range_is_nonnull.
4004 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
4005 special case VR_ANTI_RANGE.
4006 (vrp_finalize): Same.
4007 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4009 (range_is_nonnull): Remove.
4010 * vr-values.c (vrp_stmt_computes_nonzero): Use
4011 range_includes_zero_p instead of range_is_nonnull.
4012 (extract_range_basic): Pass value_range to range_includes_zero_p
4013 instead of range_is_nonnull.
4015 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
4017 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4018 * emit-rtl.h (rtl_data): Remove return_bnd.
4019 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4020 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4021 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4022 (POINTER_BOUNDS_MODE): Remove definition.
4023 (make_pointer_bounds_mode): Remove.
4024 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4025 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4026 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4027 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4028 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4029 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4030 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4032 * config/i386/i386-modes.def (BND32, BND64): Remove.
4033 * config/i386/i386.c (dbx_register_map): Remove bound registers.
4034 (dbx64_register_map): Ditto.
4035 (svr4_dbx_register_map): Ditto.
4036 (indirect_thunk_bnd_needed): Remove.
4037 (indirect_thunks_bnd_used): Ditto.
4038 (indirect_return_bnd_needed): Ditto.
4039 (indirect_return_via_cx_bnd): Ditto.
4040 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4041 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4042 (output_indirect_thunk): Ditto. Remove need_prefix argument.
4043 (output_indirect_thunk_function): Remove handling of
4044 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4045 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4046 (ix86_save_reg): Remove handling of crtl->return_bnd.
4047 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4048 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4049 and UNSPEC_BNDLX_ADDR.
4050 (ix86_output_indirect_branch_via_reg): Remove handling of
4051 indirect_thunk_prefix_bnd.
4052 (ix86_output_indirect_branch_via_push): Ditto.
4053 (ix86_output_function_return): Ditto.
4054 (ix86_output_indirect_function_return): Ditto.
4055 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4056 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4057 (CALL_USED_REGISTERS): Ditto.
4058 (REG_ALLOC_ORDER): Update for removal of bound registers.
4059 (HI_REGISTER_NAMES): Ditto.
4060 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4061 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4062 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4063 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4064 (FIRST_PSEUDO_REG): Update.
4065 (BND): Remove mode iterator.
4066 * config/i386/predicates.md (bnd_mem_operator): Remove.
4068 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4070 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4073 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4075 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4076 the case in which the permute needs only a single element and
4077 repeats for every vector of the result. Extend that case to
4078 handle variable-length vectors.
4079 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4081 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
4084 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4085 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4087 2018-08-24 Richard Biener <rguenther@suse.de>
4089 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4090 bb_flags_allocated members.
4091 (auto_flag): New RAII class for allocating flags.
4092 (auto_edge_flag): New RAII class for allocating edge flags.
4093 (auto_bb_flag): New RAII class for allocating bb flags.
4094 * cfgloop.c (verify_loop_structure): Allocate temporary edge
4096 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4097 in favor of temporarily allocated BB flag.
4098 * hsa-brig.c: Re-order includes.
4099 * hsa-dump.c: Likewise.
4100 * hsa-regalloc.c: Likewise.
4101 * print-rtl.c: Likewise.
4102 * profile-count.c: Likewise.
4104 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
4107 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4110 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4113 * wide-int-range.cc (wide_int_range_div): Do not ignore result
4114 from wide_int_range_multiplicative_op.
4116 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4118 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4121 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
4123 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4126 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
4128 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4129 full GENERIC expression used for replacement.
4131 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
4133 * tree-vrp.c (abs_extent_range): Remove.
4134 (extract_range_into_wide_ints): Pass wide ints by reference.
4135 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4136 Pass wide ints by reference in all calls to
4137 extract_range_into_wide_ints.
4138 * wide-int-range.cc (wide_int_range_div): New.
4139 * wide-int-range.h (wide_int_range_div): New.
4140 (wide_int_range_includes_zero_p): New.
4141 (wide_int_range_zero_p): New.
4143 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
4145 * config/aarch64/aarch64.md (arches): New enum.
4146 (arch): New enum attr.
4147 (arch_enabled): New attr.
4148 (enabled): Now uses arch_enabled only.
4149 (simd, sve, fp16): Removed attribute.
4150 (fp): Attr now defined in terms of 'arch'.
4151 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4152 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4153 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4154 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4155 attributes into 'arch'.
4156 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4157 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4158 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4159 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4160 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4161 'simd' attribute into 'arch'.
4162 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4163 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4164 Convert use of 'fp' attribute to 'arch'.
4165 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4166 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4168 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4169 (different modes) Merge 'fp' and 'simd' into 'arch'.
4170 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4173 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
4175 PR rtl-optimization/87026
4176 * expmed.c (canonicalize_comparison): If we can no longer create
4177 pseudoregisters, don't.
4179 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
4182 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4184 * config/arm/arm.c (speculation_barrier_libfunc): New static
4186 (arm_init_libfuncs): Initialize it.
4187 (arm_emit_speculation_barrier): New function.
4188 * config/arm/arm.md (speculation_barrier): Call
4189 arm_emit_speculation_barrier for architectures that do not have
4191 (speculation_barrier_insn): Only match on Armv7 or later.
4193 2018-08-23 Richard Biener <rguenther@suse.de>
4196 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4199 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4201 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4202 of single-vector TBLs.
4203 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4206 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4209 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4210 aarch64_evpc_tbl guard.
4212 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
4214 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4217 2018-08-22 Martin Sebor <msebor@redhat.com>
4220 * tree-pretty-print.c (pretty_print_string): Add argument.
4221 (dump_generic_node): Call to pretty_print_string with string size.
4223 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
4225 PR rtl-optimization/86771
4226 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4227 of two SETs into those two SETs, one to be placed at i2, if that SETs
4228 destination is modified between i2 and i3.
4230 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4232 PR tree-optimization/86725
4233 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4235 (vect_analyze_scalar_cycles_1): Check it.
4237 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4239 PR tree-optimization/86725
4240 * tree-vect-loop.c (vect_is_simple_reduction): When treating
4241 an outer loop phi as a double reduction, make sure that the
4242 single user of the phi result is an inner loop phi.
4244 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4246 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4247 grouped stores with gaps to a strided group.
4249 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4251 * tree-vect-stmts.c (get_group_load_store_type)
4252 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4253 first statement in a group.
4255 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4257 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4258 the sequence used in gcc/gcc.c.
4260 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4263 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4266 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4268 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4269 Darwin10-specific unwinder-shim.
4270 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4271 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4272 New to cater for Darwin10 Rosetta.
4274 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4276 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4277 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4279 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4284 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4285 (dwarf2out_switch_text_section): Generate a local label for the
4286 second function sub-section and apply it as the second FDE start
4288 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4289 second sub-section start.
4291 2018-08-22 Richard Biener <rguenther@suse.de>
4293 PR tree-optimization/86988
4294 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4296 2018-08-22 Richard Biener <rguenther@suse.de>
4298 PR tree-optimization/86945
4299 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4301 2018-08-22 Alexandre Oliva <oliva@adacore.com>
4303 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4304 a comment about how uses of r2 for .sdata2 come about.
4306 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
4308 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4310 2018-08-21 Marek Polacek <polacek@redhat.com>
4312 PR c++/86981, Implement -Wpessimizing-move.
4313 * doc/invoke.texi: Document -Wpessimizing-move.
4315 2018-08-21 Jan Hubicka <jh@suse.cz>
4317 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4318 * tree.h (is_redundant_typedef): Remove.
4319 * dwarf2out.c (is_redundant_typedef): Turn into static function.
4321 2018-08-21 Jan Hubicka <jh@suse.cz>
4323 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4326 2018-08-21 Tamar Christina <tamar.christina@arm.com>
4328 * expmed.c (extract_low_bits): Reject invalid subregs early.
4330 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4333 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4336 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4338 * config/vxworks.h: Guard vxworks_asm_out_constructor and
4339 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4340 * config/vxworks.c: Likewise.
4342 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4344 * config/vxworks.c: Set targetm.have_ctors_dtors
4345 if HAVE_INITFINI_ARRAY_SUPPORT.
4346 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4347 if HAVE_INITFINI_ARRAY_SUPPORT.
4349 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4351 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4352 default search path for VxWorks < 7.
4354 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4356 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4357 (get_format_string): Refer to c_getstr.
4359 2018-08-21 Tom de Vries <tdevries@suse.de>
4361 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4362 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4363 (debuginfo_early_stop): Declare.
4364 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4365 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4366 (debuginfo_early_stop): New function.
4367 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4368 and debuginfo_early_stop.
4369 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4370 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4371 (general_init): Call debuginfo_early_init.
4372 (finalize): Call debuginfo_fini.
4373 (do_compile): Call debuginfo_init.
4374 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4376 (@item -fdump-debug, @item -fdump-earlydebug): Add.
4378 2018-08-21 Tom de Vries <tdevries@suse.de>
4380 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4381 flag_dump_noaddr and flag_dump_unnumbered.
4383 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
4385 * wide-int-range.cc (wide_int_range_abs): New.
4386 (wide_int_range_order_set): Rename from wide_int_range_min_max.
4387 * wide-int-range.h (wide_int_range_abs): New.
4388 (wide_int_range_min_max): New.
4389 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4390 case to call wide_int_range_abs.
4391 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4392 (extract_range_from_abs_expr): Delete.
4394 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
4397 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4398 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4401 2018-08-20 Nathan Sidwell <nathan@acm.org>
4402 Jeff Law <law@redhat.com>
4404 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4405 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4407 2018-08-20 David Malcolm <dmalcolm@redhat.com>
4410 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4411 (decl_attributes): Likewise.
4412 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4414 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4415 * diagnostic-core.h (class auto_diagnostic_group): New class.
4416 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4417 (diagnostic_report_diagnostic): Handle the first diagnostics within
4419 (emit_diagnostic): Add auto_diagnostic_group instance.
4421 (inform_n): Likewise.
4422 (warning): Likewise.
4423 (warning_at): Likewise.
4424 (warning_n): Likewise.
4425 (pedwarn): Likewise.
4426 (permerror): Likewise.
4428 (error_n): Likewise.
4429 (error_at): Likewise.
4431 (fatal_error): Likewise.
4432 (internal_error): Likewise.
4433 (internal_error_no_backtrace): Likewise.
4434 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4435 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4436 * diagnostic.h (struct diagnostic_context): Add fields
4437 "diagnostic_group_nesting_depth",
4438 "diagnostic_group_emission_count", "begin_group_cb",
4440 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4441 Add auto_diagnostic_group instance(s).
4442 (find_explicit_erroneous_behavior): Likewise.
4443 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4444 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4445 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4446 (gimplify_va_arg_expr): Likewise.
4447 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4448 (HSA_SORRY_AT): Likewise.
4449 * ipa-devirt.c (compare_virtual_tables): Likewise.
4450 (warn_odr): Likewise.
4451 * multiple_target.c (expand_target_clones): Likewise.
4452 * opts-common.c (cmdline_handle_error): Likewise.
4453 * reginfo.c (globalize_reg): Likewise.
4454 * substring-locations.c (format_warning_n_va): Likewise.
4455 * tree-inline.c (expand_call_inline): Likewise.
4456 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4457 * tree-ssa-loop-niter.c
4458 (do_warn_aggressive_loop_optimizations): Likewise.
4459 * tree-ssa-uninit.c (warn_uninit): Likewise.
4460 * tree.c (warn_deprecated_use): Likewise.
4462 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
4465 * config/i386/i386.md (eh_return): Always update EH return
4466 address in word_mode.
4468 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
4470 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4471 TARGET_SPLIT_COMPLEX_ARG is defined.
4473 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4475 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4477 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4480 * expr.c (expand_assignment): Assert that bitpos is positive.
4481 (store_field): Likewise
4482 (expand_expr_real_1): Make sure that bitpos is positive.
4483 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4486 2018-08-20 Nathan Sidwell <nathan@acm.org>
4488 * Makefile.in (CPP_ID_DATA_H): Delete.
4489 (CPP_INTERNAL_H): Don't add it.
4490 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4491 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4493 2018-08-20 Richard Biener <rguenther@suse.de>
4495 PR tree-optimization/78655
4496 * tree-vrp.c (extract_range_from_binary_expr_1): Make
4497 pointer + offset nonnull if either operand is nonnull work.
4499 2018-08-20 Tom de Vries <tdevries@suse.de>
4501 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4502 unless the referenced die describes the added property using
4503 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
4504 Otherwise, add a DW_AT_location to the referenced die.
4506 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
4509 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4510 register_operand when calling ix86_set_reg_reg_cost.
4511 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4512 Set *total to 0 for operands that satisfy x86_64_immediate_operand
4513 predicate and to 1 otherwise.
4515 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
4517 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4518 emit a diagnostic that it is not supported and reset the option.
4519 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4520 supported and consume the option. (ASM_FINAL_SPEC): New.
4522 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4524 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4527 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
4529 C-SKY port: Documentation
4531 * doc/extend.texi (C-SKY Function Attributes): New section.
4532 * doc/invoke.texi (Option Summary): Add C-SKY options.
4533 (C-SKY Options): New section.
4534 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4536 2018-08-17 Jojo <jijie_rong@c-sky.com>
4537 Huibin Wang <huibin_wang@c-sky.com>
4538 Sandra Loosemore <sandra@codesourcery.com>
4539 Chung-Lin Tang <cltang@codesourcery.com>
4541 C-SKY port: Backend implementation
4543 * config/csky/*: New.
4544 * common/config/csky/*: New.
4546 2018-08-17 Jojo <jijie_rong@c-sky.com>
4547 Huibin Wang <huibin_wang@c-sky.com>
4548 Sandra Loosemore <sandra@codesourcery.com>
4549 Chung-Lin Tang <cltang@codesourcery.com>
4550 Andrew Jenner <andrew@codesourcery.com>
4552 C-SKY port: Configury
4554 * config.gcc (csky-*-*): New.
4555 * configure.ac: Add csky to targets for dwarf2 debug_line support.
4556 * configure: Regenerated.
4558 2018-08-17 David Malcolm <dmalcolm@redhat.com>
4560 * dump-context.h: Include "dumpfile.h".
4561 (dump_context::dump_printf_va): Convert final param from va_list
4562 to va_list *. Convert from ATTRIBUTE_PRINTF to
4563 ATTRIBUTE_GCC_DUMP_PRINTF.
4564 (dump_context::dump_printf_loc_va): Likewise.
4565 * dumpfile.c: Include "stringpool.h".
4566 (make_item_for_dump_printf_va): Delete.
4567 (make_item_for_dump_printf): Delete.
4568 (class dump_pretty_printer): New class.
4569 (dump_pretty_printer::dump_pretty_printer): New ctor.
4570 (dump_pretty_printer::emit_items): New member function.
4571 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4573 (dump_pretty_printer::emit_item): New member function.
4574 (dump_pretty_printer::stash_item): New member function.
4575 (dump_pretty_printer::format_decoder_cb): New member function.
4576 (dump_pretty_printer::decode_format): New member function.
4577 (dump_context::dump_printf_va): Reimplement in terms of
4578 dump_pretty_printer.
4579 (dump_context::dump_printf_loc_va): Convert final param from va_list
4581 (dump_context::begin_scope): Reimplement call to
4582 make_item_for_dump_printf.
4583 (dump_printf): Update for change to dump_printf_va.
4584 (dump_printf_loc): Likewise.
4585 (selftest::test_capture_of_dump_calls): Convert "stmt" from
4586 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
4587 with %T, %E, and %G.
4588 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4589 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4590 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4591 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4592 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4593 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4594 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4595 within a dump_printf_loc call to "%wu".
4596 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4597 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
4598 missing space after "=".
4599 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4600 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4601 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4602 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4603 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
4604 duplicate "vectorized" from message.
4606 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
4608 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4609 polyNxK_t element's TYPE_STRING_FLAG.
4611 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4613 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4614 (they were unnamed before). Fix comments.
4616 2018-08-17 Nathan Sidwell <nathan@acm.org>
4618 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4620 2018-08-17 Richard Biener <rguenther@suse.de>
4622 PR tree-optimization/86841
4623 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4625 2018-08-17 Martin Liska <mliska@suse.cz>
4627 * common.opt: Remove Warn, Init and Report for options with
4628 Ignore/Deprecated flag. Warning is done automatically for
4630 * config/i386/i386.opt: Likewise.
4631 * config/ia64/ia64.opt: Likewise.
4632 * config/rs6000/rs6000.opt: Likewise.
4633 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4634 Remove usage of flag_check_pointer_bounds.
4635 * lto-wrapper.c (merge_and_complain): Do not handle
4636 OPT_fcheck_pointer_bounds.
4637 (append_compiler_options): Likewise.
4638 * opt-functions.awk: Do not handle Deprecated.
4639 * optc-gen.awk: Check that Var, Report and Init are not
4640 used for an option with Ignore/Deprecated flag.
4641 * opts-common.c (decode_cmdline_option): Do not report
4643 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4645 * opts.h (struct cl_option): Remove cl_deprecated flag.
4646 (CL_ERR_DEPRECATED): Remove error enum value.
4648 2018-08-17 Richard Biener <rguenther@suse.de>
4651 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4652 across a va-arg-pack using call adjust its return value accordingly.
4654 2018-08-16 Martin Sebor <msebor@redhat.com>
4656 PR tree-optimization/86853
4657 * gimple-ssa-sprintf.c (struct format_result): Rename member.
4658 (struct fmtresult): Add member and initialize it in ctors.
4659 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
4660 (format_string): Handle %S the same as %ls. Set MAYFAIL.
4661 (format_directive): Set POSUNDER4K when MAYFAIL is set.
4662 (parse_directive): Handle %C same as %c.
4663 (sprintf_dom_walker::compute_format_length): Adjust.
4664 (is_call_safe): Adjust.
4666 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4668 * builtins.c (c_strlen): Add new parameter eltsize. Use it
4669 for determining how to count the elements.
4670 * builtins.h (c_strlen): Adjust prototype.
4671 * expr.c (string_constant): Add new parameter mem_size.
4672 Set *mem_size appropriately.
4673 * expr.h (string_constant): Adjust protoype.
4674 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4675 * gimple-fold.h (get_range_strlen): Adjust prototype.
4676 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4677 (format_string): Call get_string_length with eltsize.
4679 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4681 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4682 to emit the span, rather than setting it as the prefix.
4684 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4686 * diagnostic-show-locus.c (layout::start_annotation_line): Add
4687 "margin_char" parameter, defaulting to space. Use it in place
4688 of pp_space for the initial part of the margin.
4689 (layout::print_leading_fixits): Use '+' when filling the margin
4690 of line-insertion fix-it hints.
4692 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4694 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4697 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4699 * config/rs6000/altivec.md: Don't set length attribute to the default
4701 * config/rs6000/darwin.md: Ditto.
4702 * config/rs6000/dfp.md: Ditto.
4703 * config/rs6000/htm.md: Ditto.
4704 * config/rs6000/rs6000.md: Ditto.
4705 * config/rs6000/sync.md: Ditto.
4706 * config/rs6000/vsx.md: Ditto.
4708 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4710 * config/rs6000/altivec.md: Don't set length attribute to the default
4711 value, for branch instructions.
4712 * config/rs6000/darwin.md: Ditto.
4713 * config/rs6000/rs6000.md: Ditto.
4715 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4717 * config/rs6000/rs6000.md (length): Always define as const_int 4.
4718 (unnamed conditional branch define_insn): Set length to 4 or 8
4719 depending on offset.
4720 (<bd>_<mode>): Similar, for alternative 0.
4721 (<bd>tf_<mode>): Ditto.
4723 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4725 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4727 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
4729 * doc/rtl.texi: Replace old RTX class names with new names.
4732 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
4734 * expmed.h (canonicalize_comparison): New declaration.
4735 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4736 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4737 * optabs.c (prepare_cmp_insn): Likewise.
4738 * rtl.h (unsigned_condition_p): New function which checks if a
4739 comparison operator is unsigned.
4741 2018-08-16 Nathan Sidwell <nathan@acm.org>
4743 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4744 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4746 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4749 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4750 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4751 (mov<mov>): ..this and enable unconditionally.
4753 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4755 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4757 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
4759 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4760 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4761 "Common" with "Target".
4763 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4765 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4766 * doc/invoke.texi (mmitigate-rop): Remove.
4767 * config/i386/i386.c: Do not include "regrename.h".
4768 (ix86_rop_should_change_byte_p, reg_encoded_number)
4769 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4771 (ix86_reorg): Remove call to ix86_mitigate_rop.
4772 * config/i386/i386.md (attr "modrm_class"): Remove.
4773 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4774 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4775 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4777 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4779 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4780 allow folding of mergeh() and mergel() for the float and double types.
4781 (fold_mergehl_helper): Rework to handle building a permute tree
4784 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4786 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4789 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4791 * common.opt (fdiagnostics-show-labels): New option.
4792 * diagnostic-show-locus.c (class layout_range): Add field
4794 (class layout): Add field "m_show_labels_p".
4795 (layout_range::layout_range): Add param "label" and use it to
4797 (make_range): Pass in NULL for new "label" param of layout_range's
4799 (layout::layout): Initialize m_show_labels_p.
4800 (layout::maybe_add_location_range): Pass in loc_range->m_label
4801 when constructing layout_range instances.
4802 (struct line_label): New struct.
4803 (layout::print_any_labels): New member function.
4804 (layout::print_line): Call it if label-printing is enabled.
4805 (selftest::test_one_liner_labels): New test.
4806 (selftest::test_diagnostic_show_locus_one_liner): Call it.
4807 * diagnostic.c (diagnostic_initialize): Initialize
4808 context->show_labels_p.
4809 * diagnostic.h (struct diagnostic_context): Add field
4811 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4812 -fno-diagnostics-show-labels.
4813 * dwarf2out.c (gen_producer_string): Add
4814 OPT_fdiagnostics_show_labels to the ignored options.
4815 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4817 (gcc_rich_location::maybe_add_expr): Likewise.
4818 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4819 label" param, defaulting to NULL.
4820 (gcc_rich_location::add_expr): Add "label" param.
4821 (gcc_rich_location::maybe_add_expr): Likewise.
4822 (class text_range_label): New class.
4823 (class range_label_for_type_mismatch): New class.
4824 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4825 of format_warning_va.
4826 (fmtwarn_n): Likewise for new params of format_warning_n_va.
4827 * lto-wrapper.c (merge_and_complain): Add
4828 OPT_fdiagnostics_show_labels to the "pick one setting" options.
4829 (append_compiler_options): Likewise to the dropped options.
4830 (append_diag_options): Likewise to the passed-on options.
4831 * opts.c (common_handle_option): Handle the new option.
4832 * selftest-diagnostic.c
4833 (test_diagnostic_context::test_diagnostic_context): Enable
4835 * substring-locations.c: Include "gcc-rich-location.h".
4836 (format_warning_n_va): Add "fmt_label" and "param_label" params
4837 and use them as appropriate.
4838 (format_warning_va): Add "fmt_label" and "param_label" params,
4839 passing them on to format_warning_n_va.
4840 (format_warning_at_substring): Likewise.
4841 (format_warning_at_substring_n): Likewise.
4842 * substring-locations.h (format_warning_va): Add "fmt_label" and
4843 "param_label" params.
4844 (format_warning_n_va): Likewise.
4845 (format_warning_at_substring): Likewise.
4846 (format_warning_at_substring_n): Likewise.
4847 * toplev.c (general_init): Initialize global_dc->show_labels_p.
4849 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
4852 * builtins.c (expand_builtin_memcmp): Do not expand the call
4853 when overflow is detected.
4855 2018-08-15 Martin Sebor <msebor@redhat.com>
4857 PR tree-optimization/71625
4858 * config/aarch64/aarch64-builtins.c
4859 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4861 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
4863 * config/s390/s390.c (s390_reorg): Remove loop.
4865 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4868 (darwin_function_switched_text_sections): Delete.
4869 * gcc/config/darwin.h
4870 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4872 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4875 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4876 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
4877 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4879 2018-08-15 Martin Liska <mliska@suse.cz>
4881 PR tree-optimization/86925
4882 * predict.c (expr_expected_value_1): When taking
4883 later predictor, assign also probability.
4884 Use fold_build2_initializer_loc in order to fold
4885 the expression in -frounding-math.
4887 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
4889 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4891 (expand_vec_perm_1): Try the new method.
4893 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
4896 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4897 Check whether lra_live_max_point is 0 before dividing.
4899 2018-08-14 Martin Sebor <msebor@redhat.com>
4901 PR tree-optimization/86650
4902 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4903 (vrp_prop::check_mem_ref): Same.
4905 2018-08-13 Liu Hao <lh_mouse@126.com>
4907 * pretty-print.c (eat_esc_sequence): Swap the foreground and
4908 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4909 and clear it thereafter, as it only works for DBCS.
4911 2018-08-13 Liu Hao <lh_mouse@126.com>
4913 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4914 handle returned by _get_osf_handle().
4916 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4918 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4919 for folding vec_perm.
4921 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4923 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4924 Add support for gimple-folding of vec_pack() and vec_unpack()
4927 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4929 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4930 vec_xst variants to the list.
4931 (rs6000_gimple_fold_builtin): Add support for folding unaligned
4932 vector loads and stores.
4934 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
4936 * config.gcc (rs6000-ibm-aix4.x): Delete.
4937 (rs6000-ibm-aix5.1): Delete.
4938 (rs6000-ibm-aix5.2): Delete.
4939 (rs6000-ibm-aix5.3): Delete.
4940 * config/rs6000/aix43.h: Delete.
4941 * config/rs6000/aix51.h: Delete.
4942 * config/rs6000/aix52.h: Delete.
4943 * config/rs6000/t-aix43: Delete.
4945 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
4947 * config/s390/s390.c (s390_decompose_constant_pool_ref):
4949 (s390_decompose_address): Factor out constant pool ref
4952 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4954 * config/nds32/nds32-predicates.c
4955 (nds32_can_use_bclr_p): Change return type as bool.
4956 (nds32_can_use_bset_p): Ditto.
4957 (nds32_can_use_btgl_p): Ditto.
4958 (nds32_can_use_bitci_p): Ditto.
4959 * config/nds32/nds32-protos.h
4960 (nds32_can_use_bclr_p): Change declaration.
4961 (nds32_can_use_bset_p): Ditto.
4962 (nds32_can_use_btgl_p): Ditto.
4963 (nds32_can_use_bitci_p): Ditto.
4965 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4967 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4968 Support -msched-prolog-epilog option.
4969 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4971 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4973 * common/config/nds32/nds32-common.c
4974 (nds32_option_optimization_table): Enalbe -malways-align.
4976 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4978 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4980 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4981 OPT_misr_secure_ case.
4982 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4983 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4984 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4986 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4987 (nds32_isr_info): New field security_level.
4988 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4989 * config/nds32/nds32.md (return_internal): Consider critical attribute.
4990 * config/nds32/nds32.opt (misr-secure): New option.
4991 * config/nds32/nds32_init.inc: New file.
4992 * config/nds32/nds32_isr.h: New file.
4994 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
4996 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4997 Update comment for atomic instructions.
4998 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4999 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5001 (atomic_loaddi): Revise fence expansion to only emit fence prior to
5002 load for __ATOMIC_SEQ_CST model.
5003 (atomic_loaddi_1): Remove float register target.
5004 (atomic_storedi): Handle CONST_INT values.
5005 (atomic_storedi_1): Remove float register source. Add special case
5007 (memory_barrier): New expander and insn.
5009 2018-08-11 Jakub Jelinek <jakub@redhat.com>
5011 PR tree-optimization/86835
5012 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5013 new_stmt after def_gsi, make sure to insert new_square_stmt after
5014 that stmt, not 2 stmts before it.
5016 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
5019 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5022 2018-08-10 Martin Liska <mliska@suse.cz>
5025 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5027 * builtins.c (expand_builtin_expect_with_probability):
5029 (expand_builtin_expect_with_probability): New function.
5030 (build_builtin_expect_predicate): Add new argumnet probability
5031 for BUILT_IN_EXPECT_WITH_PROBABILITY.
5032 (fold_builtin_expect):
5035 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5036 * builtins.h (fold_builtin_expect): Set new argument.
5037 * doc/extend.texi: Document __builtin_expect_with_probability.
5038 * doc/invoke.texi: Likewise.
5039 * gimple-fold.c (gimple_fold_call): Pass new argument.
5040 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5041 also BUILT_IN_EXPECT_WITH_PROBABILITY.
5042 * predict.c (get_predictor_value): New function.
5043 (expr_expected_value): Add new argument probability. Assume
5044 that predictor and probability are always non-null.
5045 (expr_expected_value_1): Likewise. For __builtin_expect and
5046 __builtin_expect_with_probability set probability. Handle
5047 combination in binary expressions.
5048 (tree_predict_by_opcode): Simplify code by simply calling
5049 get_predictor_value.
5050 (pass_strip_predict_hints::execute): Add handling of
5051 BUILT_IN_EXPECT_WITH_PROBABILITY.
5052 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5054 * tree.h (DECL_BUILT_IN_P): New function.
5056 2018-08-10 Martin Liska <mliska@suse.cz>
5058 PR tree-optimization/85799
5059 * passes.def: Add argument for pass_strip_predict_hints.
5060 * predict.c (class pass_strip_predict_hints): Add new argument
5062 (strip_predictor_early): New function.
5063 (pass_strip_predict_hints::execute): Call the function to
5065 (strip_predict_hints): New function.
5066 * predict.def: Fix comment.
5068 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
5070 * Makefile.in: Clarify which tm.texi to copy over to assert the
5071 right to grant a GFDL license for all.
5073 2018-08-09 Jeff Law <law@redhat.com>
5075 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5078 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
5080 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5083 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5085 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5086 reductions for variable-length vectors.
5088 2018-08-09 David Malcolm <dmalcolm@redhat.com>
5091 * common.opt (fdiagnostics-show-line-numbers): New option.
5092 * diagnostic-show-locus.c (class layout): Add fields
5093 "m_show_line_numbers_p" and "m_linenum_width";
5094 (num_digits): New function.
5095 (test_num_digits): New function.
5096 (layout::layout): Initialize new fields. Update m_x_offset
5097 logic to handle any left margin.
5098 (layout::print_source_line): Print line number when requested.
5099 (layout::start_annotation_line): New member function.
5100 (layout::print_annotation_line): Call it.
5101 (layout::print_leading_fixits): Likewise.
5102 (layout::print_trailing_fixits): Likewise. Update calls to
5103 move_to_column for new parameter.
5104 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5105 it to potentially call start_annotation_line.
5106 (layout::show_ruler): Call start_annotation_line.
5107 (selftest::test_line_numbers_multiline_range): New selftest.
5108 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5109 and selftest::test_line_numbers_multiline_range.
5110 * diagnostic.c (diagnostic_initialize): Initialize
5111 show_line_numbers_p.
5112 * diagnostic.h (struct diagnostic_context): Add field
5113 "show_line_numbers_p".
5114 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5115 -fno-diagnostics-show-line-numbers.
5116 * dwarf2out.c (gen_producer_string): Add
5117 OPT_fdiagnostics_show_line_numbers to the ignored options.
5118 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5119 one setting" options.
5120 (append_compiler_options): Likewise to the dropped options.
5121 (append_diag_options): Likewise to the passed-on options.
5122 * opts.c (common_handle_option): Handle the new option.
5123 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5125 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
5127 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5128 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
5129 third argument of type "const signed char" to descriptions of
5130 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5131 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5132 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5133 __builtin_bcdsub_ov functions.
5135 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5137 PR tree-optimization/86858
5138 * tree-vect-loop.c (vect_is_simple_reduction): Restore
5139 flow_bb_inside_loop_p calls.
5141 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5143 PR tree-optimization/86871
5144 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5145 instead of gimple_assign_lhs.
5147 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
5150 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5151 register constraint to operand 0.
5152 (add<mode>3_carryinC): Likewise.
5153 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5155 2018-08-09 Martin Liska <mliska@suse.cz>
5158 * common.opt: Remove extra line.
5160 2018-08-09 Martin Liska <mliska@suse.cz>
5162 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5163 at the end of a line, make first letter capital and end up
5164 a sentence with a dot.
5165 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5166 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5167 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5168 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5169 (PARAM_MAX_ISL_OPERATIONS): Likewise.
5170 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5171 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5172 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5173 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5174 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5175 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5176 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5177 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5178 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5179 (PARAM_TREE_REASSOC_WIDTH): Likewise.
5180 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5181 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5182 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5184 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
5187 * config/s390/s390.c (s390_option_override_internal): Reduce the
5188 stack-clash-protection-probe-interval param if it would be too big
5191 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
5194 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5195 * config/m68k/m68k.c (handle_move_double): Don't call
5196 m68k_final_prescan_insn.
5197 (m68k_adjust_decorated_operand): Renamed from
5198 m68k_final_prescan_insn, remove first and third operand and
5200 (print_operand): Call it.
5201 (print_operand_address): Call it.
5203 2018-08-08 Nathan Sidwell <nathan@acm.org>
5205 * diagnostic.c (diagnostic_report_current_module): Use
5206 linemap_included_from & linemap_included_from_linemap.
5208 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
5210 * config/aarch64/aarch64-cores.def: Add phecda core.
5211 * config/aarch64/aarch64-tune.md: Regenerate.
5212 * doc/invoke.texi: Add phecda core.
5214 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5217 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5219 * config/s390/s390.md ("movti"): Add more alternatives for
5220 constant to GPR copies.
5222 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5224 * config/s390/s390.c: Fix whitespace damage throughout the file.
5225 * config/s390/s390.h: Likewise.
5226 * config/s390/tpf.h: Likewise.
5228 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5230 * config/s390/s390.c (s390_loadrelative_operand_p):
5231 Remove TARGET_CPU_ZARCH usages.
5232 (s390_rtx_costs): Likewise.
5233 (s390_legitimate_constant_p): Likewise.
5234 (s390_cannot_force_const_mem): Likewise.
5235 (legitimate_reload_constant_p): Likewise.
5236 (s390_preferred_reload_class): Likewise.
5237 (legitimize_pic_address): Likewise.
5238 (legitimize_tls_address): Likewise.
5239 (s390_split_branches): Removed.
5240 (s390_add_execute): Removed.
5241 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5242 (s390_mainpool_start): Likewise.
5243 (s390_mainpool_finish): Likewise.
5244 (s390_mainpool_cancel): Removed.
5245 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5246 (s390_chunkify_cancel): Likewise.
5247 (s390_return_addr_rtx): Likewise.
5248 (s390_register_info): Remove split_branches_pending_p uages.
5249 (s390_optimize_register_info): Likewise.
5250 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5251 split_branches_pending_p usages.
5252 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5253 (s390_load_got): Likewise.
5254 (s390_expand_split_stack_prologue): Likewise.
5255 (output_asm_nops): Likewise.
5256 (s390_function_profiler): Likewise.
5257 (s390_emit_call): Likewise.
5258 (s390_conditional_register_usage): Likewise.
5259 (s390_optimize_prologue): Likewise.
5260 (s390_reorg): Remove TARGET_CPU_ZARCH and
5261 split_branches_pending_p usages.
5262 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5264 (s390_output_indirect_thunk_function): Likewise.
5265 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5266 (TARGET_CPU_ZARCH_P): Removed.
5267 (struct machine_function): Remove split_branches_pending_p.
5268 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5270 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5272 * common/config/s390/s390-common.c (processor_flags_table):
5274 * config.gcc: Remove with_arch/with_tune support.
5275 * config/s390/2064.md: Remove cpu attribute comparisons.
5276 * config/s390/driver-native.c (s390_host_detect_local_cpu):
5278 * config/s390/linux.h (ASM_SPEC):
5279 Remove -march support.
5280 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5281 Use a table to get an arch level.
5282 * config/s390/s390-opts.h (enum processor_type):
5284 * config/s390/s390.c
5285 (processor_table): Remove entries, add arch_level values.
5286 (s390_issue_rate): Remove cases.
5287 (s390_option_override): Adjust
5288 s390_option_override_internal() call.
5289 (s390_option_override_internal): Remove deprecation warning.
5290 (s390_valid_target_attribute_tree): Adjust
5291 s390_option_override_internal() call.
5292 * config/s390/s390.h (struct s390_processor):
5293 Share with s390-c.c, add arch_level field.
5294 * config/s390/s390.md:
5295 Remove occurrences in cpu attribute.
5296 * config/s390/s390.opt: Remove -march/-mtune support.
5297 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5298 * doc/invoke.texi: Remove deprecation warning.
5300 2018-08-08 Luis Machado <luis.machado@linaro.org>
5302 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5304 (qdf24xx_tunings): Set vector cost structure to
5305 qdf24xx_vector_cost.
5307 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5308 <register_sextend>: Set to 3.
5310 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
5313 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5314 * config/aarch64/aarch64-simd.md
5315 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5316 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5317 (aarch64_frecpx<mode>): ...this new pattern.
5318 * config/aarch64/aarch64-simd-builtins.def: Remove comment
5319 about aarch64_frecp<FRECP:frecp_suffix><mode>.
5321 2018-08-07 Martin Liska <mliska@suse.cz>
5324 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5325 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5326 * predict.def (PRED_MALLOC_NONNULL): New predictor.
5327 * doc/extend.texi: Document that malloc attribute adds
5330 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
5333 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5334 Define to speculation_safe_value_not_needed.
5336 2018-08-06 Jeff Law <law@redhat.com>
5338 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5339 the vr_values instance to cprop_into_stmt.
5340 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5341 (cprop_operand): Also query EVRP to determine if OP is a constant.
5343 2018-08-06 Nathan Sidwell <nathan@acm.org>
5345 * diagnostic.c (diagnostic_report_current_module): Reroll
5346 included-at loop. Translate text.
5348 2018-08-06 David Malcolm <dmalcolm@redhat.com>
5350 * function-tests.c (selftest::test_expansion_to_rtl): Call
5351 free_after_compilation.
5353 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5355 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5357 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
5359 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5360 loops with memory block operations from getting unrolled.
5362 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
5365 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5366 Define to speculation_safe_value_not_needed.
5368 2018-08-06 Jeff Law <law@redhat.com>
5370 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5373 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5376 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5377 with all enabled __intN types.
5379 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5381 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5383 * alias.c (record_set): Check for clobber high.
5384 * cfgexpand.c (expand_gimple_stmt): Likewise.
5385 * combine-stack-adj.c (single_set_for_csa): Likewise.
5386 * combine.c (find_single_use_1): Likewise.
5387 (set_nonzero_bits_and_sign_copies): Likewise.
5388 (get_combine_src_dest): Likewise.
5389 (is_parallel_of_n_reg_sets): Likewise.
5390 (try_combine): Likewise.
5391 (record_dead_and_set_regs_1): Likewise.
5392 (reg_dead_at_p_1): Likewise.
5393 (reg_dead_at_p): Likewise.
5394 * dce.c (deletable_insn_p): Likewise.
5395 (mark_nonreg_stores_1): Likewise.
5396 (mark_nonreg_stores_2): Likewise.
5397 * df-scan.c (df_find_hard_reg_defs): Likewise.
5398 (df_uses_record): Likewise.
5399 (df_get_call_refs): Likewise.
5400 * dwarf2out.c (mem_loc_descriptor): Likewise.
5401 * haifa-sched.c (haifa_classify_rtx): Likewise.
5402 * ira-build.c (create_insn_allocnos): Likewise.
5403 * ira-costs.c (scan_one_insn): Likewise.
5404 * ira.c (equiv_init_movable_p): Likewise.
5405 (rtx_moveable_p): Likewise.
5406 (interesting_dest_for_shprep): Likewise.
5407 * jump.c (mark_jump_label_1): Likewise.
5408 * postreload-gcse.c (record_opr_changes): Likewise.
5409 * postreload.c (reload_cse_simplify): Likewise.
5410 (struct reg_use): Add source expr.
5411 (reload_combine): Check for clobber high.
5412 (reload_combine_note_use): Likewise.
5413 (reload_cse_move2add): Likewise.
5414 (move2add_note_store): Likewise.
5415 * print-rtl.c (print_pattern): Likewise.
5416 * recog.c (decode_asm_operands): Likewise.
5417 (store_data_bypass_p): Likewise.
5418 (if_test_bypass_p): Likewise.
5419 * regcprop.c (kill_clobbered_value): Likewise.
5420 (kill_set_value): Likewise.
5421 * reginfo.c (reg_scan_mark_refs): Likewise.
5422 * reload1.c (maybe_fix_stack_asms): Likewise.
5423 (eliminate_regs_1): Likewise.
5424 (elimination_effects): Likewise.
5425 (mark_not_eliminable): Likewise.
5426 (scan_paradoxical_subregs): Likewise.
5427 (forget_old_reloads_1): Likewise.
5428 * reorg.c (find_end_label): Likewise.
5429 (try_merge_delay_insns): Likewise.
5430 (redundant_insn): Likewise.
5431 (own_thread_p): Likewise.
5432 (fill_simple_delay_slots): Likewise.
5433 (fill_slots_from_thread): Likewise.
5434 (dbr_schedule): Likewise.
5435 * resource.c (update_live_status): Likewise.
5436 (mark_referenced_resources): Likewise.
5437 (mark_set_resources): Likewise.
5438 * rtl.c (copy_rtx): Likewise.
5439 * rtlanal.c (reg_referenced_p): Likewise.
5440 (single_set_2): Likewise.
5441 (noop_move_p): Likewise.
5442 (note_stores): Likewise.
5443 * sched-deps.c (sched_analyze_reg): Likewise.
5444 (sched_analyze_insn): Likewise.
5446 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5448 * cse.c (invalidate_reg): New function extracted from...
5449 (invalidate): ...here.
5450 (canonicalize_insn): Check for clobber high.
5451 (invalidate_from_clobbers): invalidate clobber highs.
5452 (invalidate_from_sets_and_clobbers): Likewise.
5453 (count_reg_usage): Check for clobber high.
5454 (insn_live_p): Likewise.
5455 * cselib.c (cselib_expand_value_rtx_1):Likewise.
5456 (cselib_invalidate_regno): Check for clobber in setter.
5457 (cselib_invalidate_rtx): Pass through setter.
5458 (cselib_invalidate_rtx_note_stores):
5459 (cselib_process_insn): Check for clobber high.
5460 * cselib.h (cselib_invalidate_rtx): Add operand.
5462 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5464 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5465 (mark_not_eliminable): Likewise.
5466 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5467 * lra-lives.c (process_bb_lives): Check for clobber high.
5468 * lra.c (new_insn_reg): Remember clobber highs.
5469 (collect_non_operand_hard_regs): Check for clobber high.
5470 (lra_set_insn_recog_data): Likewise.
5471 (add_regs_to_insn_regno_info): Likewise.
5472 (lra_update_insn_regno_info): Likewise.
5474 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5476 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5477 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5479 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5481 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5482 (copy_insn_1): Likewise.
5483 (gen_hard_reg_clobber_high): New gen function.
5484 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5485 * genemit.c (gen_exp): Likewise.
5486 (gen_emit_seq): Pass through info.
5487 (gen_insn): Check for CLOBBER_HIGH.
5488 (gen_expand): Pass through info.
5489 (gen_split): Likewise.
5490 (output_add_clobbers): Likewise.
5491 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5492 (remove_clobbers): Likewise.
5493 * rtl.h (gen_hard_reg_clobber_high): New declaration.
5495 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5497 * doc/rtl.texi (clobber_high): Add.
5498 (parallel): Add in clobber high
5499 * rtl.c (rtl_check_failed_code3): Add function.
5500 * rtl.def (CLOBBER_HIGH): Add expression.
5501 * rtl.h (RTL_CHECKC3): Add macro.
5502 (rtl_check_failed_code3): Add declaration.
5503 (XC3EXP): Add macro.
5505 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
5508 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5509 cfun->machine->max_used_stack_alignment if needed.
5511 2018-08-04 Martin Sebor <msebor@redhat.com>
5513 PR tree-optimization/86571
5514 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5517 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
5520 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5523 2018-08-03 Jeff Law <law@redhat.com>
5526 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5527 Define to speculation_safe_value_not_needed.
5529 2018-08-03 David Malcolm <dmalcolm@redhat.com>
5531 * doc/gcov.texi (-x): Remove duplicate "to".
5532 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5533 (-Wif-not-aligned): Remove duplicate "is".
5534 (-flto): Remove duplicate "the".
5535 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5536 duplicate "v5.00.b".
5537 (MSP430 Options): Remove duplicate "and" from the description
5538 of "-mgprel-sec=regexp".
5539 (x86 Options): Remove duplicate copies of "vmldLog102" and
5540 vmlsLog104 from description of "-mveclibabi=type".
5542 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
5544 * internal-fn.h (first_commutative_argument): Declare.
5545 * internal-fn.c (first_commutative_argument): New function.
5546 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5547 restrictions for pattern statements. Use first_commutative_argument
5548 to look for commutative operands in calls to internal functions.
5550 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
5552 * Makefile.in (wide-int-range.o): New.
5553 * tree-vrp.c: Move all the wide_int_* functions to...
5554 * wide-int-range.cc: ...here.
5555 * tree-vrp.h: Move all the wide_int_* prototypes to...
5556 * wide-int-range.h: ...here.
5558 2018-08-03 Tom de Vries <tdevries@suse.de>
5560 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5562 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5563 * except.c (output_function_exception_table): Do early exit if
5564 targetm_common.except_unwind_info (&global_options) == UI_NONE.
5566 2018-08-03 Martin Liska <mliska@suse.cz>
5568 * predict.c (dump_prediction): Change to 2 digits
5571 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
5573 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5575 (falkor_am_1_gtov_gtov): ... a new insn reservation.
5577 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
5579 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5580 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5581 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5582 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5583 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5585 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5587 * diagnostic-show-locus.c (diagnostic_show_locus): Use
5588 pp_take_prefix when saving the existing prefix.
5589 * diagnostic.c (diagnostic_append_note): Likewise.
5590 * langhooks.c (lhd_print_error_function): Likewise.
5591 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5592 param's type. Free the existing prefix.
5593 (pp_take_prefix): New function.
5594 (pretty_printer::pretty_printer): Drop the prefix parameter.
5595 Rename the length parameter to match the comment.
5596 (pretty_printer::~pretty_printer): Free the prefix.
5597 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5599 (struct pretty_printer): Drop the "const" from "prefix" field's
5600 type and clarify memory management.
5601 (pp_set_prefix): Drop the "const" from the 2nd param.
5602 (pp_take_prefix): New decl.
5604 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5606 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5607 for word_mode_ok here instead of passing as argument.
5608 (expand_block_compare): Change select_block_compare_mode() call.
5609 (expand_strncmp_gpr_sequence): New function.
5610 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5612 2018-08-02 Jeff Law <law@redhat.com>
5615 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5616 Define to speculation_safe_value_not_needed.
5619 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5620 Define to speculation_safe_value_not_needed.
5622 2018-08-02 Tom de Vries <tdevries@suse.de>
5625 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5626 function. Return UI_TARGET unconditionally.
5627 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5628 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5630 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5632 * genemit.c (print_overload_test): Fix typo.
5634 2018-08-02 Richard Biener <rguenther@suse.de>
5636 PR tree-optimization/86816
5637 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5638 which checks for value availability before querying it.
5639 (gvn_uses_equal): Use it.
5640 (same_succ_hash): Likewise.
5641 (gimple_equal_p): Likewise.
5643 2018-08-02 Nick Clifton <nickc@redhat.com>
5646 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5647 Define to speculation_safe_value_not_needed.
5650 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5651 Define to speculation_safe_value_not_needed.
5654 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5655 Define to speculation_safe_value_not_needed.
5658 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5659 Define to speculation_safe_value_not_needed.
5662 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5663 Define to speculation_safe_value_not_needed.
5666 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5667 Define to speculation_safe_value_not_needed.
5670 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5671 Define to speculation_safe_value_not_needed.
5674 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5675 Define to speculation_safe_value_not_needed.
5678 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5679 speculation_safe_value_not_needed.
5682 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5683 to speculation_safe_value_not_needed.
5685 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5687 * doc/md.texi: Expand the documentation of instruction names
5688 to mention port-local uses. Document '@' in pattern names.
5689 * read-md.h (overloaded_instance, overloaded_name): New structs.
5691 (md_reader::handle_overloaded_name): New member function.
5692 (md_reader::get_overloads): Likewise.
5693 (md_reader::m_first_overload): New member variable.
5694 (md_reader::m_next_overload_ptr): Likewise.
5695 (md_reader::m_overloads_htab): Likewise.
5696 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5697 m_next_overload_ptr and m_overloads_htab.
5698 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5699 (get_mode_token, get_code_token, get_int_token): New functions.
5700 (map_attr_string): Add an optional argument that passes back
5701 the associated iterator.
5702 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5703 (md_reader::handle_overloaded_name, add_overload_instance): New
5705 (apply_iterators): Handle '@' names. Report an error if '@'
5706 is used without iterators.
5707 (initialize_iterators): Initialize the new iterator_group fields.
5708 * genopinit.c (handle_overloaded_code_for)
5709 (handle_overloaded_gen): New functions.
5710 (main): Use them to print declarations of maybe_code_for_* and
5711 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5712 * genemit.c (print_overload_arguments, print_overload_test)
5713 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5714 (main): Use it to print definitions of maybe_code_for_* and
5715 maybe_gen_* functions.
5716 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5717 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5718 instead of explicit mode checks.
5719 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5720 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5721 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5722 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5723 (aarch64_expand_compare_and_swap): Likewise
5724 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5725 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5726 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5727 (aarch64_constant_pool_reload_icode): Delete.
5728 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5729 instead of aarch64_constant_pool_reload_icode. Use
5730 code_for_aarch64_reload_mov instead of explicit mode checks.
5731 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5732 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5733 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5734 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5735 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5736 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5737 (aarch64_atomic_load_op_code): Delete.
5738 (aarch64_emit_atomic_load_op): Likewise.
5739 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5740 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
5741 instead of aarch64_emit_atomic_load_op.
5742 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5743 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5744 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5745 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5746 character before the pattern name.
5747 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5748 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5749 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5750 (aarch64_frecps<mode>): Likewise.
5751 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5752 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5753 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5754 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5755 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5757 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5759 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5760 Allow HFmode constants if TARGET_FP_F16INST.
5762 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
5765 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5766 No longer check last store for clobber of address register.
5768 2018-08-02 Martin Liska <mliska@suse.cz>
5770 PR gcov-profile/86817
5771 * gcov.c (process_all_functions): New function.
5773 (process_file): Move functions processing to
5774 process_all_functions.
5776 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5778 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5779 "const" to the "gimple *" and "rtx_insn *" parameters.
5780 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5782 (dump_location_t::dump_location_t): Likewise.
5784 2018-08-01 Martin Sebor <msebor@redhat.com>
5786 PR tree-optimization/86650
5787 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5788 rather than a "gcall *". Directly pass the data of interest
5789 to percent_K_format, rather than building a temporary CALL_EXPR
5791 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5792 (gimple_fold_builtin_strncat): Adjust.
5793 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5794 gcall* argument with gimple*.
5795 * gimple-ssa-warn-restrict.c (check_call): Same.
5796 (wrestrict_dom_walker::before_dom_children): Same.
5797 (builtin_access::builtin_access): Same.
5798 (check_bounds_or_overlap): Same
5799 (maybe_diag_overlap): Same.
5800 (maybe_diag_offset_bounds): Same.
5801 * tree-diagnostic.c (default_tree_printer): Move usage of
5802 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5804 * tree-pretty-print.c (percent_K_format): Add argument.
5805 * tree-pretty-print.h: Add argument.
5806 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5807 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5808 (maybe_diag_stxncpy_trunc): Same.
5809 (handle_builtin_stxncpy): Same.
5810 (handle_builtin_strcat): Same.
5812 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5814 * match.pd: Optimise pointer range checks.
5816 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5818 PR tree-optimization/86758
5819 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5820 to remove pattern statements.
5822 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5824 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5825 result of dfs_enumerate_from when constructing stmt_vec_infos,
5826 instead of additionally calling get_loop_body.
5828 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5830 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5832 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5833 When creating an iv, assert that the step is not known to be zero.
5834 (vect_setup_realignment): Update call accordingly.
5835 * tree-vect-stmts.c (vectorizable_store): Likewise.
5836 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
5838 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5840 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5841 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5842 (vectorizable_reduction): Likewise.
5843 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5844 (vect_detect_hybrid_slp_stmts): Likewise.
5845 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5847 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
5849 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5850 (wide_int_set_zero_nonzero_bits): ...this.
5851 (zero_nonzero_bits_from_vr): Rename to...
5852 (vrp_set_zero_nonzero_bits): ...this.
5853 (extract_range_from_multiplicative_op_1): Abstract wide int
5855 (wide_int_range_multiplicative_op): ...here.
5856 (extract_range_from_binary_expr_1): Extract wide int binary
5857 operations into their own functions.
5858 (wide_int_range_lshift): New.
5859 (wide_int_range_can_optimize_bit_op): New.
5860 (wide_int_range_shift_undefined_p): New.
5861 (wide_int_range_bit_xor): New.
5862 (wide_int_range_bit_ior): New.
5863 (wide_int_range_bit_and): New.
5864 (wide_int_range_trunc_mod): New.
5865 (extract_range_into_wide_ints): New.
5866 (vrp_shift_undefined_p): New.
5867 (extract_range_from_multiplicative_op): New.
5868 (vrp_can_optimize_bit_op): New.
5869 * tree-vrp.h (value_range::dump): New.
5870 (wide_int_range_multiplicative_op): New.
5871 (wide_int_range_lshift):New.
5872 (wide_int_range_shift_undefined_p): New.
5873 (wide_int_range_bit_xor): New.
5874 (wide_int_range_bit_ior): New.
5875 (wide_int_range_bit_and): New.
5876 (wide_int_range_trunc_mod): New.
5877 (zero_nonzero_bits_from_bounds): Rename to...
5878 (wide_int_set_zero_nonzero_bits): ...this.
5879 (zero_nonzero_bits_from_vr): Rename to...
5880 (vrp_set_zero_nonzero_bits): ...this.
5881 (range_easy_mask_min_max): Rename to...
5882 (wide_int_range_can_optimize_bit_op): this.
5883 * vr-values.c (simplify_bit_ops_using_ranges): Rename
5884 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5886 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5888 * tree-vectorizer.h (vect_orig_stmt): New function.
5889 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5890 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5891 (vect_create_epilog_for_reduction): Likewise.
5892 (vectorizable_live_operation): Likewise.
5893 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5894 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5895 * tree-vect-stmts.c (vectorizable_call): Likewise.
5896 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5898 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5900 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5902 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5903 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5904 (vect_transform_loop): Likewise.
5905 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5907 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5909 * tree-vectorizer.h (vect_schedule_slp): Return void.
5910 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5911 (vect_schedule_slp): Likewise.
5913 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5915 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5917 (vect_transform_loop): Update calls accordingly. Schedule SLP
5918 instances before the main loop, if any exist.
5920 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5922 PR tree-optimization/86749
5923 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5924 If the lhs is used in a COND_EXPR, check that it is being used
5925 as the "then" or "else" value.
5927 2018-08-01 Tom de Vries <tdevries@suse.de>
5930 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5931 speculation_safe_value_not_needed.
5933 2018-08-01 Richard Biener <rguenther@suse.de>
5935 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5938 2018-08-01 Martin Liska <mliska@suse.cz>
5940 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5941 format how successful transformation is dumped.
5942 (gimple_mod_pow2_value_transform): Likewise.
5943 (gimple_mod_subtract_transform): Likewise.
5944 (gimple_stringops_transform): Likewise.
5946 2018-08-01 Martin Liska <mliska@suse.cz>
5949 * value-prof.c (interesting_stringop_to_profile_p):
5950 Simplify the code and add BUILT_IN_MEMMOVE.
5951 (gimple_stringops_transform): Likewise.
5953 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
5955 * config/aarch64/aarch64-simd.md
5956 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5957 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5958 use GPI iterator instead of SI mode.
5960 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5962 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5963 rs6000_speculation_barrier.
5964 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5965 new barrier pattern name.
5967 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5969 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5970 (speculation_barrier): New insn.
5972 2018-08-01 Richard Biener <rguenther@suse.de>
5975 * graphite.h: Include isl/id.h and isl/space.h to allow build
5978 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
5981 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5982 mode for COFF targets.
5983 * defaults.h (TARGET_COFF): Define.
5984 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5985 TARGET_COFF): Define.
5986 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5987 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5988 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5990 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5992 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5993 * gimple-streamer-out.c (output_bb): Save it.
5994 * lto-streamer-in.c (input_struct_function_base): Restore
5995 instance discriminator if available. Create map on demand.
5996 * lto-streamer-out.c (output_struct_function_base): Save it if
5998 * final.c (decl_to_instance_map): Document LTO strategy.
6000 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6001 Olivier Hainque <hainque@adacore.com>
6003 * debug.h (decl_to_instance_map_t): New type.
6004 (decl_to_instance_map): Declare.
6005 (maybe_create_decl_to_instance_map): New inline function.
6006 * final.c (bb_discriminator, last_bb_discriminator): New statics,
6007 to track basic block discriminators.
6008 (final_start_function_1): Initialize them.
6009 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6011 (decl_to_instance_map): New variable.
6012 (map_decl_to_instance, maybe_set_discriminator): New functions.
6013 (notice_source_line): Set discriminator.
6015 2018-07-31 Ian Lance Taylor <iant@golang.org>
6017 * targhooks.c (default_have_speculation_safe_value): Add
6020 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6022 * dump-context.h: Include "pretty-print.h".
6023 (dump_context::refresh_dumps_are_enabled): New decl.
6024 (dump_context::emit_item): New decl.
6025 (class dump_context): Add fields "m_test_pp" and
6027 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6028 (temp_dump_context::get_dumped_text): New decl.
6029 (class temp_dump_context): Add field "m_pp".
6030 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6031 (dump_context::refresh_dumps_are_enabled): ...and add a test for
6033 (set_dump_file): Update for above change.
6034 (set_alt_dump_file): Likewise.
6035 (dump_loc): New overload, taking a pretty_printer *.
6036 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
6037 to any test pretty-printer.
6038 (make_item_for_dump_gimple_stmt): New function, adapted from
6039 optinfo::add_gimple_stmt.
6040 (dump_context::dump_gimple_stmt): Call it, and use the result,
6041 eliminating the direct usage of dump_file and alt_dump_file in
6042 favor of indirectly using them via emit_item.
6043 (make_item_for_dump_gimple_expr): New function, adapted from
6044 optinfo::add_gimple_expr.
6045 (dump_context::dump_gimple_expr): Call it, and use the result,
6046 eliminating the direct usage of dump_file and alt_dump_file in
6047 favor of indirectly using them via emit_item.
6048 (make_item_for_dump_generic_expr): New function, adapted from
6050 (dump_context::dump_generic_expr): Call it, and use the result,
6051 eliminating the direct usage of dump_file and alt_dump_file in
6052 favor of indirectly using them via emit_item.
6053 (make_item_for_dump_printf_va): New function, adapted from
6054 optinfo::add_printf_va.
6055 (make_item_for_dump_printf): New function.
6056 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6057 and use the result, eliminating the direct usage of dump_file and
6058 alt_dump_file in favor of indirectly using them via emit_item.
6059 (make_item_for_dump_dec): New function.
6060 (dump_context::dump_dec): Call it, and use the result,
6061 eliminating the direct usage of dump_file and alt_dump_file in
6062 favor of indirectly using them via emit_item.
6063 (make_item_for_dump_symtab_node): New function, adapted from
6064 optinfo::add_symtab_node.
6065 (dump_context::dump_symtab_node): Call it, and use the result,
6066 eliminating the direct usage of dump_file and alt_dump_file in
6067 favor of indirectly using them via emit_item.
6068 (dump_context::begin_scope): Reimplement, avoiding direct usage
6069 of dump_file and alt_dump_file in favor of indirectly using them
6071 (dump_context::emit_item): New member function.
6072 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6073 Set up test pretty-printer on the underlying context. Call
6074 refresh_dumps_are_enabled.
6075 (temp_dump_context::~temp_dump_context): Call
6076 refresh_dumps_are_enabled.
6077 (temp_dump_context::get_dumped_text): New member function.
6078 (selftest::verify_dumped_text): New function.
6079 (ASSERT_DUMPED_TEXT_EQ): New macro.
6080 (selftest::test_capture_of_dump_calls): Run all tests twice, with
6081 and then without optinfo enabled. Add uses of
6082 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
6083 * dumpfile.h: Update comment for the dump_* API.
6084 * optinfo-emit-json.cc
6085 (selftest::test_building_json_from_dump_calls): Update for new
6086 param for temp_dump_context ctor.
6087 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6088 and "m_owned" field.
6089 (optinfo_item::~optinfo_item): Likewise.
6090 (optinfo::add_item): New member function.
6091 (optinfo::emit): Update comment.
6092 (optinfo::add_string): Delete.
6093 (optinfo::add_printf): Delete.
6094 (optinfo::add_printf_va): Delete.
6095 (optinfo::add_gimple_stmt): Delete.
6096 (optinfo::add_gimple_expr): Delete.
6097 (optinfo::add_tree): Delete.
6098 (optinfo::add_symtab_node): Delete.
6099 (optinfo::add_dec): Delete.
6100 * optinfo.h (class dump_context): New forward decl.
6101 (optinfo::add_item): New decl.
6102 (optinfo::add_string): Delete.
6103 (optinfo::add_printf): Delete.
6104 (optinfo::add_printf_va): Delete.
6105 (optinfo::add_gimple_stmt): Delete.
6106 (optinfo::add_gimple_expr): Delete.
6107 (optinfo::add_tree): Delete.
6108 (optinfo::add_symtab_node): Delete.
6109 (optinfo::add_dec): Delete.
6110 (optinfo::add_poly_int): Delete.
6111 (optinfo_item::optinfo_item): Remove "owned" param.
6112 (class optinfo_item): Remove field "m_owned".
6114 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6117 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6118 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6119 requested variable alignment.
6120 (expand_one_ssa_partition): Likewise.
6121 (expand_one_var): Likewise.
6123 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6125 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6126 to speculation_safe_value_not_needed.
6128 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6130 * targhooks.h (speculation_safe_value_not_needed): New prototype.
6131 * targhooks.c (speculation_safe_value_not_needed): New function.
6132 * target.def (have_speculation_safe_value): Update documentation.
6133 * doc/tm.texi: Regenerated.
6135 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6137 * config/aarch64/iterators.md (ALLI_TI): New iterator.
6138 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6140 (despeculate_copy<ALLI:mode>_insn): New insn.
6141 (despeculate_copyti_insn): New insn.
6142 (despeculate_simple<ALLI:mode>): New insn
6143 (despeculate_simpleti): New insn.
6144 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6146 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6147 aarch64_speculation_safe_value.
6148 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6150 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6152 * config/aarch64/aarch64-speculation.cc: New file.
6153 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6154 before pass_reorder_blocks.
6155 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6157 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6158 X14 and X15 when tracking speculation.
6159 * config/aarch64/aarch64.md (register name constants): Add
6160 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6161 (unspec): Add UNSPEC_SPECULATION_TRACKER.
6162 (speculation_barrier): New insn attribute.
6163 (cmp<mode>): Allow SP in comparisons.
6164 (speculation_tracker): New insn.
6165 (speculation_barrier): Add speculation_barrier attribute.
6166 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6167 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6168 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6170 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6172 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6173 aarch64_track_speculation is true.
6174 (tb<optab><mode>1): Likewise.
6175 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6176 generate CB[N]Z when tracking speculation.
6177 (aarch64_split_compare_and_swap): Likewise.
6178 (aarch64_split_atomic_op): Likewise.
6180 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6182 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6184 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6186 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6187 (speculation_barrier): New insn.
6189 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6191 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6192 * config/arm/arm.md (speculation_barrier): New expand.
6193 (speculation_barrier_insn): New pattern.
6195 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6197 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6198 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6199 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6200 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6202 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6203 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6204 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6205 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6206 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6207 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6208 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6209 * builtins.c (expand_speculation_safe_value): New function.
6210 (expand_builtin): Call it.
6211 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6212 * doc/extend.texi: Document __builtin_speculation_safe_value.
6213 * doc/md.texi: Document "speculation_barrier" pattern.
6214 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6215 TARGET_HAVE_SPECULATION_SAFE_VALUE.
6216 * doc/tm.texi: Regenerated.
6217 * target.def (have_speculation_safe_value, speculation_safe_value): New
6219 * targhooks.c (default_have_speculation_safe_value): New function.
6220 (default_speculation_safe_value): New function.
6221 * targhooks.h (default_have_speculation_safe_value): Add prototype.
6222 (default_speculation_safe_value): Add prototype.
6224 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6226 * dump-context.h (dump_context::dump_loc): New decl.
6227 * dumpfile.c (dump_context::dump_loc): New member function.
6228 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6229 and dump_gimple_stmt.
6230 (dump_context::dump_gimple_expr_loc): Likewise, using
6232 (dump_context::dump_generic_expr_loc): Likewise, using
6234 (dump_context::dump_printf_loc_va): Likewise, using
6236 (dump_context::begin_scope): Explicitly using the global function
6237 "dump_loc", rather than the member function.
6239 2018-07-31 Martin Sebor <msebor@redhat.com>
6241 PR tree-optimization/86741
6242 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6244 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
6246 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6247 depend on whether prefetch instructions will be emitted or not.
6248 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6249 will be emitted or not.
6250 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6251 (TARGET_SETMEM_PFD): New macros.
6253 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6255 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6256 (NULL_STMT_VEC_INFO): Delete.
6257 (stmt_vec_info::operator*): Likewise.
6258 (stmt_vec_info::operator gimple *): Likewise.
6259 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6260 of NULL_STMT_VEC_INFO.
6261 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6262 (vect_reassociating_reduction_p): Likewise.
6263 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6264 (vectorizable_store): Likewise.
6265 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6266 (vec_info::free_stmt_vec_infos): Likewise.
6268 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6270 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6271 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6273 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6275 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6276 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6277 (vec_info::free_stmt_vec_info): New private member functions.
6278 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6279 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6280 * tree-parloops.c (gather_scalar_reductions): Remove calls to
6281 set_stmt_vec_info_vec and free_stmt_vec_infos.
6282 * tree-vect-loop.c (_loop_vec_info): Remove call to
6283 set_stmt_vec_info_vec.
6284 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6285 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6286 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6287 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6288 (vec_info::free_stmt_vec_info): ...these new functions. Remove
6289 assignments in {vec_info::,}new_stmt_vec_info that are redundant
6290 with the clearing in the xcalloc.
6291 (stmt_vec_info_vec): Delete.
6292 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6293 (vectorize_loops): Likewise.
6294 (vec_info::~vec_info): Remove argument from call to
6295 free_stmt_vec_infos.
6296 (vec_info::add_stmt): Remove vinfo argument from call to
6299 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6301 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6302 rather than a gimple stmt.
6303 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
6304 information for pattern statements when passed the original
6305 statement; instead wait to be passed the pattern statement itself.
6306 Don't call set_vinfo_for_stmt here.
6307 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6308 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6309 stmt_vec_infos here.
6310 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6311 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6312 stmt_vec_infos entry.
6314 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6316 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6317 * tree-vectorizer.c (vec_info::replace_stmt): New function.
6318 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6319 * tree-vect-stmts.c (vectorizable_call): Likewise.
6320 (vectorizable_simd_clone_call): Likewise.
6322 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6324 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6325 * tree-vectorizer.c (vec_info::remove_stmt): New function.
6326 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6327 * tree-vect-loop.c (vect_transform_loop): Likewise.
6328 * tree-vect-slp.c (vect_schedule_slp): Likewise.
6329 * tree-vect-stmts.c (vect_remove_stores): Likewise.
6331 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6333 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6334 (vect_dr_stmt): Delete.
6335 * tree-vectorizer.c (vec_info::lookup_dr): New function.
6336 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6338 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6339 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6340 (vect_verify_datarefs_alignment, vect_peeling_supportable)
6341 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6342 (vect_analyze_data_refs): Likewise.
6343 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
6345 (vect_find_same_alignment_drs): Likewise.
6346 (vect_slp_analyze_node_dependences): Update calls accordingly.
6347 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6348 instead of DR_VECT_AUX.
6349 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6350 of a vector data references. Use vec_info::lookup_dr instead of
6352 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6353 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6354 instead of DR_VECT_AUX.
6356 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6358 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6360 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6362 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6363 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6364 (vect_gen_prolog_loop_niters): Likewise.
6366 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6368 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6369 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6370 (vect_known_alignment_in_bytes, vect_dr_behavior)
6371 (vect_get_scalar_dr_size): Take references as dr_vec_infos
6372 instead of data_references. Update calls to other routines for
6373 which the same change has been made.
6374 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6375 dr_vec_infos instead of stmt_vec_infos.
6376 (vect_analyze_data_ref_dependence): Update call accordingly.
6377 (vect_slp_analyze_data_ref_dependence)
6378 (vect_record_base_alignments): Use DR_VECT_AUX.
6379 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6380 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6381 (vector_alignment_reachable_p, vect_get_data_access_cost)
6382 (vect_peeling_supportable, vect_analyze_group_access_1)
6383 (vect_analyze_group_access, vect_analyze_data_ref_access)
6384 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6385 (vect_compile_time_alias, vect_small_gap_p)
6386 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6387 (vect_supportable_dr_alignment): Take references as dr_vec_infos
6388 instead of data_references. Update calls to other routines for
6389 which the same change has been made.
6390 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6391 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6392 (vect_slp_analyze_and_verify_node_alignment)
6393 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6394 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6395 (vect_setup_realignment): Use dr_vec_infos. Update calls after
6397 (_vect_peel_info::dr): Replace with...
6398 (_vect_peel_info::dr_info): ...this new field.
6399 (vect_peeling_hash_get_most_frequent)
6400 (vect_peeling_hash_choose_best_peeling): Update accordingly.
6401 (vect_peeling_hash_get_lowest_cost):
6402 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
6403 routines for which the same change has been made.
6404 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
6406 * tree-vect-loop-manip.c (get_misalign_in_elems)
6407 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
6409 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6410 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6411 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6412 (get_group_load_store_type, get_negative_load_store_type)
6413 (vect_get_data_ptr_increment, vectorizable_store)
6414 (vectorizable_load): Likewise.
6415 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6416 Update calls to other routines for which the same change has been made.
6418 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6420 * tree-vectorizer.h (vec_info::move_dr): New member function.
6421 (dataref_aux): Rename to...
6422 (dr_vec_info): ...this and add "dr" and "stmt" fields.
6423 (_stmt_vec_info::dr_aux): Update accordingly.
6424 (_stmt_vec_info::data_ref_info): Delete.
6425 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6426 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6427 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6428 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6430 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
6431 (STMT_VINFO_DR_INFO): New macro.
6432 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6433 (set_dr_misalignment): Update after rename of dataref_aux.
6434 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
6435 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6436 initialization of STMT_VINFO_DATA_REF.
6437 * tree-vectorizer.c (vec_info::move_dr): New function.
6438 * tree-vect-patterns.c (vect_recog_bool_pattern)
6439 (vect_recog_mask_conversion_pattern)
6440 (vect_recog_gather_scatter_pattern): Use it.
6441 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6442 the "dr" and "stmt" fields of dr_vec_info instead of
6443 STMT_VINFO_DATA_REF.
6445 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6447 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6448 (is_pattern_stmt_p): Use it.
6449 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6450 on pattern statements.
6452 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6454 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6455 original stmt as a stmt_vec_info rather than a gimple stmt.
6456 (vect_pattern_recog_1): Take the statement directly as a
6457 stmt_vec_info, rather than via a gimple_stmt_iterator.
6458 Update call to vect_mark_pattern_stmts.
6459 (vect_pattern_recog): Update calls accordingly.
6461 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6463 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6464 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6465 a vect_def_type for the first argument.
6466 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6467 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
6468 operand if it isn't defined by a vectorized statement.
6469 (vect_build_gather_load_calls): Remove the mask_dt argument and
6470 update calls to vect_get_vec_def_for_stmt_copy.
6471 (vectorizable_bswap): Likewise the dt argument.
6472 (vectorizable_call): Update calls to vectorizable_bswap and
6473 vect_get_vec_def_for_stmt_copy.
6474 (vectorizable_simd_clone_call, vectorizable_assignment)
6475 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6476 (vectorizable_comparison): Update calls to
6477 vect_get_vec_def_for_stmt_copy.
6478 (vectorizable_store): Likewise. Remove now-unnecessary calls to
6480 (vect_get_loop_based_defs): Remove dt argument and update call
6481 to vect_get_vec_def_for_stmt_copy.
6482 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6483 and vect_get_vec_def_for_stmt_copy.
6484 (vectorizable_load): Update calls to vect_build_gather_load_calls
6485 and vect_get_vec_def_for_stmt_copy.
6486 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6487 (vectorizable_reduction, vectorizable_live_operation): Update calls
6488 to vect_get_vec_def_for_stmt_copy.
6490 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6492 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6493 and gimple stmt arguments with a stmt_vec_info.
6494 (vect_record_base_alignments): Update calls accordingly.
6495 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6496 and gimple stmt arguments with a stmt_vec_info.
6497 (vect_build_slp_tree_1): Remove vinfo argument and update call
6498 to vect_record_max_nunits.
6499 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6500 and vect_record_max_nunits.
6502 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6504 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6505 file and take a stmt_vec_info instead of a gimple stmt.
6506 (supportable_widening_operation, vect_finish_replace_stmt)
6507 (vect_finish_stmt_generation, vect_get_store_rhs)
6508 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6509 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6510 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6511 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6512 (vect_create_data_ref_ptr, bump_vector_ptr)
6513 (vect_permute_store_chain, vect_setup_realignment)
6514 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6515 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6516 (vectorizable_reduction, vectorizable_induction)
6517 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6518 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6519 than gimple stmts as arguments.
6520 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6521 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6522 (can_group_stmts_p, vect_check_gather_scatter)
6523 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6524 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6525 (vect_permute_load_chain, vect_shift_permute_load_chain)
6526 (vect_transform_grouped_load)
6527 (vect_record_grouped_load_vectors): Likewise.
6528 * tree-vect-loop.c (vect_fixup_reduc_chain)
6529 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6530 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6531 (vectorizable_reduction, vectorizable_induction)
6532 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6533 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6534 (vect_get_load_store_mask): Likewise.
6535 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6536 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6537 * tree-vect-stmts.c (vect_mark_relevant)
6538 (is_simple_and_all_uses_invariant)
6539 (exist_non_indexing_operands_for_use_p, process_use)
6540 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6541 (vect_get_vec_def_for_operand, vect_get_vec_defs)
6542 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6543 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6544 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6545 (get_negative_load_store_type, get_load_store_type)
6546 (vect_check_load_store_mask, vect_check_store_rhs)
6547 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6548 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6549 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6550 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6551 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6552 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6553 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6554 (supportable_widening_operation): Likewise.
6556 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6558 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6559 a stmt_vec_info instead of a gcall.
6560 (vect_check_gather_scatter): Update call accordingly.
6561 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6563 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6564 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6565 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6566 instead of a gimple stmt.
6567 (vect_transform_loop): Update calls accordingly.
6568 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6569 stmt_vec_infos instead of gimple stmts.
6570 (vect_analyze_slp_instance): Update use accordingly.
6571 * tree-vect-stmts.c (read_vector_array, write_vector_array)
6572 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6573 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6574 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6575 (vect_gen_widened_results_half, vect_get_loop_based_defs)
6576 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6577 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6578 down to subroutines.
6580 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6582 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6583 of the worklist from a vector of gimple stmts to a vector of
6585 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6586 (vect_mark_stmts_to_be_vectorized): Likewise
6588 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6590 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6591 statement before passing it to vect_analyze_stmt.
6592 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6593 the chain of phi vector definitions. Track the exit phi via its
6595 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6596 STMT_VINFO_REDUC_DEF.
6597 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6598 stmt_vec_infos to handle the statement chains.
6599 (vect_get_slp_defs): Record the first statement in the node
6600 using a stmt_vec_info.
6601 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6602 statements here and pass their stmt_vec_info down to subroutines.
6603 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6604 down to vect_finish_stmt_generation.
6605 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6606 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6607 stmt_vec_infos to subroutines.
6608 (vect_remove_stores): Use stmt_vec_infos to handle the statement
6611 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6613 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6614 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6615 (vect_permute_store_chain, vect_setup_realignment)
6616 (vect_permute_load_chain, vect_shift_permute_load_chain)
6617 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6618 stmts internally, and when passing values to other vectorizer routines.
6619 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6620 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6621 (vect_analyze_loop_operations, get_initial_def_for_reduction)
6622 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6623 (vectorizable_reduction, vectorizable_induction)
6624 (vectorizable_live_operation, vect_transform_loop_stmt)
6625 (vect_transform_loop): Likewise.
6626 * tree-vect-patterns.c (vect_reassociating_reduction_p)
6627 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6628 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6629 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6630 (vect_slp_analyze_node_operations_1): Likewise.
6631 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6632 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6633 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6634 (vect_finish_stmt_generation_1, get_group_load_store_type)
6635 (get_load_store_type, vect_build_gather_load_calls)
6636 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6637 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6638 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6639 (vectorizable_store, vectorizable_load, vectorizable_condition)
6640 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6641 (supportable_widening_operation): Likewise.
6642 (vect_get_vector_types_for_stmt): Likewise.
6643 * tree-vectorizer.h (vect_dr_behavior): Likewise.
6645 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6647 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6648 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6649 (vect_permute_store_chain, vect_permute_load_chain)
6650 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6651 repeated stmt_vec_info lookups.
6652 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6653 (vect_update_ivs_after_vectorizer): Likewise.
6654 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6655 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6656 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6657 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6658 (vect_bb_slp_scalar_cost): Likewise.
6659 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6661 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6663 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6664 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6665 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6666 to get gassigns and gcalls, rather than operating on generc gimple
6668 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6669 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6670 (vectorizable_load, vect_analyze_stmt): Likewise.
6671 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6673 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6675 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6676 return stmt_vec_infos rather than gimple stmts. Do not accept
6678 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6680 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6681 Update use of get_later_stmt.
6682 (vect_get_constant_vectors): Update call accordingly.
6683 (vect_schedule_slp_instance): Likewise
6684 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6685 (vect_slp_analyze_instance_dependence): Likewise.
6686 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6688 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6690 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6691 (stmt_info_for_cost::stmt_info): ...this new field.
6692 (add_stmt_costs): Update accordingly.
6693 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6694 (vect_get_known_peeling_cost): Likewise.
6695 (vect_estimate_min_profitable_iters): Likewise.
6696 * tree-vect-stmts.c (record_stmt_cost): Likewise.
6698 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6700 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6701 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6702 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6704 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6706 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6708 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6709 a gimple stmt to a stmt_vec_info.
6710 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6712 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6714 * tree-vectorizer.h (vec_info::grouped_stores): Change from
6715 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6716 (_loop_vec_info::reduction_chains): Likewise.
6717 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6719 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6721 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6723 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6724 a gimple stmt to a stmt_vec_info.
6725 (_stmt_vec_info::next_element): Likewise.
6726 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6727 (vect_slp_analyze_and_verify_node_alignment)
6728 (vect_analyze_group_access_1, vect_analyze_group_access)
6729 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6730 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6731 (vect_supportable_dr_alignment): Update accordingly.
6732 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6733 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6734 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6735 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6736 * tree-vect-slp.c (vect_build_slp_tree_1)
6737 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6738 (vect_split_slp_store_group, vect_analyze_slp_instance)
6739 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6740 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6741 (get_group_load_store_type, get_load_store_type)
6742 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6743 (vect_transform_stmt, vect_remove_stores): Likewise.
6745 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6747 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6749 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6750 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6751 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6752 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6753 (vector_alignment_reachable_p, vect_get_data_access_cost)
6754 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6755 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6756 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6757 (vect_analyze_group_access_1, vect_analyze_group_access)
6758 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6759 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6760 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6761 result of vect_dr_stmt and use the stmt_vec_info instead of
6762 the associated gimple stmt.
6763 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6764 (vect_gen_prolog_loop_niters): Likewise.
6765 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6767 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6769 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6770 to a vec<stmt_vec_info>.
6771 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6772 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6774 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6775 to a vec<stmt_vec_info>.
6776 (bst_traits::value_type, bst_traits::value_type): Likewise.
6777 (bst_traits::hash): Update accordingly.
6778 (vect_get_and_check_slp_defs): Change the stmts parameter from
6779 a vec<gimple *> to a vec<stmt_vec_info>.
6780 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6781 (vect_build_slp_tree): Likewise.
6782 (vect_build_slp_tree_2): Likewise. Update uses of
6783 SLP_TREE_SCALAR_STMTS.
6784 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6785 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6786 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6787 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6788 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6789 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6790 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6791 (vect_get_constant_vectors, vect_get_slp_defs)
6792 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6793 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6794 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6795 instead of gimple stmts.
6796 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6797 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6798 (vect_slp_analyze_instance_dependence): Update uses of
6799 SLP_TREE_SCALAR_STMTS.
6800 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6801 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6802 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6803 (get_initial_defs_for_reduction): Likewise.
6804 (vect_create_epilog_for_reduction): Likewise.
6805 (vectorize_fold_left_reduction): Likewise.
6806 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6807 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6808 (can_vectorize_live_stmts): Likewise.
6810 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6812 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6813 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6814 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6816 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6818 (gather_scalar_reductions): Update after above interface changes.
6819 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6820 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6822 (vect_force_simple_reduction): Likewise.
6823 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6824 LOOP_VINFO_REDUCTIONS.
6825 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6827 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6829 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6830 a gimple stmt to a stmt_vec_info.
6831 * tree-vect-loop.c (vect_active_double_reduction_p)
6832 (vect_force_simple_reduction, vectorizable_reduction): Update
6835 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6837 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6838 vec<gimple *> to a vec<stmt_vec_info>.
6839 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6840 the reduction_phis argument from a vec<gimple *> to a
6842 (vectorizable_reduction): Likewise the phis local variable that
6843 is passed to vect_create_epilog_for_reduction. Update for new type
6844 of SLP_TREE_VEC_STMTS.
6845 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6846 (vectorizable_live_operation): Likewise.
6847 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6848 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6850 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6852 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6853 a gimple stmt to a stmt_vec_info.
6854 (vectorizable_condition, vectorizable_live_operation)
6855 (vectorizable_reduction, vectorizable_induction): Pass back the
6856 vectorized statement as a stmt_vec_info.
6857 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6858 use of STMT_VINFO_VEC_STMT.
6859 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6860 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6861 as stmt_vec_infos rather than gimple stmts.
6862 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6864 (vectorizable_live_operation): Likewise.
6865 (vectorizable_reduction, vectorizable_induction): Likewise,
6866 updating use of STMT_VINFO_VEC_STMT.
6867 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6868 of STMT_VINFO_VEC_STMT.
6869 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6870 (vectorizable_simd_clone_call, vectorizable_conversion)
6871 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6872 (vectorizable_store, vectorizable_load, vectorizable_condition)
6873 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6874 from a gimple stmt to a stmt_vec_info.
6875 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
6876 pointer to a stmt_vec_info to the vectorizable_* routines.
6878 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6880 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6881 a gimple stmt to a stmt_vec_info.
6882 (is_pattern_stmt_p): Update accordingly.
6883 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6884 (vect_record_grouped_load_vectors): Likewise.
6885 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6886 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6887 (vect_model_reduction_cost): Likewise.
6888 (vect_create_epilog_for_reduction): Likewise.
6889 (vectorizable_reduction, vectorizable_induction): Likewise.
6890 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6891 Return the stmt_vec_info for the pattern statement.
6892 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6893 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6894 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6895 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6896 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6897 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6898 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6899 (free_stmt_vec_info, vect_is_simple_use): Likewise.
6901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6903 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6904 (vect_finish_stmt_generation): Likewise.
6905 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6906 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6907 (vect_build_gather_load_calls): Use the return value of the above
6908 functions instead of a separate call to vinfo_for_stmt. Use narrow
6909 scopes for the input gimple stmt and wider scopes for the associated
6910 stmt_vec_info. Use vec_info::lookup_def when setting these
6911 stmt_vec_infos from an SSA_NAME definition.
6912 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6913 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6914 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6915 (vectorizable_store, vectorizable_load, vectorizable_condition)
6916 (vectorizable_comparison): Likewise.
6917 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6918 (vectorizable_reduction): Likewise.
6920 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6922 * tree-vectorizer.h (vect_is_simple_use): Add an optional
6923 stmt_vec_info * parameter before the optional gimple **.
6924 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6925 (process_use, vect_get_vec_def_for_operand_1): Update callers.
6926 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6927 * tree-vect-loop.c (vectorizable_reduction): Likewise.
6928 (vectorizable_live_operation): Likewise.
6929 * tree-vect-patterns.c (type_conversion_p): Likewise.
6930 (vect_look_through_possible_promotion): Likewise.
6931 (vect_recog_rotate_pattern): Likewise.
6932 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6934 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6936 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6937 a typedef to a wrapper class.
6938 (NULL_STMT_VEC_INFO): New macro.
6939 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6940 (stmt_vec_info::operator*): New function.
6941 (stmt_vec_info::operator gimple *): Likewise.
6942 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6943 (add_stmt_costs): Likewise.
6944 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6945 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6946 (vect_get_known_peeling_cost): Likewise.
6947 (vect_estimate_min_profitable_iters): Likewise.
6948 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6949 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6950 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6951 (vectorizable_store, free_stmt_vec_infos): Likewise.
6952 (new_stmt_vec_info): Change return type of xcalloc to
6955 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6957 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6958 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6959 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6960 a single_imm_use-based sequence.
6961 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6963 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6965 * tree-vectorizer.h (vec_info::lookup_def): Declare.
6966 * tree-vectorizer.c (vec_info::lookup_def): New function.
6967 * tree-vect-patterns.c (vect_get_internal_def): Use it.
6968 (vect_widened_op_tree): Likewise.
6969 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6970 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6971 (vectorizable_reduction): Likewise.
6972 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6974 (vect_is_slp_reduction): Update calls accordingly. Use
6975 vec_info::lookup_def.
6976 (vect_is_simple_reduction): Likewise
6977 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6979 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6981 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6982 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6983 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6985 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6986 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6987 (vect_update_vf_for_slp, vect_analyze_loop_operations)
6988 (vect_is_slp_reduction, vectorizable_induction)
6989 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6990 * tree-vect-patterns.c (vect_init_pattern_stmt):
6991 (vect_determine_min_output_precision_1, vect_determine_precisions)
6992 (vect_pattern_recog): Likewise.
6993 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6994 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6995 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6996 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6997 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6998 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
6999 info field from a loop to a loop_vec_info.
7001 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7003 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7004 (vec_info::add_stmt): Declare.
7005 * tree-vectorizer.c (vec_info::add_stmt): New function.
7006 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7007 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7008 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7009 (vectorizable_induction): Likewise.
7010 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7011 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7012 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7013 (vectorizable_load): Likewise.
7014 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7015 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7016 (vect_recog_gather_scatter_pattern): Likewise.
7017 (append_pattern_def_seq): Likewise. Remove a check that is
7018 performed by add_stmt itself.
7020 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7022 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7023 which make_ssa_name was called with new_stmt before new_stmt
7026 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7028 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7030 (vect_is_slp_reduction): ...here...
7031 (vect_is_simple_reduction): ...and here. Remove repetition of tests
7032 that are already known to be false.
7034 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7036 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7037 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
7038 STMT_VINFO_NUM_SLP_USES when it's true.
7039 (vect_free_slp_instance): Add a final_p parameter and pass it to
7041 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7042 (vect_analyze_slp_instance): Likewise.
7043 (vect_slp_analyze_operations): Likewise.
7044 (vect_slp_analyze_bb_1): Likewise.
7045 * tree-vectorizer.c (vec_info): Likewise.
7046 * tree-vect-loop.c (vect_transform_loop): Likewise.
7048 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7050 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7051 function is not called for second and subsequent members of
7054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7056 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7057 cases for nested loops from here to ...
7058 (vect_create_epilog_for_reduction): ...here. Only call
7059 vect_is_simple_use for inner-loop reductions.
7061 2018-07-31 Martin Liska <mliska@suse.cz>
7063 PR gcov-profile/85338
7064 PR gcov-profile/85350
7065 PR gcov-profile/85372
7066 * profile.c (struct location_triplet): New.
7067 (struct location_triplet_hash): Likewise.
7068 (output_location): Do not output a BB that
7069 is already recorded for a line.
7070 (branch_prob): Use streamed_locations.
7072 2018-07-31 Martin Liska <mliska@suse.cz>
7074 PR gcov-profile/85370
7075 * coverage.c (coverage_begin_function): Do not mark target
7076 clones as artificial functions.
7078 2018-07-31 Martin Liska <mliska@suse.cz>
7080 PR gcov-profile/83813
7081 PR gcov-profile/84758
7082 PR gcov-profile/85217
7083 PR gcov-profile/85332
7084 * profile.c (branch_prob): Do not record GOTO expressions
7085 for GIMPLE statements which locations are already streamed.
7087 2018-07-31 Olivier Hainque <hainque@adacore.com>
7089 * gcc.c (handle_spec_function): Accept a soft_matched_part
7090 argument, as do_spec_1. Pass it down to ...
7091 (eval_spec_function): Accept a soft_matched_part argument,
7092 and pass it down to ...
7093 (do_spec_2): Accept a soft_matched_part argument, and pass
7094 it down to do_spec_1.
7095 (do_spec_1): Pass soft_matched_part to handle_spec_function.
7096 (handle_braces): Update call to handle_spec_function.
7097 (driver::set_up_specs): Update calls to do_spec_2.
7098 (compare_debug_dump_opt_spec_function): Likewise.
7099 (compare_debug_self_opt_spec_function): Likewise.
7101 2018-07-31 Olivier Hainque <hainque@adacore.com>
7103 * common.opt (nolibc): New option.
7104 * doc/invoke.texi (Link Options): Document it.
7105 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7106 * config/alpha/linux.h: Likewise.
7107 * config/arc/elf.h: Likewise.
7108 * config/arm/uclinux-elf.h: Likewise.
7109 * config/arm/unknown-elf.h: Likewise.
7110 * config/avr/avrlibc.h: Likewise.
7111 * config/bfin/bfin.h: Likewise.
7112 * config/bfin/linux.h: Likewise.
7113 * config/bfin/uclinux.h: Likewise.
7114 * config/darwin.h: Likewise.
7115 * config/darwin10.h: Likewise.
7116 * config/darwin12.h: Likewise.
7117 * config/gnu-user.h: Likewise.
7118 * config/lm32/uclinux-elf.h: Likewise.
7119 * config/pa/pa-hpux11.h: Likewise.
7120 * config/pa/pa64-hpux.h: Likewise.
7121 * config/sparc/sparc.h: Likewise.
7123 2018-07-31 Olivier Hainque <hainque@adacore.com>
7125 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7126 undefined variables.
7128 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7131 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7134 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7136 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7137 terminated string literal.
7139 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7141 PR rtl-optimization/85160
7142 * combine.c (is_just_move): New function.
7143 (try_combine): Allow combining two instructions into two if neither of
7144 the original instructions was a move.
7146 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
7149 * doc/extend.texi (Global Register Variables): Discourage use of type
7151 (Local Register Variables): Likewise.
7153 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
7155 PR tree-optimization/86506
7156 * hwint.h (ceil_log2): Resync with hwint.c implementation.
7158 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
7161 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7162 hard_regno, make sure no insn between `from` and `to` clobbers it.
7164 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
7165 Tom de Vries <tdevries@suse.de>
7167 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7168 (PTX_DEFAULT_RUNTIME_DIM): ... this.
7169 (nvptx_goacc_validate_dims): Set default worker and gang dims to
7170 PTX_DEFAULT_RUNTIME_DIM.
7171 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7173 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
7175 * config/pa/pa.c (pa_output_addr_vec): Align address table.
7176 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7177 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7179 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
7181 * config/rs6000/constraints.md (wG constraint): Delete, no longer
7183 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7184 predicate to reflect toc fusion has been deleted.
7185 (toc_fusion_mem_raw): Delete, no longer used.
7186 (toc_fusion_mem_wrapped): Likewise.
7187 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7189 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7190 Delete, no longer used.
7191 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7192 meant to be used for toc fusion.
7193 (rs6000_debug_print_mode): Delete toc fusion debugging.
7194 (rs6000_debug_reg_global): Likewise.
7195 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7196 fusion and secondary reload support that were never used.
7197 (rs6000_option_override_internal): Delete TOC fusion, that was only
7198 partially defined, and it did not work unless you also used the
7200 (rs6000_legitimate_address_p): Delete TOC fusion support.
7201 (rs6000_opt_masks): Likewise.
7202 (fusion_wrap_memory_address): Delete function, no longer used.
7203 (fusion_split_address); Delete TOC fusion support.
7204 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7205 longer used with toc fusion being deleted.
7206 (TARGET_TOC_FUSION_FP): Likewise.
7207 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7209 (toc fusion spliter): Delete TOC fusion support.
7210 (toc_fusionload_<mode>): Likewise.
7211 (toc_fusionload_di): Likewise.
7212 (fusion_gpr_load_<mode>): Delete generator function, this insn no
7213 longer needs to be named. Rename predicate to delete TOC fusion.
7214 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7215 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7216 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7217 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7218 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7220 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
7222 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7223 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
7224 __int128 in built-in function prototypes.
7225 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7226 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7228 2018-07-27 Martin Sebor <msebor@redhat.com>
7230 PR tree-optimization/86696
7231 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7232 types, including enums.
7233 (handle_char_store): Be prepared for the above function to fail.
7235 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
7237 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7238 when optimization level is lower than 2 or optimize for size.
7240 2018-07-26 Martin Sebor <msebor@redhat.com>
7242 PR tree-optimization/86043
7243 PR tree-optimization/86042
7244 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7245 (get_string_cst_length): Rename...
7246 (get_min_string_length): ...to this. Add argument.
7247 (handle_char_store): Extend to handle multi-character stores by
7249 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
7250 * tree.h (initializer_zerop): Add argument.
7252 2018-07-26 Jakub Jelinek <jakub@redhat.com>
7255 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7256 declare target to variables if they have always,{to,from,tofrom} map
7259 2018-07-26 Martin Liska <mliska@suse.cz>
7262 * lto-wrapper.c: Add linker_output as prefix
7263 for ltrans_output_file.
7265 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
7267 PR rtl-optimization/85805
7268 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7269 value for hard registers if that was written in the same mode.
7271 2018-07-26 Martin Liska <mliska@suse.cz>
7273 PR gcov-profile/86536
7274 * gcov.c (format_gcov): Use printf format %.*f directly
7275 and do not handle special values.
7277 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7279 * common/config/arc/arc-common.c (arc_option_optimization_table):
7280 Update default optimizations for size.
7282 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7284 * config/arc/arc.md (movsf_insn): Add short instruction selection.
7285 * config/arc/constraints.md (CfZ): New constraint.
7286 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7287 (subsf3_fpu): Likewise.
7288 (cmpsf_fpu): Likewise.
7289 (cmpsf_fpu_uneq): Likewise.
7291 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7293 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7295 (arc_is_uncached_mem_p): uncached applies to both the variable and
7298 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7300 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7303 2018-07-25 David Malcolm <dmalcolm@redhat.com>
7305 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7306 field "m_scopes" from vec to auto_vec.
7308 2018-07-25 Martin Liska <mliska@suse.cz>
7310 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7313 2018-07-25 Richard Biener <rguenther@suse.de>
7316 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7317 special wrt context_die late.
7318 (gen_subprogram_die): Re-use DIEs in local scope.
7320 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
7322 PR tree-optimization/86644
7323 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
7325 2018-07-25 Martin Liska <mliska@suse.cz>
7328 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7329 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7331 2018-07-25 Martin Liska <mliska@suse.cz>
7334 * params.def: Explain ASan abbreviation and provide
7335 a documentation link.
7337 2018-07-24 Martin Sebor <msebor@redhat.com>
7339 PR tree-optimization/86622
7340 PR tree-optimization/86532
7341 * builtins.h (string_length): Declare.
7342 * builtins.c (c_strlen): Correct handling of non-constant offsets.
7343 (check_access): Be prepared for non-constant length ranges.
7344 (string_length): Make extern.
7345 * expr.c (string_constant): Only handle the minor non-constant
7346 array index. Use string_constant to compute the length of
7347 a generic string constant.
7349 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
7351 PR tree-optimization/86618
7352 * tree-vect-stmts.c (vectorizable_call): Don't take the address
7353 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7355 2018-07-24 David Malcolm <dmalcolm@redhat.com>
7357 PR tree-optimization/86636
7358 * json.cc (json::object::set): Fix comment. Add assertions.
7359 (json::array::append): Move here from json.h. Add comment and an
7361 (json::string::string): Likewise.
7362 * json.h (json::array::append): Move to json.cc.
7363 (json::string::string): Likewise.
7364 * optinfo-emit-json.cc
7365 (optrecord_json_writer::impl_location_to_json): Assert that we
7366 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7367 wrapper around it. Expand the location once, rather than three
7369 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7370 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7372 (optrecord_json_writer::optinfo_to_json): Likewise, in four
7373 places. Fix some overlong lines.
7375 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
7377 * config/aarch64/aarch64-simd.md
7378 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7379 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7380 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7381 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7382 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7383 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7384 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7385 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7386 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7388 2018-07-24 Jakub Jelinek <jakub@redhat.com>
7391 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7392 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
7393 and abs_d == d, do the power of two handling if profitable.
7395 2018-07-24 Richard Biener <rguenther@suse.de>
7397 * match.pd: Add BIT_FIELD_REF canonicalizations.
7399 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7402 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7404 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7406 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7407 terminated STRING_CST object.
7409 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7411 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7413 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7415 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7416 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7417 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7418 the elements into a register.
7419 (rs6000_split_v4si_init_di_reg): Delete.
7420 (rs6000_split_v4si_init): Delete.
7421 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7422 (vsx_init_v4si): Rewrite as a define_expand.
7424 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7426 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7427 zero_extend argument from memory): New.
7429 2018-07-22 Martin Sebor <msebor@redhat.com>
7432 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7433 diagnosing calls with unknown arguments unless -Walloca-larger-than
7434 is restricted to less than PTRDIFF_MAX bytes.
7436 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
7438 * doc/gcov.texi (Invoking Gcov): Editorial changes.
7440 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7442 * pretty-print.c (text_info::set_location): Remove redundant
7443 "line_table" parameter from call to rich_location::set_range.
7445 2018-07-20 Martin Sebor <msebor@redhat.com>
7448 * builtins.c (expand_builtin_alloca): Adjust.
7449 * calls.c (alloc_max_size): Simplify.
7450 * cgraphunit.c (cgraph_node::expand): Adjust.
7451 * common.opt (larger_than_size, warn_frame_larger_than): Remove
7453 (frame_larger_than_size): Same.
7454 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7455 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7457 * doc/invoke.texi (GCC Command Options): Document option arguments.
7458 Explain byte-size arguments and suffixes.
7459 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7460 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7461 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7462 * doc/options.texi (UInteger): Expand.
7463 (Host_Wide_Int, ByteSize): Document new properties.
7464 * final.c (final_start_function_1): Include sizes in an error message.
7465 * function.c (frame_offset_overflow): Same.
7466 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7467 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7468 Diagnose unbounded alloca calls only for limits of less than
7470 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
7471 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
7473 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
7474 only for limits of less than PTRDIFF_MAX.
7475 * langhooks-def.h (lhd_handle_option): Change function argument
7477 * langhooks.c (lhd_handle_option): Same.
7478 * langhooks.h (handle_option): Same.
7479 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7481 (var_type, var_type_struct): Same.
7482 (var_set): Handle ByteSize flag.
7483 * optc-gen.awk: Add comments to output to ease debugging. Make
7484 use of HOST_WIDE_INT where appropriate.
7485 * opts-gen-save.awk: Use %lx to format unsigned long.
7486 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7487 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7488 arguments. Parse bytes-size suffixes.
7489 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7490 (enum_value_to_arg): Same.
7491 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
7492 (handle_option): Adjust.
7493 (generate_option): Change function argument to HOST_WIDE_INT.
7494 (cmdline_handle_error): Adjust.
7495 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7496 (set_option): Change function argument to HOST_WIDE_INT.
7497 (option_enabled): Handle cl_host_wide_int.
7498 (get_option_state): Handle CLVC_SIZE.
7499 (control_warning_option): Same.
7500 * opts.c (common_handle_option): Change function argument to
7501 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
7502 OPT_Wvla_larger_than_.
7503 * opts.h (enum cl_var_type): Add an enumerator.
7504 * stor-layout.c (layout_decl): Print a more meaningful warning.
7505 * toplev.c (output_stack_usage): Adjust.
7507 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
7509 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7510 call to inline_expand_builtin_string_cmp.
7511 (expand_builtin_strcmp): Likewise.
7512 (expand_builtin_strncmp): Likewise.
7513 (inline_string_cmp): Delete the last parameter, change char_type_node
7514 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7516 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7517 the inlining expansion on target where the type of the call has same or
7518 narrower precision than unsigned char.
7520 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7522 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7523 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7524 * common.opt (fsave-optimization-record): New option.
7525 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7526 * doc/invoke.texi (-fsave-optimization-record): New option.
7527 * dumpfile.c: Include "optinfo-emit-json.h".
7528 (struct kv_pair): Move to coretypes.h.
7529 (optgroup_options): Make non-static.
7530 (dump_context::end_scope): Call
7531 optimization_records_maybe_pop_dump_scope.
7532 * dumpfile.h (optgroup_options): New decl.
7533 * json.cc: New file.
7535 * optinfo-emit-json.cc: New file.
7536 * optinfo-emit-json.h: New file.
7537 * optinfo.cc: Include "optinfo-emit-json.h".
7538 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7539 (optinfo_enabled_p): Check optimization_records_enabled_p.
7540 (optinfo_wants_inlining_info_p): Likewise.
7541 * optinfo.h: Update comment.
7542 * profile-count.c (profile_quality_as_string): New function.
7543 * profile-count.h (profile_quality_as_string): New decl.
7544 (profile_count::quality): New accessor.
7545 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7546 and optinfo_emit_json_cc_tests.
7547 * selftest.h (selftest::json_cc_tests): New decl.
7548 (selftest::optinfo_emit_json_cc_tests): New decl.
7549 * toplev.c: Include "optinfo-emit-json.h".
7550 (compile_file): Call optimization_records_finish.
7551 (do_compile): Call optimization_records_start.
7552 * tree-ssa-live.c: Include optinfo.h.
7553 (remove_unused_scope_block_p): Retain inlining information if
7554 optinfo_wants_inlining_info_p returns true.
7556 2018-07-20 Richard Biener <rguenther@suse.de>
7559 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7560 to cover -flto-partition=none.
7562 2018-07-20 Martin Liska <mliska@suse.cz>
7564 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7565 (get_decl_source_range): Remove unused function.
7567 2018-07-20 Richard Biener <rguenther@suse.de>
7569 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7570 (struct vn_phi_s): Likewise.
7571 (struct vn_reference_s): Likewise.
7572 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7573 for searching the slot of an entry known to be in the hash itself.
7574 (vn_phi_hasher::equal): Likewise.
7575 (vn_reference_hasher::equal): Likewise.
7576 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7578 (optimistic_info, current_info): Remove, keeping only valid_info.
7579 (vn_reference_lookup_1): Remove fallback lookup.
7580 (vn_reference_lookup_2): Likewise.
7581 (vn_nary_op_lookup_1): Likewise.
7582 (vn_phi_lookup): Likewise.
7583 (vn_nary_build_or_lookup_1): Make sure to not chain the built
7585 (vn_reference_insert): Adjust, chain the inserted hash element
7586 at last_inserted_ref.
7587 (vn_reference_insert_pieces): Likewise.
7588 (visit_reference_op_call): Likewise.
7589 (vn_nary_op_insert_into): Chain the inserted hash element at
7591 (vn_nary_op_insert_pieces): Adjust.
7592 (vn_nary_op_insert): Likewise.
7593 (vn_nary_op_insert_stmt): Likewise.
7594 (vn_phi_insert): Adjust, chain the inserted hash element at
7596 (process_scc): Remove clearing and copying the optimistic
7597 table. Instead remove elements inserted during an optimistic
7598 iteration from the single table we maintain.
7599 (init_scc_vn): Adjust.
7600 (free_scc_vn): Likewise.
7601 (sccvn_dom_walker::record_cond): Likewise.
7602 (sccvn_dom_walker::after_dom_children): Likewise.
7604 2018-07-19 Martin Sebor <msebor@redhat.com>
7606 PR tree-optimization/84047
7607 PR tree-optimization/83776
7608 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7609 (check_array_bounds): Call it.
7611 2018-07-19 Martin Sebor <msebor@redhat.com>
7613 * align.h (align_flags): Use member initialization.
7615 2018-07-19 David Malcolm <dmalcolm@redhat.com>
7617 * Makefile.in (OBJS): Add optinfo.o.
7618 * coretypes.h (class symtab_node): New forward decl.
7619 (struct cgraph_node): New forward decl.
7620 (class varpool_node): New forward decl.
7621 * dump-context.h: New file.
7622 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7624 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7625 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7626 (set_alt_dump_file): Likewise.
7627 (dump_context::~dump_context): New dtor.
7628 (dump_gimple_stmt): Move implementation to...
7629 (dump_context::dump_gimple_stmt): ...this new member function.
7630 Add the stmt to any pending optinfo, creating one if need be.
7631 (dump_gimple_stmt_loc): Move implementation to...
7632 (dump_context::dump_gimple_stmt_loc): ...this new member function.
7633 Start a new optinfo and add the stmt to it.
7634 (dump_gimple_expr): Move implementation to...
7635 (dump_context::dump_gimple_expr): ...this new member function.
7636 Add the stmt to any pending optinfo, creating one if need be.
7637 (dump_gimple_expr_loc): Move implementation to...
7638 (dump_context::dump_gimple_expr_loc): ...this new member function.
7639 Start a new optinfo and add the stmt to it.
7640 (dump_generic_expr): Move implementation to...
7641 (dump_context::dump_generic_expr): ...this new member function.
7642 Add the tree to any pending optinfo, creating one if need be.
7643 (dump_generic_expr_loc): Move implementation to...
7644 (dump_context::dump_generic_expr_loc): ...this new member
7645 function. Add the tree to any pending optinfo, creating one if
7647 (dump_printf): Move implementation to...
7648 (dump_context::dump_printf_va): ...this new member function. Add
7649 the text to any pending optinfo, creating one if need be.
7650 (dump_printf_loc): Move implementation to...
7651 (dump_context::dump_printf_loc_va): ...this new member function.
7652 Start a new optinfo and add the stmt to it.
7653 (dump_dec): Move implementation to...
7654 (dump_context::dump_dec): ...this new member function. Add the
7655 value to any pending optinfo, creating one if need be.
7656 (dump_context::dump_symtab_node): New member function.
7657 (dump_context::get_scope_depth): New member function.
7658 (dump_context::begin_scope): New member function.
7659 (dump_context::end_scope): New member function.
7660 (dump_context::ensure_pending_optinfo): New member function.
7661 (dump_context::begin_next_optinfo): New member function.
7662 (dump_context::end_any_optinfo): New member function.
7663 (dump_context::s_current): New global.
7664 (dump_context::s_default): New global.
7665 (dump_scope_depth): Delete global.
7666 (dumpfile_ensure_any_optinfo_are_flushed): New function.
7667 (dump_symtab_node): New function.
7668 (get_dump_scope_depth): Reimplement in terms of dump_context.
7669 (dump_begin_scope): Likewise.
7670 (dump_end_scope): Likewise.
7671 (selftest::temp_dump_context::temp_dump_context): New ctor.
7672 (selftest::temp_dump_context::~temp_dump_context): New dtor.
7673 (selftest::verify_item): New function.
7674 (ASSERT_IS_TEXT): New macro.
7675 (ASSERT_IS_TREE): New macro.
7676 (ASSERT_IS_GIMPLE): New macro.
7677 (selftest::test_capture_of_dump_calls): New test.
7678 (selftest::dumpfile_c_tests): Call it.
7679 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7680 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7681 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7682 descriptive comment.
7683 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7684 (dump_node, dump_bb): Move these unrelated decls.
7685 (class dump_manager): Add leading comment.
7686 * optinfo.cc: New file.
7687 * optinfo.h: New file.
7689 2018-07-19 Michael Collison <michael.collison@arm.com>
7690 Richard Henderson <rth@redhat.com>
7692 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7693 (subti): Handle op1 zero.
7694 (subvti4, usub4ti4): New.
7695 (*sub<GPI>3_compare1_imm): New.
7696 (sub<GPI>3_carryinCV): New.
7697 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7698 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7700 2018-07-19 Michael Collison <michael.collison@arm.com>
7701 Richard Henderson <rth@redhat.com>
7703 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7704 (addti3): Create simpler code if low part is already known to be 0.
7705 (addvti4, uaddvti4): New.
7706 (*add<GPI>3_compareC_cconly_imm): New.
7707 (*add<GPI>3_compareC_cconly): New.
7708 (*add<GPI>3_compareC_imm): New.
7709 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7710 handle constants within this pattern..
7711 (*add<GPI>3_compareV_cconly_imm): New.
7712 (*add<GPI>3_compareV_cconly): New.
7713 (*add<GPI>3_compareV_imm): New.
7714 (add<GPI>3_compareV): New.
7715 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7716 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7717 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7718 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7719 with 'comparison' operator.
7720 (*add<GPI>3_compareV_cconly_imm): Ditto.
7721 (*add<GPI>3_compareV_cconly): Ditto.
7722 (*add<GPI>3_compareV_imm): Ditto.
7723 (add<GPI>3_compareV): Ditto.
7724 (add<mode>3_carryinC): Ditto.
7725 (*add<mode>3_carryinC_zero): Ditto.
7726 (*add<mode>3_carryinC): Ditto.
7727 (add<mode>3_carryinV): Ditto.
7728 (*add<mode>3_carryinV_zero): Ditto.
7729 (*add<mode>3_carryinV): Ditto.
7731 2018-07-19 Michael Collison <michael.collison@arm.com>
7732 Richard Henderson <rth@redhat.com>
7734 * config/aarch64/aarch64-modes.def (CC_V): New.
7735 * config/aarch64/aarch64-protos.h
7736 (aarch64_addti_scratch_regs): Declare
7737 (aarch64_subvti_scratch_regs): Declare.
7738 (aarch64_expand_subvti): Declare.
7739 (aarch64_gen_unlikely_cbranch): Declare
7740 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7741 for signed overflow using CC_Vmode.
7742 (aarch64_get_condition_code_1): Handle CC_Vmode.
7743 (aarch64_gen_unlikely_cbranch): New function.
7744 (aarch64_addti_scratch_regs): New function.
7745 (aarch64_subvti_scratch_regs): New function.
7746 (aarch64_expand_subvti): New function.
7748 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7750 * config/aarch64/aarch64-option-extensions.def: New entry for profile
7752 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7753 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7756 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7759 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7760 address check not strict.
7762 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7764 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7766 (store_pair_lanes<mode>): Likewise.
7767 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7768 enum value 'ADDR_QUERY_LDP_STP_N'.
7769 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7770 (aarch64_print_address_internal): Add declaration.
7771 (aarch64_print_ldpstp_address): Remove.
7772 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7773 (aarch64_print_operand): Change printing of 'y'.
7774 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7775 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7776 'true' rather than '1'.
7777 * config/aarch64/constraints.md (Uml): Likewise.
7778 (Uml): Rename to Umn.
7781 2018-07-19 Richard Biener <rguenther@suse.de>
7783 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7785 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7786 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7787 (vn_reference_hasher): Likewise.
7788 (struct vn_tables_s): Remove obstack and alloc-pool members.
7789 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7790 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7791 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7792 (vn_reference_insert_pieces): Likewise.
7793 (alloc_vn_nary_op_noinit): Adjust.
7794 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7795 (vn_phi_eq): Adjust.
7796 (shared_lookup_phiargs): Remove.
7797 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7798 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7799 (visit_reference_op_call): Likewise.
7800 (copy_nary, copy_phi, copy_reference): Remove.
7801 (process_scc): Rewind the obstack when iterating. Do not
7802 copy the elements to valid_info but just move them from one
7803 hashtable to the other.
7804 (allocate_vn_table): Adjust.
7805 (free_vn_table): Likewise.
7806 (init_scc_vn): Likewise.
7807 (free_scc_vn): Likewise.
7809 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
7812 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7813 indirect_return as function type attribute.
7814 (ix86_attribute_table): Change indirect_return to function
7816 * doc/extend.texi: Update indirect_return attribute.
7818 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
7820 * wide-int.h (widest2_int): New.
7821 * gimple-fold.c (arith_overflowed_p): Use it.
7822 * tree.h (widest2_int_cst): New.
7823 * tree-vrp.c (wide_int_binop_overflow): Rename from
7824 vrp_int_const_binop.
7825 Rewrite to work on trees.
7826 (extract_range_from_multiplicative_op_1): Abstract code to...
7827 (wide_int_range_min_max): ...here.
7828 (wide_int_range_cross_product): ...and here.
7829 (extract_range_from_binary_expr_1): Abstract overflow code to...
7830 (wide_int_range_mult_wrapping): ...here.
7831 * tree-vrp.h (wide_int_range_cross_product): New.
7832 (wide_int_range_mult_wrapping): New.
7834 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
7835 Julia Koval <julia.koval@intel.com>
7837 * config/i386/x86-tune-costs.h (skylake_memcpy,
7838 skylake_memset): Replace rep_prefix with unrolling for size 512.
7840 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
7843 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7844 comparision with EQ_EXPR in last stmt.
7846 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
7848 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7849 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7850 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7851 previously known as "PowerPC AltiVec Built-in Functions". Move
7852 some material to new subsubsections "PowerPC AltiVec Built-in
7853 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7855 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7856 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7857 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7858 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7860 2018-07-18 Richard Biener <rguenther@suse.de>
7862 PR tree-optimization/86557
7863 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7866 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
7868 * config/s390/s390.c (s390_function_profiler): Generate CFI.
7870 2018-07-17 Jeff Law <law@redhat.com>
7872 * config/arm/arm.c (get_label_padding): Update for recent
7873 changes to label_to_alignment.
7875 PR tree-optimization/86010
7876 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7878 * config/mips/mips.c (vr4130_align_insns): Update for recent
7879 changes to label_to_alignment.
7881 * config/frv/frv.c (frv_label_align): Update for recent changes
7882 to label_to_alignment.
7884 * config/nios2/nios2.c (nios2_label_align): Update for recent
7885 changes which dropped ALIGN_LABELS_LOG.
7887 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
7889 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7890 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7892 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
7894 * config/arc/arc.c (arc_label_align): Use align_labels instead of
7895 deprecated align_labels_log.
7897 2018-07-17 Richard Biener <rguenther@suse.de>
7900 * dwarf2out.c (init_sections_and_labels): Always generate
7901 a debug_line_str_section for early LTO debug.
7902 (dwarf2out_finish): Reset debug_line_str_hash output early.
7903 Bump counter for extra dwarf5 .debug_loc labels to not conflict
7905 (dwarf2out_early_finish): Output debug_line_str.
7907 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7909 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7910 index register on z196 or later.
7912 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7914 * config/s390/s390.c (s390_default_align): Set default function
7916 (s390_override_options_after_change): Call s390_default align.
7917 (s390_option_override_internal): Call s390_default_align.
7918 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7920 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7923 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7924 to _looptemp_ clauses, other than the first two.
7926 2018-07-17 Martin Liska <mliska@suse.cz>
7928 * opts.c: Do not enable OPT_falign_* for -Os.
7930 2018-07-17 Martin Liska <mliska@suse.cz>
7932 * align.h (MAX_CODE_ALIGN): New.
7933 (MAX_CODE_ALIGN_VALUE): New.
7934 * common/config/i386/i386-common.c (ix86_handle_option):
7935 (MAX_CODE_ALIGN): Moved to align.h.
7936 * final.c (MAX_CODE_ALIGN): Likewise.
7937 * opts.c (parse_and_check_align_values):
7938 (MAX_CODE_ALIGN): Likewise.
7939 (MAX_CODE_ALIGN_VALUE): Likewise.
7941 2018-07-17 Martin Liska <mliska@suse.cz>
7943 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7944 in order to fulfil coding style.
7945 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7946 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7947 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7948 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7949 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7950 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7951 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7952 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7953 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7955 2018-07-17 Martin Liska <mliska@suse.cz>
7957 * align.h: New file.
7958 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7960 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7961 align_flags of label_to_alignment.
7962 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7964 * config/m68k/m68k.c: Do not use removed align_labels_value and
7966 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7967 (LOOP_ALIGN): Likewise.
7968 (LABEL_ALIGN): Likewise.
7969 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7970 Remove not used macro.
7971 (rs6000_loop_align): Change return type to align_flags.
7972 (rs6000_loop_align_max_skip): Remove.
7973 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7974 Change return type to align_flags.
7975 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7976 Remove not used macro.
7977 (rs6000_loop_align): Change return type to align_flags.
7978 (rs6000_loop_align_max_skip): Remove.
7979 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7980 * config/rx/rx-protos.h (rx_align_for_label): Make it
7982 * config/rx/rx.c (rx_align_for_label): Change return type
7984 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7986 into align_flags class.
7987 (LABEL_ALIGN): Likewise.
7988 (LOOP_ALIGN): Likewise.
7989 * config/s390/s390.c (s390_label_align): Use align_flags
7991 (s390_asm_output_function_label): Likewise.
7992 * config/sh/sh.c (sh_override_options_after_change):
7993 Use align_flags class directly without macros.
7994 (find_barrier): Likewise.
7995 (barrier_align): Likewise.
7996 (sh_loop_align): Likewise.
7997 * config/spu/spu.c (spu_option_override):
7998 Use align_flags_tuple::get_value instead of removed macros.
7999 (spu_sched_init): Likewise.
8000 * config/spu/spu.h (GTY): Likewise.
8001 * config/visium/visium.c (visium_option_override):
8002 Set "8" as default secondary alignment.
8003 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8004 in order to guarantee secondary alignment of 8.
8005 * coretypes.h: Include align.h header file.
8006 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8007 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8008 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8009 * doc/tm.texi.in: Likewise.
8010 * final.c (struct label_alignment): Remove not used structure.
8011 (LABEL_ALIGN): Change type to align_flags.
8012 (LOOP_ALIGN): Likewise.
8013 (JUMP_ALIGN): Likewise.
8014 (default_loop_align_max_skip): Remove.
8015 (default_label_align_max_skip): Likewise.
8016 (default_jump_align_max_skip): Likewise.
8017 (default_label_align_after_barrier_max_skip):
8018 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8019 (LABEL_TO_MAX_SKIP): Remove.
8020 (label_to_alignment): Return align_flags type instead of integer.
8021 (label_to_max_skip): Remove.
8022 (align_fuzz): Use align_flags type.
8023 (compute_alignments): Use align_flags type and use align_flags::max
8024 to combine multiple alignments.
8025 (grow_label_align): Grow vec instead of C array.
8026 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8027 (shorten_branches): Use align_flags type and use align_flags::max
8028 to combine multiple alignments.
8029 (final_scan_insn_1): Remove usage of secondary alignment that comes
8030 from label alignment, but instead use proper secondary alignment
8031 which is computed in grow_label_align.
8032 * flags.h (struct align_flags_tuple): Move to align.h.
8033 (struct align_flags): Likewise.
8034 (state_align_loops): Rename to align_loops.
8035 (state_align_jumps): Rename to align_jumps.
8036 (state_align_labels): Rename to align_labels.
8037 (state_align_functions): Rename to align_functions.
8038 (align_loops_log): Remove.
8039 (align_jumps_log): Remove.
8040 (align_labels_log): Remove.
8041 (align_functions_log): Remove.
8042 (align_loops_max_skip): Remove.
8043 (align_jumps_max_skip): Remove.
8044 (align_labels_max_skip): Remove.
8045 (align_functions_max_skip): Remove.
8046 (align_loops_value): Remove.
8047 (align_jumps_value): Remove.
8048 (align_labels_value): Remove.
8049 (align_functions_value): Remove.
8050 * output.h (label_to_alignment): Change return type to align_flags.
8051 (label_to_max_skip): Remove.
8052 * target.def: Remove loop_align_max_skip, label_align_max_skip,
8053 jump_align_max_skip macros.
8054 * targhooks.h (default_loop_align_max_skip): Remove.
8055 (default_label_align_max_skip): Likewise.
8056 (default_jump_align_max_skip): Likewise.
8057 (default_label_align_after_barrier_max_skip): Remove.
8058 * toplev.c (read_log_maxskip): Use ::normalize function.
8059 (parse_N_M): Remove not used argument and also call ::normalize.
8060 (parse_alignment_opts): Do not pass unused arguments.
8061 * varasm.c (assemble_start_function): Use directly align_functions
8062 instead of removed macros.
8063 * system.h: Do not poison removed macros.
8065 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8068 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8069 and cond temporaries don't have reference type if iterator has
8070 pointer type. For init use &for_pre_body instead of pre_p if
8071 for_pre_body is non-empty.
8073 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8075 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8076 double-double modes to SFmode directly directly.
8077 (trunc<mode>sf2_fprs): Delete.
8079 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8081 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8082 for conversions between IFmode and the decimal floating point modes.
8083 (init_float128_ieee): Use the correct names for conversions between
8084 KFmode and the decimal floating point modes.
8086 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8088 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8089 for the conversions between TDmode and IFmode.
8090 (init_float128_ieee): Use more correct names for the conversions
8091 between TDmode and KFmode.
8093 2018-07-16 Jakub Jelinek <jakub@redhat.com>
8095 PR tree-optimization/86526
8096 * builtins.c (expand_builtin_memcmp): Formatting fixes.
8097 (inline_expand_builtin_string_cmp): Likewise.
8098 (inline_string_cmp): Likewise. Use c_readstr instead of
8099 builtin_memcpy_read_str. Add unit_mode temporary.
8101 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8104 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8105 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8107 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8109 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8110 Alphabetize prototypes of built-in functions, separating out
8111 built-in functions that are listed in this section but should be
8112 described elsewhere.
8114 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
8117 * expmed.c (emit_store_flag): Do not emit setcc followed by a
8118 conditional move when trapping comparison was split to a
8119 non-trapping one (and vice versa).
8121 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8123 * config/s390/s390.c (s390_function_profiler): Generate nops
8124 instead of profiler call sequences.
8125 * config/s390/s390.opt: Add the new option.
8127 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8129 * config/s390/s390.c (s390_function_profiler): Generate
8130 __mcount_loc section.
8131 * config/s390/s390.opt: Add the new option.
8133 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8135 * common.opt: Add the new warning.
8136 * config/s390/s390.c (s390_function_profiler): Emit "brasl
8137 %r0,__fentry__" when -mfentry is specified.
8138 (s390_option_override_internal): Disallow -mfentry for 31-bit
8140 * config/s390/s390.opt: Add the new option.
8142 2018-07-16 Richard Biener <rguenther@suse.de>
8145 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8146 for function-local FUNCTION_DECL and RESULT_DECL immediately.
8148 2018-07-16 Martin Liska <mliska@suse.cz>
8151 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8154 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8156 * config/arc/arcHS.md: Update ARCHS scheduling rules.
8158 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8160 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8162 * config/arc/arc-cpus.def (hs4x): New cpu.
8164 * config/arc/arc-tables.opt: Regenerate.
8165 * config/arc/arc.c (arc_sched_issue_rate): New function.
8166 (TARGET_SCHED_ISSUE_RATE): Define.
8167 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8168 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8170 (attr tune): Add ARCHS4x tune values.
8171 (attr tune_dspmpy): Define.
8172 (*tst): Correct instruction type.
8173 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8174 * config/arc/arcHS4x.md: New file.
8175 * config/arc/fpu.md: Update instruction type attributes.
8176 * config/arc/t-multilib: Regenerate.
8178 2018-07-16 Tom de Vries <tdevries@suse.de>
8181 * var-tracking.c (vt_initialize): Fix pre_dec handling.
8183 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8185 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8188 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
8190 PR tree-optimization/86514
8191 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8192 conversion to a boolean type from a type with greater precision.
8194 2018-07-16 Tom de Vries <tdevries@suse.de>
8196 * var-tracking.c (vt_initialize): Print adjusted insn slim if
8197 dump_flags request TDF_SLIM.
8199 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
8201 * fold-const.c (int_const_binop_1): Abstract...
8202 (wide_int_binop): ...wide int code here.
8203 (poly_int_binop): ...poly int code here.
8204 Abstract the rest of int_const_binop_1 into int_const_binop.
8205 * fold-const.h (wide_int_binop): New.
8206 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8207 Remove useless PLUS/MINUS_EXPR case.
8208 (zero_nonzero_bits_from_vr): Move wide int code...
8209 (zero_nonzero_bits_from_bounds): ...here.
8210 (extract_range_from_binary_expr_1): Move mask optimization code...
8211 (range_easy_mask_min_max): ...here.
8212 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8213 (range_easy_mask_min_max): New.
8215 2018-07-15 Jeff Law <law@redhat.com>
8218 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8221 2018-07-14 Jim Wilson <jimw@sifive.com>
8223 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8225 2018-07-14 Paul Koning <ni1d@arrl.net>
8227 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8229 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
8231 * lto-streamer-out.c (copy_function_or_variable): Dump info about
8234 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
8235 Steve Munroe <munroesj52@gmail.com>
8237 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8238 (_mm_andnot_si128): Likewise.
8239 (_mm_or_si128): Likewise.
8240 (_mm_xor_si128): Likewise.
8242 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
8245 * builtins.c (expand_builtin_memcmp): Inline the calls first
8246 when result_eq is false.
8247 (expand_builtin_strcmp): Inline the calls first.
8248 (expand_builtin_strncmp): Likewise.
8249 (inline_string_cmp): New routine. Expand a string compare
8250 call by using a sequence of char comparison.
8251 (inline_expand_builtin_string_cmp): New routine. Inline expansion
8252 a call to str(n)cmp/memcmp.
8253 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8255 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8257 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
8259 * config/arm/driver-arm.c: Include arm-native.h.
8260 (host_detect_local_cpu): Use auto-generated data tables.
8261 (vendors, arm_cpu_table): Delete. Move part information to ...
8262 * config/arm/arm-cpus.in: ... here.
8263 * config/arm/parsecpu.awk (gen_native): New function.
8264 (vendor, part): New CPU fields.
8265 (END): Add support for building the native CPU detection tables.
8266 * config/arm/t-arm (arm-native.h): Add build rule.
8267 (driver-arm.o): Add dependency on arm-native.h.
8269 2018-07-13 Richard Biener <rguenther@suse.de>
8272 * match.pd (addr1 - addr2): Allow either of the operand to
8275 2018-07-13 Tom de Vries <tdevries@suse.de>
8277 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8280 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
8282 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8283 arrays instead of numbered variables.
8285 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
8287 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8288 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8290 2018-07-13 Richard Biener <rguenther@suse.de>
8293 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8294 instead of get_context_die.
8296 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
8297 Richard Biener <rguenther@suse.de>
8300 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8301 that the loop latch destination where phi is defined.
8303 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
8305 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8306 (riscv_expand_epilogue): Add assertion to check interrupt mode.
8307 (riscv_set_current_function): Extract getting interrupt type to new
8309 (riscv_get_interrupt_type): New function.
8310 (riscv_merge_decl_attributes): New function, checking interrupt type is
8312 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8314 2018-07-12 Paul Koning <ni1d@arrl.net>
8316 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8319 2018-07-12 Paul Koning <ni1d@arrl.net>
8321 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8322 zero reference, add doloop_end instead.
8323 * doc/md.texi (decrement_and_branch_until_zero): Remove.
8324 (Looping patterns): Remove decrement_and_branch_until_zero. Add
8325 detail for doloop_end.
8327 2018-07-12 Martin Sebor <msebor@redhat.com>
8330 * attribs.c (decl_attributes): Reject conflicting attributes before
8331 calling attribute handlers.
8333 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8335 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8337 (gcc::dump_manager::get_dump_file_name): likewise.
8338 (dump_begin): Likewise.
8339 * dumpfile.h (dump_begin): Update prototype.
8340 (gcc::dump_manager::get_dump_file_name,
8341 gcc::dump_manager::get_dump_file_name): Update prototype.
8343 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8345 * internal-fn.h (vectorizable_internal_fn_p): New function.
8346 * tree-vect-slp.c (compatible_calls_p): Likewise.
8347 (vect_build_slp_tree_1): Remove nops argument. Handle calls
8348 to internal functions.
8349 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8351 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8353 * fold-const.h (inverse_conditions_p): Declare.
8354 * fold-const.c (inverse_conditions_p): New function.
8355 * match.pd: Use inverse_conditions_p. Add folds of view_converts
8356 that test the inverse condition of a conditional internal function.
8357 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8358 * internal-fn.c (internal_fn_mask_index): Handle conditional
8360 (vectorized_internal_fn_supported_p): New function.
8361 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8362 (any_pred_load_store): Replace with...
8363 (need_to_predicate): ...this new variable.
8364 (redundant_ssa_names): New variable.
8365 (ifcvt_can_use_mask_load_store): Move initial checks to...
8366 (ifcvt_can_predicate): ...this new function. Handle tree codes
8367 for which a conditional internal function exists.
8368 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8369 instead of ifcvt_can_use_mask_load_store. Update after variable
8371 (predicate_load_or_store): New function, split out from
8372 predicate_mem_writes.
8373 (check_redundant_cond_expr): New function.
8374 (value_available_p): Likewise.
8375 (predicate_rhs_code): Likewise.
8376 (predicate_mem_writes): Rename to...
8377 (predicate_statements): ...this. Use predicate_load_or_store
8378 and predicate_rhs_code.
8379 (combine_blocks, tree_if_conversion): Update after above name changes.
8380 (ifcvt_local_dce): Handle redundant_ssa_names.
8381 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8382 general conditional functions.
8383 * tree-vect-stmts.c (vectorizable_call): Likewise.
8385 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8386 Alan Hayward <alan.hayward@arm.com>
8387 David Sherwood <david.sherwood@arm.com>
8389 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8390 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8391 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8392 plus and minus and convert them into IFN_COND_FMA-based sequences.
8393 (convert_mult_to_fma): Handle conditional plus and minus.
8395 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8397 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8398 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8399 (cond_fnms_optab): New optabs.
8400 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8402 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8403 * internal-fn.h (get_conditional_internal_fn): Declare.
8404 (get_unconditional_internal_fn): Likewise.
8405 * internal-fn.c (cond_ternary_direct): New macro.
8406 (expand_cond_ternary_optab_fn): Likewise.
8407 (direct_cond_ternary_optab_supported_p): Likewise.
8408 (FOR_EACH_COND_FN_PAIR): Likewise.
8409 (get_conditional_internal_fn): New function.
8410 (get_unconditional_internal_fn): Likewise.
8411 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8412 (gimple_match_op::gimple_match_op): Add a new overload for 5
8414 (gimple_match_op::set_op): Likewise.
8415 (gimple_resimplify5): Declare.
8416 * genmatch.c (decision_tree::gen): Generate simplifications for
8418 * gimple-match-head.c (gimple_simplify): Define an overload for
8419 5 operands. Handle calls with 5 arguments in the top-level overload.
8420 (convert_conditional_op): Handle conversions from unconditional
8421 internal functions to conditional ones.
8422 (gimple_resimplify5): New function.
8423 (build_call_internal): Pass a fifth operand.
8424 (maybe_push_res_to_seq): Likewise.
8425 (try_conditional_simplification): Try converting conditional
8426 internal functions to unconditional internal functions.
8427 Handle 3-operand unconditional forms.
8428 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8429 Define ternary equivalents of the current rules for binary conditional
8431 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8433 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8434 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8435 (optab): Handle them.
8436 (SVE_COND_FP_TERNARY): New int iterator.
8437 (sve_fmla_op, sve_fmad_op): New int attributes.
8438 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8439 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8440 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8442 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8444 * target.def (preferred_else_value): New target hook.
8445 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8446 * doc/tm.texi: Regenerate.
8447 * targhooks.h (default_preferred_else_value): Declare.
8448 * targhooks.c (default_preferred_else_value): New function.
8449 * internal-fn.h (conditional_internal_fn_code): Declare.
8450 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8451 (get_conditional_internal_fn): Use it.
8452 (conditional_internal_fn_code): New function.
8453 * gimple-match.h (gimple_match_cond): New struct.
8454 (gimple_match_op): Add a cond member function.
8455 (gimple_match_op::gimple_match_op): Update all forms to take a
8457 * genmatch.c (expr::gen_transform): Use the same condition as res_op
8458 for the suboperation, but don't specify a particular else_value.
8459 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8460 (visit_nary_op, visit_reference_op_load): Pass
8461 gimple_match_cond::UNCOND to the gimple_match_op constructor.
8462 * gimple-match-head.c: Include tree-eh.h
8463 (convert_conditional_op): New function.
8464 (maybe_resimplify_conditional_op): Likewise.
8465 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8466 (gimple_resimplify2): Likewise.
8467 (gimple_resimplify3): Likewise.
8468 (gimple_resimplify4): Likewise.
8469 (maybe_push_res_to_seq): Return null for conditional operations.
8470 (try_conditional_simplification): New function.
8471 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
8473 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8475 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8477 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8479 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8481 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8483 (hash_tree): Do not hash DECL_FCONTEXT
8484 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8485 Do not stream DECL_FCONTEXT.
8486 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8487 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8489 2018-07-12 Richard Biener <rguenther@suse.de>
8492 * dwarf2out.c (gen_block_die): Only output blocks when they have
8493 at least one !DECL_IGNORED_P variable.
8495 2018-07-12 Richard Biener <rguenther@suse.de>
8498 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8499 Remove -mieee-fp handling.
8501 2018-07-12 Richard Biener <rguenther@suse.de>
8503 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8504 left-over from last patch.
8506 2018-07-12 Jakub Jelinek <jakub@redhat.com>
8508 PR tree-optimization/86492
8509 * gimple-ssa-store-merging.c
8510 (imm_store_chain_info::coalesce_immediate_stores): Call
8511 check_no_overlap even for the merge_overlapping case. Formatting fix.
8513 2018-07-12 Richard Biener <rguenther@suse.de>
8516 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8517 move possibly trapping operations into the conditional.
8519 2018-07-12 Richard Biener <rguenther@suse.de>
8521 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8522 (vn_lookup_simplify_result): Remove recursion limit applied
8524 (vn_nary_build_or_lookup_1): Adjust.
8525 (try_to_simplify): Likewise.
8526 * gimple-match-head.c (gimple_resimplify1): Instead apply one
8528 (gimple_resimplify2): Likewise.
8529 (gimple_resimplify3): Likewise.
8530 (gimple_resimplify4): Likewise.
8532 2018-07-11 Jakub Jelinek <jakub@redhat.com>
8534 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8535 Use __mmask64 type instead of __mmask8 for __M argument.
8536 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8537 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8539 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8540 __mmask16 for __M argument.
8541 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8542 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8543 to __mmask16 instead of __mmask8.
8544 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8545 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8546 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8547 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8548 instead of __mmask16 for __U argument.
8549 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8550 __mmask16 instead of __mmask8 for __U argument.
8551 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8553 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8555 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8557 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8559 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8561 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8563 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8565 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8567 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8569 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8571 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8572 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8573 return type as well as __M argument type and all casts from __mmask8
8575 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8576 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8577 return type as well as __M argument type and all casts from __mmask8
8579 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8580 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8581 return type as well as __M argument type and all casts from __mmask8
8583 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8584 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8585 return type as well as __M argument type and all casts from __mmask8
8587 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8588 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8591 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
8593 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8594 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8597 2018-07-11 Paul Koning <ni1d@arrl.net>
8599 * doc/md.texi (define_subst): Document how multiple occurrences of
8600 the same argument in the replacement pattern are handled.
8602 2018-07-11 Paul Koning <ni1d@arrl.net>
8604 * doc/extend.texi (Common Variable Attributes): Move "mode" into
8606 (Common Type Attributes): Add "mode" attribute.
8608 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
8610 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8611 stream DECL_ORIGINAL_TYPE.
8612 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8613 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8614 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8615 Do not walk original type.
8616 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8618 (write_ts_decl_non_common_tree_pointers): Do not stream
8620 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8621 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8623 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8625 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8626 thread twice from the same starting edge.
8628 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8630 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8631 * gimple.c (gimple_call_nonnull_result_p): ...here...
8632 (gimple_call_nonnull_arg): ...and here.
8633 * gimple.h (gimple_call_nonnull_result_p): New.
8634 (gimple_call_nonnull_arg): New.
8636 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
8638 * config/arm/arm-cpus.in: Move information from fpu field of each
8639 cpu definition to the isa field.
8640 * config/arm/parsecpu.awk (fpu): Delete match rule.
8641 (gen_comm_data): Don't add bits from the CPU's FPU entry.
8643 2018-07-11 Richard Biener <rguenther@suse.de>
8646 * dwarf2out.c (init_sections_and_labels): Use
8647 output_asm_line_debug_info consistently.
8648 (dwarf2out_early_finish): Likewise.
8649 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8652 2018-07-11 Richard Biener <rguenther@suse.de>
8654 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8655 Rework father_bb setting in a way to avoid propagating constants
8656 multiple times on a loop body.
8658 2018-07-10 Mark Wielaard <mark@klomp.org>
8661 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8663 2018-07-10 Richard Biener <rguenther@suse.de>
8665 * hash-map.h (hash_map::iterator::operator*): Return
8666 references to key and value.
8668 2018-07-10 Jakub Jelinek <jakub@redhat.com>
8671 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8672 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8673 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8674 (gimplify_omp_for): For composite loops, move outer
8675 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8676 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
8677 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
8678 TREE_LIST for both the original class iterator and the "last" helper
8679 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
8680 loop, remember has_decl_expr from outer composite loops for the
8681 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8683 2018-07-09 Martin Sebor <msebor@redhat.com>
8687 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8688 accessing implicitly initialized array elements.
8689 * expr.c (string_constant): Handle string initializers of
8690 character arrays within aggregates.
8691 * gimple-fold.c (fold_array_ctor_reference): Add argument.
8692 Store element offset. As a special case, handle zero size.
8693 (fold_nonarray_ctor_reference): Same.
8694 (fold_ctor_reference): Add argument. Store subobject offset.
8695 * gimple-fold.h (fold_ctor_reference): Add argument.
8697 2018-07-09 Paul Koning <ni1d@arrl.net>
8699 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8700 (pdp11_insn_cost): New function.
8701 (pdp11_md_asm_adjust): New function.
8702 (TARGET_INVALID_WITHIN_DOLOOP): Define.
8703 (pdp11_rtx_costs): Update to match machine better.
8704 (output_addr_const_pdp11): Correct format mismatch warnings.
8705 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8706 * config/pdp11/pdp11.md: General change to add base_cost and/or
8707 length attributes for use by new pdp11_insn_cost function.
8708 (MIN_BRANCH): Correct definition.
8710 (doloop_end): Use standard pattern name for looping pattern.
8711 (doloop_end_nocc): New.
8712 (movsf): Add another constraint alternative.
8713 (zero_extendqihi2): Add constraint alternatives for not in place
8715 (zero_extendhisi2): Remove.
8716 (shift patterns): Add CC handling variants.
8720 (define_peephole2): New peephole to recognize mov that sets CC for
8723 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8725 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8726 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8729 2018-07-09 Jakub Jelinek <jakub@redhat.com>
8732 * real.c (real_nextafter): Return true if result is denormal.
8734 2018-07-09 Martin Liska <mliska@suse.cz>
8736 * common.opt: Add back wrongly removed attribute.
8738 2018-07-09 Richard Biener <rguenther@suse.de>
8741 * dwarf2out.c (gen_block_die): For an early generated DIE
8742 always output high/low PC attributes.
8744 2018-07-09 Tom de Vries <tdevries@suse.de>
8746 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8747 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8748 onto VAR_DECL with abstract origin.
8750 2018-07-07 Jim Wilson <jimw@sifive.com>
8752 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8754 2018-07-07 Tom de Vries <tdevries@suse.de>
8756 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8758 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
8760 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8762 (combine_bound): Use wide-int overflow calculation instead of
8764 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8766 * fold-const.c (int_const_binop_2): Same.
8767 (extract_muldiv_1): Same.
8768 (fold_div_compare): Same.
8769 (fold_abs_const): Same.
8771 * poly-int.h (add): Same.
8775 * predict.c (predict_iv_comparison): Same.
8776 * profile-count.c (slow_safe_scale_64bit): Same.
8777 * simplify-rtx.c (simplify_const_binary_operation): Same.
8778 * tree-chrec.c (tree_fold_binomial): Same.
8779 * tree-data-ref.c (split_constant_offset_1): Same.
8780 * tree-if-conv.c (idx_within_array_bound): Same.
8781 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8782 * tree-ssa-phiopt.c (minmax_replacement): Same.
8783 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8784 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8785 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8786 * wide-int.cc (wi::add_large): Same.
8787 (wi::mul_internal): Same.
8788 (wi::sub_large): Same.
8789 (wi::divmod_internal): Same.
8790 * wide-int.h: Change overflow type to overflow_type for neg, add,
8791 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8792 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8793 mul_internal, divmod_internal.
8794 (overflow_type): New enum.
8795 (accumulate_overflow): New.
8797 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8799 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8800 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8802 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8804 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8805 argument is checked for zero before entering loop, avoid checking again.
8807 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8809 * gimplify.h (generic_expr_could_trap_p): Set as global function.
8810 * gimplify.h (generic_expr_could_trap_p): Likwise.
8811 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8813 2018-07-06 Jakub Jelinek <jakub@redhat.com>
8815 PR tree-optimization/86401
8816 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8817 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8818 (fold_bit_and_mask): ... here. New helper function for match.pd.
8819 * fold-const.h (fold_bit_and_mask): Declare.
8820 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8822 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
8825 * target.def (translate_mode_attribute): New hook.
8826 * targhooks.h (default_translate_mode_attribute): Declare.
8827 * targhooks.c (default_translate_mode_attribute): New function.
8828 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8829 * doc/tm.texi: Regenerate.
8830 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8831 (rs6000_translate_mode_attribute): New function.
8833 2018-07-06 Paul Koning <ni1d@arrl.net>
8835 * doc/md.texi (define_split): Document DONE and FAIL.
8836 (define_peephole2): Ditto.
8838 2018-07-05 Jeff Law <law@redhat.com>
8840 PR tree-optimization/86010
8841 * tree-ssa-dse.c (compute_trims): More aggressively trim at
8842 both the head and tail of mem* and str* calls.
8844 2018-07-05 Jim Wilson <jimw@sifive.com>
8846 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8848 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
8850 * config/aarch64/aarch64-simd.md: correct flags text for
8851 MIN_EXPR replacement.
8853 2018-07-05 James Clarke <jrtc27@jrtc27.com>
8855 * configure: Regenerated.
8857 2018-07-05 Carl Love <cel@us.ibm.com>
8859 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8860 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8861 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8862 VSX_BUILTIN_DOUBLEL_V4SF.
8864 2018-07-05 Martin Sebor <msebor@redhat.com>
8867 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8868 than its domain to compute its the upper bound of a char array.
8870 2018-07-05 Nathan Sidwell <nathan@acm.org>
8872 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8873 * doc/cpp.texi: Update comment.
8874 * doc/tm.texi: Rebuilt.
8875 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8876 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8877 * doc/extend.texi (Backwards Compatibility): Clarify it is system
8878 headers affected by extern "C".
8879 * system.h: Poison NO_IMPLICIT_EXTERN_C.
8880 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8881 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8882 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8883 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8884 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8885 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8886 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8887 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8888 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8889 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8890 config/sparc/sp64-elf.h, config/spu/spu.h,
8891 config/stormy16/stormy16.h, config/v850/v850.h,
8892 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8893 define NO_IMPLICIT_EXTERN_C.
8894 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8896 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8899 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8900 instead of GET_MODE_SIZE when comparing Units.
8902 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8905 * rtlanal.c (set_noop_p): Constrain on mode change,
8906 include hard-reg-set.h
8908 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8910 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8912 2018-07-05 Jakub Jelinek <jakub@redhat.com>
8915 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8918 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8920 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8922 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8925 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8927 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8929 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
8932 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8934 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
8936 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8937 neg_*_op* variables.
8939 2018-07-04 Martin Liska <mliska@suse.cz>
8941 * tree-switch-conversion.c: Define
8942 max_ratio_for_speed and max_ratio_for_size constants.
8944 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8945 Martin Liska <mliska@suse.cz>
8950 * common.opt: Rename align options with 'str_' prefix.
8951 * common/config/i386/i386-common.c (set_malign_value): New
8953 (ix86_handle_option): Use it to set -falign-* options/
8954 * config/aarch64/aarch64-protos.h (struct tune_params): Change
8955 type from int to string.
8956 * config/aarch64/aarch64.c: Update default values from int
8958 * config/alpha/alpha.c (alpha_override_options_after_change):
8960 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8961 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8962 max skip conditionally.
8963 * config/i386/freebsd.h (SUBALIGN_LOG): New.
8964 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8965 max skip conditionally.
8966 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8967 max skip conditionally.
8968 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8969 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8970 max skip conditionally.
8971 * config/i386/i386.c (struct ptt): Change type from int to
8973 (ix86_default_align): Set default values.
8974 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8975 max skip conditionally.
8976 * config/i386/iamcu.h (SUBALIGN_LOG): New.
8977 (ASM_OUTPUT_MAX_SKIP_ALIGN):
8978 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8979 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8980 max skip conditionally.
8981 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8982 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8983 * config/i386/x86-64.h (SUBALIGN_LOG): New.
8984 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8985 max skip conditionally.
8986 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8987 * config/ia64/ia64.c (ia64_option_override): Set default values
8988 for alignment options.
8989 * config/m68k/m68k.c: Handle new str_align_* options.
8990 * config/mips/mips.c (mips_set_compression_mode): Change
8992 (mips_option_override): Set default values for options.
8993 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8995 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8997 * config/rx/rx.c (rx_option_override): Likewise.
8998 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8999 (LABEL_ALIGN): Use align_labels_log.
9000 (LOOP_ALIGN): Use align_loops_align.
9001 * config/s390/s390.c (s390_asm_output_function_label): Use new
9003 * config/sh/sh.c (sh_override_options_after_change):
9004 Change type of constants.
9005 * config/spu/spu.c (spu_sched_init): Likewise.
9006 * config/sparc/sparc.c (sparc_option_override): Set default
9008 * config/visium/visium.c (visium_option_override): Likewise.
9009 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9010 emit p2align format with last argument if it's not needed.
9011 * doc/invoke.texi: Document extended format of -falign-*.
9012 * final.c: Use align_labels alignment.
9013 * flags.h (struct target_flag_state): Change type to use
9015 (struct align_flags_tuple): New.
9016 (struct align_flags): Likewise.
9017 (align_loops_log): Redefine macro to use new types.
9018 (align_loops_max_skip): Redefine macro to use new types.
9019 (align_jumps_log): Redefine macro to use new types.
9020 (align_jumps_max_skip): Redefine macro to use new types.
9021 (align_labels_log): Redefine macro to use new types.
9022 (align_labels_max_skip): Redefine macro to use new types.
9023 (align_functions_log): Redefine macro to use new types.
9024 (align_loops): Redefine macro to use new types.
9025 (align_jumps): Redefine macro to use new types.
9026 (align_labels): Redefine macro to use new types.
9027 (align_functions): Redefine macro to use new types.
9028 (align_functions_max_skip): Redefine macro to use new types.
9029 (align_loops_value): New macro.
9030 (align_jumps_value): New macro.
9031 (align_labels_value): New macro.
9032 (align_functions_value): New macro.
9033 * function.c (invoke_set_current_function_hook): Propagate
9034 alignment values from flags to global variables default in
9036 * ipa-icf.c (sem_function::equals_wpa): Use
9037 cl_optimization_option_eq instead of memcmp.
9038 * lto-streamer.h (cl_optimization_stream_out): Support streaming
9040 (cl_optimization_stream_in): Likewise.
9041 * optc-save-gen.awk: Support strings in cl_optimization.
9042 * opth-gen.awk: Likewise.
9043 * opts.c (finish_options): Remove error checking of invalid
9045 (MAX_CODE_ALIGN): Remove.
9046 (MAX_CODE_ALIGN_VALUE): Likewise.
9047 (parse_and_check_align_values): New function.
9048 (check_alignment_argument): Likewise.
9049 (common_handle_option): Use check_alignment_argument.
9050 * opts.h (parse_and_check_align_values): Declare.
9051 * toplev.c (init_alignments): Remove.
9052 (read_log_maxskip): New.
9053 (parse_N_M): Likewise.
9054 (parse_alignment_opts): Likewise.
9055 (backend_init_target): Remove usage of init_alignments.
9056 * toplev.h (parse_alignment_opts): Declare.
9057 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9059 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9060 * tree.c (cl_option_hasher::equal): New.
9061 * varasm.c: Use new global macros.
9063 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9065 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9066 Use a simpler align directive also if MAXSKIP = ALIGN-1.
9067 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9068 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9069 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9070 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9071 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9072 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9073 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9074 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9075 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9076 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9077 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9079 2018-07-04 Martin Liska <mliska@suse.cz>
9080 Jonathan Wakely <jwakely@redhat.com>
9082 * coverage.c: Use correct type.
9083 * doc/invoke.texi: Language correction.
9085 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
9088 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9089 ENDBRANCH for non-tail call which may return via indirect branch.
9090 * doc/extend.texi: Document indirect_return attribute.
9092 2018-07-03 Martin Sebor <msebor@redhat.com>
9094 PR tree-optimization/86274
9095 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9097 (format_floating): Correct handling of infinities and NaNs.
9099 2018-07-03 Martin Sebor <msebor@redhat.com>
9101 * print-tree.c (print_real_cst): New function.
9102 (print_node_brief): Call it.
9103 (print_node): Ditto.
9105 2018-07-03 Jeff Law <law@redhat.com>
9107 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9108 into a single pattern.
9110 * config/h8300/h8300.md (ors code_iterator): New.
9111 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9112 a single pattern and single splitter.
9113 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9114 (iorqi3_1, xorqi3_1): Likewise.
9115 (iorqi3, xorqi3 expanders): Similarly.
9117 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9118 (movmd_internal) into a single pattern using the P mode iterator.
9119 (movmd splitters): Similarly.
9120 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9121 (movsd splitters): Similarly.
9123 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9124 ADDB, ADDW and ADDL into a single ADD attribute which selects the
9125 right table based on the size of the operand.
9126 * config/h8300/h8300.md (length_table): Corresponding changes. All
9127 references to "addb", "addw" and "addl" changed to "add".
9128 (btst patterns): Merge two variants into a single pattern.
9129 (tstqi, tsthi): Likewise.
9130 (addhi3_incdec, addsi3_incdec): Likewise.
9131 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9132 (mulhi3, mulsi3): Likewise.
9133 (udivhi3, udivsi3): Likewise.
9134 (divhi3, divsi3): Likewise.
9135 (andorqi3, andorhi3, andorsi3): Likewise.
9137 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
9140 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9141 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9143 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9145 PR tree-optimization/85694
9146 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9147 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9149 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9150 (<u>avg<mode>3_ceil): New patterns.
9152 2018-07-03 David Malcolm <dmalcolm@redhat.com>
9154 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9155 scan-tree-dump directive.
9156 * gcc.dg/vect/slp-perm-2.c: Likewise.
9157 * gcc.dg/vect/slp-perm-3.c: Likewise.
9158 * gcc.dg/vect/slp-perm-5.c: Likewise.
9159 * gcc.dg/vect/slp-perm-6.c: Likewise.
9160 * gcc.dg/vect/slp-perm-7.c: Likewise.
9161 * gcc.dg/vect/slp-perm-8.c: Likewise.
9163 2018-07-03 Marek Polacek <polacek@redhat.com>
9166 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9168 2018-07-03 Richard Biener <rguenther@suse.de>
9171 * tree-ssa-structalias.c (find_func_clobbers): Properly
9172 handle indirect calls.
9174 2018-07-03 Jeff Law <law@redhat.com>
9176 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9177 (shifts): New code iterator.
9178 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9179 expander. Fix HImode handling on H8/SX.
9180 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9181 (subqi3, subhi3, subsi3 expanders): Likewise.
9182 (andqi3, andhi3, andsi3 expanders): Likewise.
9183 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9184 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9185 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9186 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9187 (zero_extendqihi2, zero_extendqisi2): Likewise.
9188 (extendqihi2, extendqisi2): Likewise.
9189 (rotlqi3, rotlhi3, rotlsi3): Likewise.
9190 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9191 (rotlqi3_1, rotlhi3_1): Likewise.
9192 (logicalhi3_sn, logicalsi3_sn): Likewise.
9193 (logicalhi3, logicalsi3): Likewise.
9195 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9197 * tree-vect-patterns.c (vect_recog_rotate_pattern)
9198 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9199 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9200 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9201 type to append_pattern_def_seq instead of creating a stmt_vec_info
9203 (build_mask_conversion): Likewise. Remove vinfo argument.
9204 (vect_add_conversion_to_patterm): Likewise, renaming to...
9205 (vect_add_conversion_to_pattern): ...this.
9206 (vect_recog_mask_conversion_pattern): Update call to
9207 build_mask_conversion. Pass the vector type to
9208 append_pattern_def_seq here too.
9209 (vect_recog_gather_scatter_pattern): Update call to
9210 vect_add_conversion_to_pattern.
9212 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9214 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9215 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9216 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9217 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9218 STMT_VINFO_PATTERN_DEF_SEQ to null here.
9219 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9220 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9221 append_pattern_def_seq instead of new_pattern_def_seq.
9222 (vect_recog_divmod_pattern): Do both of the above.
9223 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9226 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9228 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9229 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9230 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9231 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9232 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9233 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9234 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9235 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9236 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9237 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9238 parameter with a single stmt_vec_info.
9239 (vect_recog_func_ptr): Likewise.
9240 (vect_recog_gather_scatter_pattern): Likewise, folding in...
9241 (vect_try_gather_scatter_pattern): ...this.
9242 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9243 the stmt_vec_info of the statement to be matched. Don't clear
9244 STMT_VINFO_RELATED_STMT.
9245 (vect_pattern_recog): Update call accordingly.
9247 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9249 PR tree-optimization/85694
9250 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9251 (uavgM3_ceil): Document new optabs.
9252 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9253 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9255 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9256 (savg_ceil_optab): New optabs.
9257 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9258 (vect_vect_recog_func_ptrs): Add it.
9259 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9260 constant directly from the associated lhs.
9262 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9264 * tree-vect-patterns.c (vect_split_statement): New function.
9265 (vect_convert_input): Use it to try to split an existing cast.
9267 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9269 * poly-int.h (print_hex): New function.
9270 * dumpfile.h (dump_dec, dump_hex): Declare.
9271 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9272 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9273 min_input_precision, operation_precision and operation_sign.
9274 * tree-vect-patterns.c (vect_get_range_info): New function.
9275 (vect_same_loop_or_bb_p, vect_single_imm_use)
9276 (vect_operation_fits_smaller_type): Delete.
9277 (vect_look_through_possible_promotion): Add an optional
9278 single_use_p parameter.
9279 (vect_recog_over_widening_pattern): Rewrite to use new
9280 stmt_vec_info infomration. Handle one operation at a time.
9281 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9282 (vect_truncatable_operation_p, vect_set_operation_type)
9283 (vect_set_min_input_precision): New functions.
9284 (vect_determine_min_output_precision_1): Likewise.
9285 (vect_determine_min_output_precision): Likewise.
9286 (vect_determine_precisions_from_range): Likewise.
9287 (vect_determine_precisions_from_users): Likewise.
9288 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9289 (vect_vect_recog_func_ptrs): Put over_widening first.
9291 (vect_pattern_recog): Call vect_determine_precisions.
9293 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9295 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9296 statements that have been replaced by further pattern statements.
9297 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9299 2018-07-03 Richard Biener <rguenther@suse.de>
9301 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9302 always set *dt. Dump vectype in vectype overload.
9303 * dumpfile.h (dump_gimple_expr): New function.
9304 (dump_gimple_expr_loc): Likewise.
9305 * dumpfile.c (dump_gimple_expr): New function.
9306 (dump_gimple_expr_loc): Likewise.
9308 2018-07-02 Jeff Law <law@redhat.com>
9310 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9311 the H8/300, H8/300H and H8/S variants into a single pattern.
9312 (movhi_h8300, movqi_h8300hs): Similarly.
9313 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9314 (QHI mode iterator): New.
9316 * config/h8300/h8300.md: Remove trailing whitespace.
9318 2018-07-02 Jim Wilson <jimw@sifive.com>
9320 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9321 instead of emit_insn for interrupt returns.
9322 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9323 (riscv_sret, riscv_uret): Likewise.
9325 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9327 * pretty-print.c (selftest::test_pp_format): Move save and restore
9328 of quotes to class auto_fix_quotes, and add an instance.
9329 * selftest.c: Include "intl.h".
9330 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9331 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9332 * selftest.h (selftest::auto_fix_quotes): New class.
9334 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
9336 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9337 (aarch64_sve_prepare_conditional_op): Remove.
9338 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9339 Allow aarch64_simd_reg_or_zero as select operand; remove
9340 the aarch64_sve_prepare_conditional_op call.
9341 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9342 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9343 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9344 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9345 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9346 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9347 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9348 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9349 and a splitters to match all of the *_any patterns.
9350 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9352 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9353 (SVE_COND_FP_BINARY_REV): Remove.
9354 (sve_int_op_rev, sve_fp_op_rev): New.
9355 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9356 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9357 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9358 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9359 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9360 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9361 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9362 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9363 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9365 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9366 Remove match_dup 1 from the inner unspec.
9367 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9369 * config/aarch64/aarch64.md (movprfx): New attr.
9370 (length): Default movprfx to 8.
9371 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9372 (*madd<SVE_I>, *msub<SVE_I): Likewise.
9373 (*<su>mul<SVE_I>3_highpart): Likewise.
9374 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9375 (*v<ASHIFT><SVE_I>3): Likewise.
9376 (*<su><MAXMIN><SVE_I>3): Likewise.
9377 (*<su><MAXMIN><SVE_F>3): Likewise.
9378 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9379 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9380 (*div<SVE_F>4): Likewise.
9382 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
9384 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9387 2018-07-02 Richard Biener <rguenther@suse.de>
9389 PR tree-optimization/86363
9390 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9391 memset argument refers to a non-variable address.
9393 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9395 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9396 {PLUS,MINUS}_EXPR code to...
9397 (adjust_symbolic_bound): ...here,
9398 (combine_bound): ...here,
9399 (set_value_range_with_overflow): ...and here.
9401 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9403 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9405 (extract_range_from_abs_expr): ...here.
9407 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
9409 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9410 -fno-omit-frame-pointer when not optimizing.
9412 2018-07-02 Martin Liska <mliska@suse.cz>
9415 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9416 (propagate_nothrow): Likewise.
9418 2018-07-02 Martin Liska <mliska@suse.cz>
9421 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9423 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9425 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9426 function in r262149, changing "loc" param from source_location to
9427 const dump_location_t &.
9428 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9429 declaration, as above.
9431 2018-07-01 Paul Koning <ni1d@arrl.net>
9433 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9434 -munit-asm, -mgnu-asm, -mdec-asm.
9435 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9436 (pdp11_output_labelref): New.
9437 (pdp11_output_def): New.
9438 (pdp11_output_addr_vec_elt): New.
9439 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
9440 %# and %@ format codes.
9441 (pdp11_option_override): New.
9442 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9443 (pdp11_output_ident): New.
9444 (pdp11_asm_named_section): New.
9445 (pdp11_asm_init_sections): New.
9446 (pdp11_file_start): New.
9447 (pdp11_file_end): New.
9448 (output_ascii): Use .ascii/.asciz for -mdec-asm.
9449 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
9450 %o, like %c but octal.
9451 (pdp11_option_override): New.
9452 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9454 (DATA_SECTION_ASM_OP): Ditto.
9455 (READONLY_DATA_SECTION_ASM_OP): New.
9456 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9457 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9458 (ASM_OUTPUT_LABELREF): Ditto.
9459 (ASM_OUTPUT_DEF): Ditto.
9460 (ASM_OUTPUT_EXTERNAL): New.
9461 (ASM_OUTPUT_SOURCE_FILENAME): New.
9462 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9463 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9464 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
9465 %# and %@ format codes.
9466 * config/pdp11/pdp11.opt (mgnu-asm): New.
9467 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9468 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9469 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9471 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
9473 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9474 dereferencing path[] beyond its length.
9476 (debug_all_paths): New.
9477 (rewire_first_differing_edge): New.
9478 (adjust_paths_after_duplication): New.
9479 (duplicate_thread_path): Call adjust_paths_after_duplication.
9481 (thread_through_all_blocks): Add new argument to
9482 duplicate_thread_path.
9484 2018-06-30 Jim Wilson <jimw@sifive.com>
9486 * config/riscv/predicates.md (p2m1_shift_operand): New.
9487 (high_mask_shift_operand): New.
9488 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9489 pattern using p2m1_shift_operand.
9490 (lshsi3_zero_extend_3+2): New combiner pattern using
9491 high_mask_shift_operand.
9493 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9495 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9497 (vect_recog_rotate_pattern): ...here.
9498 (vect_convert_input): Try to insert casts of invariants in the
9500 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9501 preheader to be empty.
9503 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9505 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9506 vector type. If given, install it in the new statement's
9508 (vect_element_precision): New function.
9509 (vect_unpromoted_value): New struct.
9510 (vect_unpromoted_value::vect_unpromoted_value): New function.
9511 (vect_unpromoted_value::set_op): Likewise.
9512 (vect_look_through_possible_promotion): Likewise.
9513 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9514 (vect_widened_op_tree, vect_convert_input): Likewise.
9515 (vect_convert_inputs, vect_convert_output): Likewise.
9516 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9517 to handle the optional cast of the multiplication result and
9518 vect_widened_op_tree to detect the widened multiplication itself.
9519 Do not require the input and output of promotion casts to have
9520 the same sign, but base the signedness of the operation on the
9521 input rather than the result. If the pattern includes two
9522 promotions, check that those promotions have the same sign.
9523 Do not restrict the MULT_EXPR handling to a double-width result;
9524 handle quadruple-width results and wider. Use vect_convert_inputs
9525 to convert the inputs to the common type.
9526 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
9527 to handle the optional cast of the ABS result. Also allow a sign
9528 change or a sign extension between the ABS and MINUS.
9529 Use vect_widened_op_tree to detect the widened subtraction and use
9530 vect_convert_inputs to convert the inputs to the common type.
9531 (vect_handle_widen_op_by_const): Delete.
9532 (vect_recog_widen_op_pattern): New function.
9533 (vect_recog_widen_mult_pattern): Use it.
9534 (vect_recog_widen_shift_pattern): Likewise.
9535 (vect_recog_widen_sum_pattern): Use
9536 vect_look_through_possible_promotion to handle the promoted
9539 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9541 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9542 the containing gimple_seq *.
9543 * gimple-iterator.h (gsi_for_stmt): Declare it.
9544 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9545 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9546 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9547 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9548 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9550 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9552 (vect_mark_pattern_stmts): ...here. Handle cases in which the
9553 statement being replaced is part of an existing pattern
9554 definition sequence, inserting the new pattern statements before
9556 (vect_pattern_recog_1): Don't return a bool. If the statement
9557 is already part of a pattern, instead apply pattern matching
9558 to the pattern definition statements. Don't clear the
9559 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9560 (vect_pattern_recog): Don't break after the first match;
9561 continue processing the pattern definition statements instead.
9562 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9564 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9566 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9567 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9568 (vect_recog_widen_sum_pattern): Use it.
9570 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9572 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9573 phi is not a pattern statement and has not been replaced by
9574 a pattern statement.
9575 * tree-vect-patterns.c (type_conversion_p): Don't check
9576 STMT_VINFO_IN_PATTERN_P.
9577 (vect_recog_vector_vector_shift_pattern): Likewise.
9578 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9579 the pattern statement rather than the original statement; check
9580 directly for a WIDEN_MULT_EXPR here.
9581 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9582 vect_is_simple_use to return the pattern statement rather
9583 than the original statement; use is_pattern_stmt_p to check
9584 for such a pattern statement.
9585 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9586 to return the pattern statement rather than the original statement;
9587 don't do the same transformation here.
9588 (vect_is_simple_use): If the defining statement has been replaced
9589 by a pattern statement, return the pattern statement instead.
9590 Remove the corresponding (local) transformation from the vectype
9593 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9595 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9596 end and default to null.
9597 * tree-vect-loop.c (vect_create_epilog_for_reduction)
9598 (vectorizable_reduction): Update calls accordingly, dropping the
9599 gimple ** argument if the passed-back statement isn't needed.
9600 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9601 (vect_recog_rotate_pattern): Likewise.
9602 (vect_recog_mask_conversion_pattern): Likewise.
9603 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9604 (vect_mask_constant_operand_p): Likewise.
9605 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9606 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9607 (get_group_load_store_type, get_load_store_type): Likewise.
9608 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9609 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9610 (vectorizable_conversion, vectorizable_assignment): Likewise.
9611 (vectorizable_shift, vectorizable_operation): Likewise.
9612 (vectorizable_store, vect_is_simple_cond): Likewise.
9613 (vectorizable_condition, vectorizable_comparison): Likewise.
9614 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9615 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9616 and move it to the end. Cope with null def_stmt_outs.
9618 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9620 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9622 2018-06-29 Jeff Law <law@redhat.com>
9624 * config/v850/v850.c (v850_legitimate_address_p): Handle large
9625 displacements for TARGET_V850E2V3 and newer.
9626 (TARGET_LRA_P): Remove. Defaults to LRA now.
9627 * config/v850/v850.md (sign23byte_load): Remove.
9628 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9629 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9631 2018-06-29 Martin Liska <mliska@suse.cz>
9634 * coverage.c (coverage_init): Mangle full path name.
9635 * doc/invoke.texi: Document the change.
9636 * gcov-io.c (mangle_path): New.
9637 * gcov-io.h (mangle_path): Likewise.
9638 * gcov.c (mangle_name): Use mangle_path for path mangling.
9640 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9642 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9643 if starting source register is not even.
9645 2018-06-29 Martin Liska <mliska@suse.cz>
9647 PR tree-optimization/86263
9648 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9649 Make edge redirection.
9651 2018-06-29 David Malcolm <dmalcolm@redhat.com>
9653 * dumpfile.c (dump_loc): Add indentation based on scope depth.
9654 (dump_scope_depth): New variable.
9655 (get_dump_scope_depth): New function.
9656 (dump_begin_scope): New function.
9657 (dump_end_scope): New function.
9658 * dumpfile.h (get_dump_scope_depth): New declaration.
9659 (dump_begin_scope): New declaration.
9660 (dump_end_scope): New declaration.
9661 (class auto_dump_scope): New class.
9662 (AUTO_DUMP_SCOPE): New macro.
9663 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9666 2018-06-29 Richard Biener <rguenther@suse.de>
9668 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9669 compute_all_dependences succeeds.
9670 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9671 exceed --param loop-max-datarefs-for-datadeps.
9673 2018-06-29 Jakub Jelinek <jakub@redhat.com>
9675 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9677 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
9680 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9681 alternative 0 in preferred_for_speed attribute.
9683 2018-06-28 Paul Koning <ni1d@arrl.net>
9685 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9686 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9687 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9688 * config/pdp11/pdp11.md: Correct "length" attribute calculation
9689 for shift insn patterns.
9691 2018-06-28 David Malcolm <dmalcolm@redhat.com>
9693 * cgraph.c (cgraph_node::get_body): Replace assignments to
9694 "dump_file" with calls to set_dump_file.
9695 * dumpfile.c (alt_dump_file): Make static, and group with...
9696 (alt_flags): ...this definition.
9697 (dumps_are_enabled): New variable.
9698 (refresh_dumps_are_enabled): New function.
9699 (set_dump_file): New function.
9700 (set_alt_dump_file): New function.
9701 (gcc::dump_manager::dump_start): Replace assignments to
9702 "dump_file" and "alt_dump_file" with calls to set_dump_file and
9704 (gcc::dump_manager::dump_finish): Likewise.
9705 * dumpfile.h (alt_dump_file): Delete decl.
9706 (dumps_are_enabled): New variable decl.
9707 (set_dump_file): New function decl.
9708 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9710 * tree-nested.c (lower_nested_functions): Replace assignments to
9711 "dump_file" with calls to set_dump_file.
9713 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9715 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9716 goto_locus of each outgoing edge of each basic block.
9718 2018-06-28 Richard Biener <rguenther@suse.de>
9720 * dwarf2out.c (decl_scope_table): Remove.
9721 (push_decl_scope): Likewise.
9722 (pop_decl_scope): Likewise.
9723 (gen_type_die_for_member): Do not call push/pop_decl_scope.
9724 (gen_struct_or_union_type_die): Likewise.
9725 (gen_tagged_type_die): Likewise.
9726 (dwarf2out_init): Do not initialize decl_scope_table.
9727 (dwarf2out_c_finalize): Do not free it.
9729 2018-06-28 Richard Biener <rguenther@suse.de>
9731 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9732 deciding whether to not re-use a DIE.
9734 2018-06-28 Richard Biener <rguenther@suse.de>
9736 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9737 DW_AT_abstract_origin attribute.
9739 2018-06-28 Martin Liska <mliska@suse.cz>
9741 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9742 Use newly introduced constants.
9743 * tree-switch-conversion.h (struct jump_table_cluster):
9744 Define max_ratio_for_size and max_ratio_for_speed.
9746 2018-06-28 Martin Liska <mliska@suse.cz>
9748 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9749 Add new checking assert to catch invalid state.
9750 (jump_table_cluster::can_be_handled): Handle single case
9752 (jump_table_cluster::is_beneficial): Bail out for such case.
9753 (bit_test_cluster::find_bit_tests):
9754 Add new checking assert to catch invalid state.
9755 (bit_test_cluster::can_be_handled): Handle single case
9757 (bit_test_cluster::is_beneficial): Bail out for such case.
9758 (switch_decision_tree::analyze_switch_statement):
9761 2018-06-28 Martin Liska <mliska@suse.cz>
9763 * common.opt: Introduce -completion option.
9764 * gcc.c (driver_handle_option): Handle it.
9765 (driver::main): Print completions if completion
9767 * opt-suggestions.c (option_proposer::get_completions):
9769 (option_proposer::suggest_completion): Likewise.
9770 (option_proposer::find_param_completions): Likewise.
9771 (verify_autocompletions): Likewise.
9772 (test_completion_valid_options): Likewise.
9773 (test_completion_valid_params): Likewise.
9774 (in_completion_p): Likewise.
9775 (empty_completion_p): Likewise.
9776 (test_completion_partial_match): Likewise.
9777 (test_completion_garbage): Likewise.
9778 (opt_proposer_c_tests): Likewise.
9779 * opt-suggestions.h: Declare new functions.
9780 * opts.c (common_handle_option): Handle OPT__completion_.
9781 * selftest-run-tests.c (selftest::run_tests): Add
9782 opt_proposer_c_tests.
9783 * selftest.c (assert_str_startswith): New.
9784 * selftest.h (assert_str_startswith): Likewise.
9785 (opt_proposer_c_tests): New.
9786 (ASSERT_STR_STARTSWITH): Likewise.
9788 2018-06-28 Martin Liska <mliska@suse.cz>
9790 * Makefile.in: Add opt-suggestions.o.
9791 * gcc-main.c: Include opt-suggestions.h.
9792 * gcc.c (driver::driver): Likewise.
9793 (driver::~driver): Remove m_option_suggestions.
9794 (driver::build_option_suggestions): Moved to option_proposer.
9795 (driver::suggest_option): Likewise.
9796 (driver::handle_unrecognized_options): Use option_proposer.
9797 * gcc.h (class driver): Add new memver m_option_proposer.
9798 * opt-suggestions.c: New file.
9799 * opt-suggestions.h: New file.
9801 2018-06-28 Martin Liska <mliska@suse.cz>
9803 * vec.h (class auto_string_vec): New (moved from auto_argvec).
9804 (auto_string_vec::~auto_string_vec): Likewise.
9806 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9808 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9809 prevent_decl_creation_for_types fields up and add reset_location field.
9810 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9811 statement if id->reset_location is true.
9812 (copy_edges_for_bb): Do not set goto_locus on the new edges if
9813 id->reset_location is true.
9814 (copy_phis_for_bb): Force input_location on the arguments if
9815 id->reset_location is true.
9816 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9817 is set on the function to be inlined.
9819 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
9821 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9823 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
9825 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9826 registers for Pmode.
9827 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9828 hard registers for the clobbered pseudo.
9830 2018-06-27 Paul Koning <ni1d@arrl.net>
9832 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9833 mutually exclusive options.
9834 * config/pdp11/constraints.md (h): New constraint.
9835 (O): Update definition to match shift code generation.
9836 (D): New constraint.
9837 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9839 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9841 (output_jump): Change arguments.
9842 (pdp11_fixed_cc_regs): New function.
9843 (pdp11_cc_mode): Ditto.
9844 (pdp11_expand_shift): Ditto.
9845 (pdp11_assemble_shift): Ditto.
9846 (pdp11_small_shift): Ditto.
9847 (pdp11_branch_cost): Remove.
9848 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9850 (pdp11_register_move_cost): Update for CC registers.
9851 (pdp11_rtx_costs): Add case for LSHIFTRT.
9852 (pdp11_output_jump): Add CCNZ mode conditional branches.
9853 (notice_update_cc_on_set): Remove.
9854 (pdp11_cc_mode): New function.
9855 (simple_memory_operand): Correct pre/post decrement case.
9856 (no_side_effect_operand): New function.
9857 (pdp11_regno_reg_class): Add CC_REGS class.
9858 (pdp11_fixed_cc_regs): New function.
9859 (pdp11_small_shift): New function.
9860 (pdp11_expand_shift): New function to expand shift insns.
9861 (pdp11_assemble_shift): New function to output shifts.
9862 (pdp11_branch_cost): Remove.
9863 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9864 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9865 (WCHAR_TYPE): Ditto.
9866 (PTRDIFF_TYPE): Ditto.
9867 (ADJUST_INSN_LENGTH): New macro.
9868 (FIXED_REGISTERS): Add CC registers.
9869 (CALL_USED_REGISTERS): Ditto.
9871 (REG_CLASS_NAMES): Ditto.
9872 (REG_CLASS_CONTENTS): Ditto.
9873 (SELECT_CC_MODE): Use new function.
9874 (TARGET_FLAGS_REGNUM): New macro.
9875 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9876 (cc0_reg_rtx): Remove.
9877 (CC_STATUS_MDEP): Remove.
9878 (CC_STATUS_MDEFP_INIT): Remove.
9879 (CC_IN_FPU): Remove.
9880 (NOTICE_UPDATE_CC): Remove.
9881 (REGISTER_NAMES): Add CC registers.
9882 (BRANCH_COST): Change to constant 1.
9883 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9885 * config/pdp11/pdp11.opt (mbcopy): Remove.
9886 (mbcopy-builtin): Remove.
9887 (mbranch-cheap): Remove.
9888 (mbranch-expensive): Remove.
9889 * config/pdp11/predicates.md (expand_shift_operand): Update to
9890 match shift code generation.
9891 (ccnz_operator): New predicate.
9892 * doc/invoke.texi (PDP-11 Options): Remove deleted options
9893 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9894 Remove non-existent option -mabshi, -mno-abshi. Document mutually
9896 * doc/md.texi (PDP-11): Document new D and h constraints. Update
9897 description of O constraint.
9899 2018-06-27 Jeff Law <law@redhat.com>
9900 Austin Law <austinklaw@gmail.com>
9902 * config/v850/v850.md (addsi3_set_flags): New pattern.
9903 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9904 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9905 (zero_extendhisi2_v850_set_flags): Likewise.
9906 (zero_extendqisi2_v850_set_flags): Likewise.
9907 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9908 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9909 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9911 * config/v850/v850-protos.h (notice_update_cc): Remove.
9912 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9913 (v850_print_operand): Handle 'D' and "d".
9914 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9915 Add handling of arithmetic/logical operations compared against zero.
9916 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9917 Do not look at v850_compare_op, instead get mode from last argument.
9918 (v850_gen_compare): Remove
9919 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9920 after reload for prologue insns.
9921 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9923 (construct_save_jarl): Likewise.
9924 (TARGET_FLAGS_REGNUM): Define.
9925 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9926 (NOTICE_UPDATE_CC): Remove.
9927 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9928 than cc0. Conditionalize on reload_completed.
9929 (cmpsi_insn, setfcc_insn): Likewise.
9930 (tst1 splitter): Turn into define_and_split which sets the flags
9932 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9933 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9934 (cstoresf4, cstoredf4): Clobber the flags.
9935 (cmpsi, cmpsf, cmpdf): Remove expanders.
9936 (setf_insn): Remove pattern.
9937 (addsi3): Turn into define_and_split which clobbers the flags after
9938 reload and a suitable pattern (addsi3_clobber_flags) for use after
9940 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9941 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9942 (ashrsi3, ashrsi3_v850e2): Likewise.
9943 (bins): Clobber the flags.
9944 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9945 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9946 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9947 (call_value_internal_short, call_value_internal_long): Likewise.
9948 (callt_save_interrupt, callt_return_interrupt): Likewise.
9949 (save_interrupt, return_interrupt): Likewise.
9950 (callt_save_all_interrupt, save_all_interrupt): Likewise.
9951 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9952 (restore_all_interrupt, _restore_all_interrupt): Likewise.
9953 (All FP comparisons): Only allow after reload has completed.
9955 (divh, divhu): Tweak output template.
9956 (branch_z_normal, branch_z_invert): Remove
9957 (branch_nz_normal, branch_nz_invert): Likewise.
9958 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9960 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9961 * config/v850/v850.c (notice_update_cc): Remove.
9962 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
9963 (CC_NO_CARRY): Likewise.
9964 (NOTICE_UPDATE_CC): Define to nothing.
9965 * config/v850/v850.md: Remove block comment on cc0 handling
9966 Remove "cc" attribute from all patterns. Remove cc_status handling
9967 from all patterns. Minor formatting fixes.
9969 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9971 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9972 (cortex-a76.cortex-a55): Likewise.
9973 * config/aarch64/aarch64-tune.md: Regenerate.
9974 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9975 cortex-a76.cortex-a55.
9977 2018-06-27 Jeff Law <law@redhat.com>
9979 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9980 (MULTILIB_DIRNAMES): Similarly.
9982 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
9984 * gimple.h (gimple_return_retbnd): Delete.
9985 (gimple_return_set_retbnd): Likewise.
9986 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9987 gimple_return_set_retbnd.
9988 * gimple-pretty-print.c (dump_gimple_return): Remove call to
9989 gimple_return_retbnd and adjust.
9990 * tree-inline.h (struct copy_body_data): Remove retbnd field.
9991 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9992 Explicitly return NULL in a couple more cases. Move assertion
9993 on debug statements and remove unreachable code.
9994 (reset_debug_binding): Do not test id->retbnd.
9995 (expand_call_inline): Do not set it.
9997 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9999 * configure.ac: Add --disable-gcov option.
10000 * configure: Regenerate.
10001 * Makefile.in: Honour @enable_gcov@.
10002 * doc/install.texi: Document --disable-gcov.
10004 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10006 * config/arm/arm-cpus.in (cortex-a76): New entry.
10007 (cortex-a76.cortex-a55): Likewise.
10008 * config/arm/arm-tables.opt: Regenerate.
10009 * config/arm/arm-tune.md: Likewise.
10010 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10011 * doc/invoke.texi (ARM Options): Document cortex-a76 and
10012 cortex-a76.cortex-a55.
10014 2018-06-27 Tamar Christina <tamar.christina@arm.com>
10017 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10019 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
10021 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10023 (EPILOGUE_USES): Likewise.
10025 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
10027 * tree-inline.c (remap_location): New function extracted from...
10028 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
10029 (copy_phis_for_bb): ...here. Call remap_location.
10030 (copy_cfg_body): Adjust call to copy_edges_for_bb.
10032 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
10034 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10035 unaligned vsx for 16B memset.
10037 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
10040 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10041 ieee128_float_type_node to long_double_type_node unless
10042 TARGET_LONG_DOUBLE_128 is set.
10044 2018-06-26 David Malcolm <dmalcolm@redhat.com>
10046 * cfgloop.c (get_loop_location): Convert return type from
10047 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10048 by implicit construction from rtx_insn *, and using
10049 dump_user_location_t::from_function_decl for the fallback case.
10050 * cfgloop.h (get_loop_location): Convert return type from
10051 location_t to dump_user_location_t.
10052 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10053 dump_printf_loc to pass in a dump_location_t rather than a
10054 location_t, via the gimple stmt.
10055 * coverage.c (get_coverage_counts): Update calls to
10056 dump_printf_loc to pass in dump_location_t rather than a
10058 * doc/optinfo.texi (Dump types): Convert example of
10059 dump_printf_loc from taking "locus" to taking "insn". Update
10060 description of the "_loc" calls to cover dump_location_t.
10061 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10063 (dump_user_location_t::dump_user_location_t): New constructors,
10064 from gimple *stmt and rtx_insn *.
10065 (dump_user_location_t::from_function_decl): New function.
10066 (dump_loc): Make static.
10067 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10068 const dump_location_t &.
10069 (dump_generic_expr_loc): Delete.
10070 (dump_printf_loc): Convert param "loc" from location_t to
10071 const dump_location_t &.
10072 (selftest::test_impl_location): New function.
10073 (selftest::dumpfile_c_tests): New function.
10074 * dumpfile.h: Include "profile-count.h".
10075 (class dump_user_location_t): New class.
10076 (struct dump_impl_location_t): New struct.
10077 (class dump_location_t): New class.
10078 (dump_printf_loc): Convert 2nd param from source_location to
10079 const dump_location_t &.
10080 (dump_generic_expr_loc): Delete.
10081 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10082 const dump_location_t &.
10083 * gimple-fold.c (fold_gimple_assign): Update call to
10084 dump_printf_loc to pass in a dump_location_t rather than a
10085 location_t, via the gimple stmt.
10086 (gimple_fold_call): Likewise.
10087 * gimple-loop-interchange.cc
10088 (loop_cand::analyze_iloop_reduction_var): Update for change to
10089 check_reduction_path.
10090 (tree_loop_interchange::interchange): Update for change to
10091 find_loop_location.
10092 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10093 change in return-type of find_loop_location.
10094 (graphite_regenerate_ast_isl): Likewise.
10095 * graphite-optimize-isl.c (optimize_isl): Likewise.
10096 * graphite.c (graphite_transform_loops): Likewise.
10097 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10098 pass in a dump_location_t rather than a location_t, via the
10100 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10101 * ipa.c (walk_polymorphic_call_targets): Likewise.
10102 * loop-unroll.c (report_unroll): Convert "locus" param from
10103 location_t to dump_location_t.
10104 (decide_unrolling): Update for change to get_loop_location's
10106 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10107 location_t to dump_user_location_t.
10108 (grid_find_single_omp_among_assignments_1): Updates calls to
10109 dump_printf_loc to pass in a dump_location_t rather than a
10110 location_t, via the gimple stmt.
10111 (grid_parallel_clauses_gridifiable): Convert "tloc" from
10112 location_t to dump_location_t. Updates calls to dump_printf_loc
10113 to pass in a dump_location_t rather than a location_t, via the
10115 (grid_inner_loop_gridifiable_p): Likewise.
10116 (grid_dist_follows_simple_pattern): Likewise.
10117 (grid_gfor_follows_tiling_pattern): Likewise.
10118 (grid_target_follows_gridifiable_pattern): Likewise.
10119 (grid_attempt_target_gridification): Convert initialization
10120 of local "grid" from memset to zero-initialization; FIXME: does
10121 this require C++11? Update call to dump_printf_loc to pass in a
10122 optinfo_location rather than a location_t, via the gimple stmt.
10123 * profile.c (read_profile_edge_counts): Updates call to
10124 dump_printf_loc to pass in a dump_location_t rather than a
10126 (compute_branch_probabilities): Likewise.
10127 * selftest-run-tests.c (selftest::run_tests): Call
10129 * selftest.h (dumpfile_c_tests): New decl.
10130 * tree-loop-distribution.c (pass_loop_distribution::execute):
10131 Update for change in return type of find_loop_location.
10132 * tree-parloops.c (parallelize_loops): Likewise.
10133 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10134 "locus" from location_t to dump_user_location_t.
10135 (canonicalize_loop_induction_variables): Likewise.
10136 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10137 for change in return type of find_loop_location.
10138 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10139 to dump_printf_loc to pass in a dump_location_t rather than a
10140 location_t, via the stmt.
10141 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10143 * tree-vect-loop-manip.c (find_loop_location): Convert return
10144 type from source_location to dump_user_location_t.
10145 (vect_do_peeling): Update for above change.
10146 (vect_loop_versioning): Update for change in type of
10148 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10149 from location_t to dump_user_location_t.
10150 (vect_estimate_min_profitable_iters): Update for change in type
10152 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10153 location_t to dump_location_t.
10154 (vect_slp_bb): Update for change in type of vect_location.
10155 * tree-vectorizer.c (vect_location): Convert from source_location
10156 to dump_user_location_t.
10157 (try_vectorize_loop_1): Update for change in vect_location's type.
10158 (vectorize_loops): Likewise.
10159 (increase_alignment): Likewise.
10160 * tree-vectorizer.h (vect_location): Convert from source_location
10161 to dump_user_location_t.
10162 (find_loop_location): Convert return type from source_location to
10163 dump_user_location_t.
10164 (check_reduction_path): Convert 1st param from location_t to
10165 dump_user_location_t.
10166 * value-prof.c (check_counter): Update call to dump_printf_loc to
10167 pass in a dump_user_location_t rather than a location_t; update
10168 call to error_at for change in type of "locus".
10169 (check_ic_target): Update call to dump_printf_loc to
10170 pass in a dump_user_location_t rather than a location_t, via the
10173 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
10175 * config/s390/s390.h (enum processor_flags): Do not use
10176 default tune parameter when -march was specified.
10178 2018-06-26 Jakub Jelinek <jakub@redhat.com>
10181 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10182 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10185 2018-06-26 Richard Biener <rguenther@suse.de>
10187 PR tree-optimization/86287
10189 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10190 (vect_analyze_loop): Initialize n_stmts.
10192 2018-06-26 Richard Biener <rguenther@suse.de>
10194 PR middle-end/86271
10195 * fold-const.c (fold_convertible_p): Pointer extension
10198 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
10201 * dwarf2out.c (loc_list_has_views): Adjust comments.
10202 (dw_loc_list): Split single cross-partition range with
10205 2018-06-25 Jeff Law <law@redhat.com>
10207 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10208 on -mbig-switch by default.
10210 * config/v850/predicates.md (const_float_1_operand): Fix match_code
10212 (const_float_0_operand): Remove unused predicate.
10213 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10214 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10215 (recipsf2): New expander. Original pattern now called
10217 (recipdf2, recipdf2_insn): Similarly.
10218 (rsqrtsf2, rsqrtsf2_insn): Similarly
10219 (rsqrtdf2, rsqrtdf2_insn): Similarly
10221 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
10223 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10224 Simplify logic for FreeBSD (twice).
10226 2018-06-25 Martin Sebor <msebor@redhat.com>
10228 PR tree-optimization/86204
10229 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10230 a strnlen result if it's less than the length of the string.
10232 2018-06-25 Martin Sebor <msebor@redhat.com>
10234 PR tree-optimization/85700
10235 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10236 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10237 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10239 2018-06-25 Martin Sebor <msebor@redhat.com>
10241 * doc/extend.texi (Zero-length arrays): Update and clarify.
10243 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
10245 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10246 added IEEE/IBM long double multilib support on PowerPC little
10247 endian Linux systems.
10248 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10249 (MULTILIB_DEFAULTS): Likewise.
10250 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10252 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10253 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10254 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10256 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
10258 PR middle-end/86311
10259 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10260 (REORDER_45): Likewise.
10262 2018-06-25 Jeff Law <law@redhat.com>
10264 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10265 dividend to 32 bits. Adjust length.
10266 (udivmodhi4): Cleanup output template. Fix length.
10268 2018-06-25 Carl Love <cel@us.ibm.com>
10270 * config/rs6000/vsx.md: Change word selector to prefered location.
10272 2018-06-25 Richard Biener <rguenther@suse.de>
10274 PR tree-optimization/86304
10275 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10276 epilogue-if-converted loops as well.
10278 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
10280 * lto-section-out.c (lto_begin_section): Do not print section
10281 name for noaddr and unnumbered dumps.
10283 2018-06-25 Richard Biener <rguenther@suse.de>
10285 * tree-vectorizer.h (struct vec_info_shared): New structure
10286 with parts split out from struct vec_info and loop_nest from
10287 struct _loop_vec_info.
10288 (struct vec_info): Adjust accordingly.
10289 (struct _loop_vec_info): Likewise.
10290 (LOOP_VINFO_LOOP_NEST): Adjust.
10291 (LOOP_VINFO_DATAREFS): Likewise.
10292 (LOOP_VINFO_DDRS): Likewise.
10293 (struct _bb_vec_info): Likewise.
10294 (BB_VINFO_DATAREFS): Likewise.
10295 (BB_VINFO_DDRS): Likewise.
10296 (struct _stmt_vec_info): Add dr_aux member.
10297 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10298 (DR_MISALIGNMENT_UNINITIALIZED): New.
10299 (set_dr_misalignment): Adjust.
10300 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10301 (vect_analyze_loop): Adjust prototype.
10302 (vect_analyze_loop_form): Likewise.
10303 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10304 Compute dependences lazily.
10305 (vect_record_base_alignments): Use shared datarefs/ddrs.
10306 (vect_verify_datarefs_alignment): Likewise.
10307 (vect_analyze_data_refs_alignment): Likewise.
10308 (vect_analyze_data_ref_accesses): Likewise.
10309 (vect_analyze_data_refs): Likewise.
10310 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10311 constructor parameter for shared part.
10312 (vect_analyze_loop_form): Pass in shared part and adjust.
10313 (vect_analyze_loop_2): Pass in storage for the number of
10314 stmts. Move loop nest finding to the caller. Compute
10316 (vect_analyze_loop): Pass in shared part.
10317 (vect_transform_loop): Verify shared datarefs are unchanged.
10318 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10319 constructor parameter for shared part.
10320 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10321 (vect_slp_bb): Verify shared datarefs are unchanged before
10323 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10325 (new_stmt_vec_info): Initialize DR_AUX misalignment to
10326 DR_MISALIGNMENT_UNINITIALIZED.
10327 * tree-vectorizer.c (vec_info::vec_info): Add constructor
10328 parameter for shared part.
10329 (vec_info::~vec_info): Adjust.
10330 (vec_info_shared::vec_info_shared): New.
10331 (vec_info_shared::~vec_info_shared): Likewise.
10332 (vec_info_shared::save_datarefs): Likewise.
10333 (vec_info_shared::check_datarefs): Likewise.
10334 (try_vectorize_loop_1): Construct shared part live for analyses
10335 of a single loop for multiple vector sizes.
10336 * tree-parloops.c (gather_scalar_reductions): Adjust.
10338 2018-06-25 Richard Biener <rguenther@suse.de>
10340 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10341 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10342 (vect_analyze_data_refs): Remove similar code from here and
10343 simplify accordingly.
10345 2018-06-25 Richard Biener <rguenther@suse.de>
10347 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10348 for reverse storage order accesses rather than asserting
10349 they cannot happen here.
10351 2018-06-25 Tom de Vries <tdevries@suse.de>
10354 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10355 Use data16 instead of .byte for insn prefix.
10357 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
10360 * parser.c (make_char_string_pack): Pass this literal chars
10361 through cpp_interpret_string.
10362 (cp_parser_userdef_numeric_literal): Check the result of
10363 make_char_string_pack.
10365 2018-06-24 Maya Rashish <coypu@sdf.org>
10367 * ginclude/stddef.h: Simplify conditions around avoiding
10368 re-definition of __size_t.
10370 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10372 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10373 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10375 2018-06-22 Maya Rashish <coypu@sdf.org>
10377 * doc/invoke.texi (mno-fancy-math-387): Update for changes
10378 made to OpenBSD and NetBSD through the years.
10380 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10382 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10383 behavior of vec_pack (vector double, vector double) to match
10384 behavior of vec_float2 (vector double, vector double).
10386 2018-06-22 Olivier Hainque <hainque@adacore.com>
10388 * gimplify.c (gimplify_function_tree): Prevent creation
10389 of a trampoline for the address of the current function
10390 passed to entry/exit instrumentation hooks.
10392 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
10395 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10398 2018-06-22 Martin Liska <mliska@suse.cz>
10400 PR tree-optimization/86263
10401 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10402 Bail out if is_enabled is false.
10403 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10405 (jump_table_cluster::is_enabled): New function.
10407 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10409 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10410 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10411 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10412 (lto_input_ts_binfo_tree_pointers): Likewise.
10413 * tree-streamer-out.c (streamer_write_tree_bitfields,
10414 write_ts_binfo_tree_pointers): Likewise.
10415 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10417 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10419 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10421 2018-06-22 Martin Liska <mliska@suse.cz>
10423 * symbol-summary.h (get): Make it pure and inline move
10424 functionality from ::get function.
10425 (get): Remove and inline into ::get and ::get_create.
10426 (get_create): Move code from ::get function.
10428 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10431 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10432 -x assembler-with-cpp.
10434 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10436 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10437 _FILE_OFFSET_BITS=64 for C++.
10439 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
10441 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10442 conversion insn that shows up when pr85657-3.c is compiled using
10443 IEEE 128-bit long double.
10444 (neg<mode>2_internal): Use the correct mode to check whether the
10445 mode is IBM extended.
10446 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10447 multiply and divide external functions from being created more
10450 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10452 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10454 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
10455 the edge can be forwarded.
10456 (cfg_layout_merge_blocks): Likewise.
10458 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10460 * except.c (finish_eh_generation): Commit edge insertions only after
10461 the EH edges have been redirected from post-landing to landing pads.
10463 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10465 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10466 create_tmp_var_for to create the FRAME decl.
10467 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10469 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10471 * tree-inline.c (copy_edges_for_bb): Minor tweak.
10472 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10473 debug statement when resetting its value.
10474 (expand_call_inline): Copy the locus of the call onto the assignment
10475 of the return value, if any. Use local variable in more cases.
10477 2018-06-21 Martin Liska <mliska@suse.cz>
10479 * ipa-pure-const.c (propagate_nothrow): Use
10480 funct_state_summaries->get.
10481 (dump_malloc_lattice): Likewise.
10482 (propagate_malloc): Likewise.
10484 2018-06-21 Richard Biener <rguenther@suse.de>
10486 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10487 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10488 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10490 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10491 BLOCK_ABSTRACT_ORIGIN unconditionally.
10493 2018-06-21 David Malcolm <dmalcolm@redhat.com>
10495 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10497 * ipa-reference.c (ipa_reference_c_finalize): Delete
10498 ipa_ref_opt_sum_summaries and set it to NULL.
10500 2018-06-21 Tom de Vries <tdevries@suse.de>
10502 PR tree-optimization/85859
10503 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10504 test with comment from bb_no_side_effects_p.
10506 2018-06-21 Richard Biener <rguenther@suse.de>
10508 PR tree-optimization/86232
10509 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10510 max for constant niter.
10512 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10514 * config/aarch64/aarch64-simd.md
10515 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10517 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10519 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10520 Make opernads of the unspec commutative.
10522 2018-06-21 Richard Biener <rguenther@suse.de>
10524 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10525 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10526 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10527 (vect_analyze_data_ref_dependence): Re-order checks to deal with
10529 (vect_record_base_alignments): Do not record base alignment
10530 for gathers or scatters.
10531 (vect_compute_data_ref_alignment): Drop return value that is always
10532 true. Bail out early for gathers or scatters.
10533 (vect_enhance_data_refs_alignment): Bail out early for gathers
10535 (vect_find_same_alignment_drs): Likewise.
10536 (vect_analyze_data_refs_alignment): Remove dead code.
10537 (vect_slp_analyze_and_verify_node_alignment): Likewise.
10538 (vect_analyze_data_refs): For possible gathers or scatters do
10539 not create an alternate DR, just check their possible validity
10540 and mark them. Adjust DECL_NONALIASED handling to not rely
10541 on DR_BASE_ADDRESS.
10542 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10543 update inits of gathers or scatters.
10544 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10545 Also copy gather/scatter flag to pattern vinfo.
10547 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
10549 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10550 behavior of vec_packsu (vector unsigned long long, vector unsigned
10551 long long) to match behavior of vec_packs with same signature.
10553 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
10554 Thomas Schwinge <thomas@codesourcery.com>
10555 Cesar Philippidis <cesar@codesourcery.com>
10557 * gimplify.c (gimplify_scan_omp_clauses): Add support for
10558 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10559 (gimplify_adjust_omp_clauses): Likewise.
10560 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10561 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10562 (gimplify_omp_target_update): Update handling of acc update and
10564 * omp-low.c (install_var_field): Remove unused parameter
10565 base_pointers_restrict.
10566 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10567 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10569 (omp_target_base_pointers_restrict_p): Delete.
10570 (scan_omp_target): Update call to scan_sharing_clauses.
10571 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10573 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10574 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10575 (convert_local_omp_clauses): Likewise.
10576 * tree-pretty-print.c (dump_omp_clause): Likewise.
10577 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
10579 (omp_clause_code_name): Likewise.
10581 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10584 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10587 PR tree-optimization/86231
10588 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
10589 anti-range don't overwrite *vr0min before using it to compute *vr0max.
10591 2018-06-20 Tom de Vries <tdevries@suse.de>
10593 PR tree-optimization/86097
10594 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10595 iv type if signedness of iv type is not the same as that of *nit.
10597 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10599 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
10600 EDGE_EH edges, verify they are all EDGE_EH.
10602 2018-06-20 Maya Rashish <coypu@sdf.org>
10604 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10606 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10608 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10609 * config/aarch64/aarch64.c (xgene1_tunings): Add
10610 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10611 (aarch64_mode_valid_for_sched_fusion_p):
10612 Allow 16-byte modes.
10613 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10614 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10616 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10618 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10619 * config/aarch64/iterators.md (VQ2): New mode iterator.
10621 2018-06-20 Martin Liska <mliska@suse.cz>
10623 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10624 Change default ratio from 10 to 8.
10626 2018-06-20 Martin Liska <mliska@suse.cz>
10628 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10630 (bit_test_cluster::find_bit_tests): Likewise.
10631 (switch_decision_tree::analyze_switch_statement): Find clusters.
10632 * tree-switch-conversion.h (struct jump_table_cluster): Document
10635 2018-06-20 Martin Liska <mliska@suse.cz>
10637 * tree-switch-conversion.c (switch_conversion::collect):
10638 Record m_uniq property.
10639 (switch_conversion::expand): Bail out for special conditions.
10640 (group_cluster::~group_cluster): New.
10641 (group_cluster::group_cluster): Likewise.
10642 (group_cluster::dump): Likewise.
10643 (jump_table_cluster::emit): New.
10644 (switch_decision_tree::fix_phi_operands_for_edges): New.
10645 (struct case_node): Remove struct.
10646 (jump_table_cluster::can_be_handled): New.
10647 (case_values_threshold): Moved to header.
10648 (reset_out_edges_aux): Likewise.
10649 (jump_table_cluster::is_beneficial): New.
10650 (bit_test_cluster::can_be_handled): Likewise.
10651 (add_case_node): Remove.
10652 (bit_test_cluster::is_beneficial): New.
10653 (case_bit_test::cmp): New.
10654 (bit_test_cluster::emit): New.
10655 (expand_switch_as_decision_tree_p): Remove.
10656 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10657 (fix_phi_operands_for_edge): Likewise.
10658 (switch_decision_tree::analyze_switch_statement): New.
10659 (compute_cases_per_edge): Move ...
10660 (switch_decision_tree::compute_cases_per_edge): ... here.
10661 (try_switch_expansion): Likewise.
10662 (switch_decision_tree::try_switch_expansion): Likewise.
10663 (record_phi_operand_mapping): Likewise.
10664 (switch_decision_tree::record_phi_operand_mapping): Likewise.
10665 (emit_case_decision_tree): Likewise.
10666 (switch_decision_tree::emit): Likewise.
10667 (balance_case_nodes): Likewise.
10668 (switch_decision_tree::balance_case_nodes): Likewise.
10669 (dump_case_nodes): Likewise.
10670 (switch_decision_tree::dump_case_nodes): Likewise.
10671 (emit_jump): Likewise.
10672 (switch_decision_tree::emit_jump): Likewise.
10673 (emit_cmp_and_jump_insns): Likewise.
10674 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10675 (emit_case_nodes): Likewise.
10676 (switch_decision_tree::emit_case_nodes): Likewise.
10677 (conditional_probability): Remove.
10678 * tree-switch-conversion.h (enum cluster_type): New.
10680 (struct cluster): Likewise.
10681 (cluster::cluster): Likewise.
10682 (struct simple_cluster): Likewise.
10683 (simple_cluster::simple_cluster): Likewise.
10684 (struct group_cluster): Likewise.
10685 (struct jump_table_cluster): Likewise.
10686 (struct bit_test_cluster): Likewise.
10687 (struct min_cluster_item): Likewise.
10688 (struct case_tree_node): Likewise.
10689 (case_tree_node::case_tree_node): Likewise.
10690 (jump_table_cluster::case_values_threshold): Likewise.
10691 (struct case_bit_test): Likewise.
10692 (struct switch_decision_tree): Likewise.
10693 (struct switch_conversion): Likewise.
10694 (switch_decision_tree::reset_out_edges_aux): Likewise.
10696 2018-06-20 Martin Liska <mliska@suse.cz>
10698 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10699 (hoist_edge_and_branch_if_true): Likewise.
10700 (expand_switch_using_bit_tests_p): Likewise.
10701 (struct case_bit_test): Likewise.
10702 (case_bit_test_cmp): Likewise.
10703 (emit_case_bit_tests): Likewise.
10704 (switch_conversion::switch_conversion): New class.
10705 (struct switch_conv_info): Remove old struct.
10706 (collect_switch_conv_info): More to ...
10707 (switch_conversion::collect): ... this.
10708 (check_range): Likewise.
10709 (switch_conversion::check_range): Likewise.
10710 (check_all_empty_except_final): Likewise.
10711 (switch_conversion::check_all_empty_except_final): Likewise.
10712 (check_final_bb): Likewise.
10713 (switch_conversion::check_final_bb): Likewise.
10714 (create_temp_arrays): Likewise.
10715 (switch_conversion::create_temp_arrays): Likewise.
10716 (free_temp_arrays): Likewise.
10717 (gather_default_values): Likewise.
10718 (switch_conversion::gather_default_values): Likewise.
10719 (build_constructors): Likewise.
10720 (switch_conversion::build_constructors): Likewise.
10721 (constructor_contains_same_values_p): Likewise.
10722 (switch_conversion::contains_same_values_p): Likewise.
10723 (array_value_type): Likewise.
10724 (switch_conversion::array_value_type): Likewise.
10725 (build_one_array): Likewise.
10726 (switch_conversion::build_one_array): Likewise.
10727 (build_arrays): Likewise.
10728 (switch_conversion::build_arrays): Likewise.
10729 (gen_def_assigns): Likewise.
10730 (switch_conversion::gen_def_assigns): Likewise.
10731 (prune_bbs): Likewise.
10732 (switch_conversion::prune_bbs): Likewise.
10733 (fix_phi_nodes): Likewise.
10734 (switch_conversion::fix_phi_nodes): Likewise.
10735 (gen_inbound_check): Likewise.
10736 (switch_conversion::gen_inbound_check): Likewise.
10737 (process_switch): Use the newly created class.
10738 (switch_conversion::expand): New.
10739 (switch_conversion::~switch_conversion): New.
10740 * tree-switch-conversion.h: New file.
10742 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10744 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10745 tree-vect-patterns.c.
10746 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10747 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
10748 (vect_recog_sad_pattern): Likewise.
10749 (vect_recog_widen_sum_pattern): Likewise.
10750 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
10751 (vect_recog_widen_shift_pattern): Remove the type_in argument.
10752 (vect_recog_rotate_pattern): Likewise.
10753 (vect_recog_mult_pattern): Likewise.
10754 (vect_recog_vector_vector_shift_pattern): Likewise.
10755 (vect_recog_divmod_pattern): Likewise.
10756 (vect_recog_mixed_size_cond_pattern): Likewise.
10757 (vect_recog_bool_pattern): Likewise.
10758 (vect_recog_mask_conversion_pattern): Likewise.
10759 (vect_try_gather_scatter_pattern): Likewise.
10760 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
10761 (vect_recog_over_widening_pattern): Likewise.
10762 (vect_recog_gather_scatter_pattern): Likewise.
10763 (vect_recog_func_ptr): Move from tree-vectorizer.h
10764 (vect_vect_recog_func_ptrs): Move further down the file.
10765 (vect_recog_func): Likewise. Remove the third argument.
10766 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10767 (vect_pattern_recog_1): Expect the pattern function to do any
10768 necessary target tests. Also expect it to provide a vector type.
10769 Remove the type_in handling.
10771 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10773 * tree-vect-patterns.c (vect_pattern_detected): New function.
10774 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10775 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10776 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10777 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10778 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10779 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10780 (vect_recog_mask_conversion_pattern)
10781 (vect_try_gather_scatter_pattern): Likewise.
10783 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10785 * tree-vect-patterns.c (vect_get_internal_def): New function.
10786 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10787 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10788 (search_type_for_mask_1): Use it.
10790 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10792 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10793 redundant WIDEN_SUM_EXPR handling.
10794 (vect_recog_sad_pattern): Likewise.
10796 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10798 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10799 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10800 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10801 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10802 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10804 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10806 * tree-vect-stmts.c (vectorizable_call): Make sure that we
10807 use the stmt_vec_info of the original bb statement for the
10808 new zero assignment, even if the call is part of a pattern.
10810 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10812 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10813 that the sequence is attached to the original statement rather
10814 than the pattern statement.
10815 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10816 PATTERN_DEF_SEQ from the original statement rather than
10817 the main pattern statement.
10818 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10819 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10820 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10822 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10824 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10825 definition statements before the early exit for statements that aren't
10827 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10829 (vect_transform_loop): ...here. Process pattern definition
10830 statements without first checking whether the main pattern
10831 statement is live or relevant.
10833 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10835 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10836 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10838 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
10840 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10841 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10842 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10843 (expand_block_compare): Change select_block_compare_mode call.
10844 (expand_strncmp_align_check): Use new functions, fix comment.
10845 (emit_final_str_compare_gpr): New function.
10846 (expand_strn_compare): Refactor and clean up code.
10847 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10849 2018-06-19 Tony Reix <tony.reix@atos.com>
10850 Damien Bergamini <damien.bergamini@atos.com>
10851 David Edelsohn <dje.gcc@gmail.com>
10853 * collect2.c (static_obj): New variable.
10854 (static_libs): New variable.
10855 (is_in_list): Uncomment declaration.
10856 (main): Track AIX libraries linked statically.
10857 (is_in_list): Uncomment definition.
10858 (scan_prog_file): Don't add AIX shared libraries initializer
10859 to constructor list if linking statically.
10861 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
10863 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10865 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10867 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
10869 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10872 2018-06-19 Martin Liska <mliska@suse.cz>
10874 * config/i386/i386.c (ix86_can_inline_p): Do not use
10875 ipa_fn_summaries::get_create.
10876 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10878 (devirtualization_time_bonus): Likewise.
10879 (ipcp_propagate_stage): Likewise.
10880 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10881 (edge_set_predicate): Likewise.
10882 (evaluate_conditions_for_known_args): Likewise.
10883 (evaluate_properties_for_edge): Likewise.
10884 (ipa_call_summary::reset): Tranform to ...
10885 (ipa_call_summary::~ipa_call_summary): ... this.
10886 (ipa_fn_summary::reset): Transform to ...
10887 (ipa_fn_summary::~ipa_fn_summary): ... this.
10888 (ipa_fn_summary_t::remove): Rename to ...
10889 (ipa_fn_summary_t::remove_callees): ... this.
10890 (ipa_fn_summary_t::duplicate): Use placement new
10891 instead of memory copy.
10892 (ipa_call_summary_t::duplicate): Likewise.
10893 (ipa_call_summary_t::remove): Remove.
10894 (dump_ipa_call_summary): Change get_create to get.
10895 (ipa_dump_fn_summary): Dump only when summary exists.
10896 (analyze_function_body): Use symbol_summary::get instead
10898 (compute_fn_summary): Likewise.
10899 (estimate_edge_devirt_benefit): Likewise.
10900 (estimate_edge_size_and_time): Likewise.
10901 (inline_update_callee_summaries): Likewise.
10902 (remap_edge_change_prob): Likewise.
10903 (remap_edge_summaries): Likewise.
10904 (ipa_merge_fn_summary_after_inlining): Likewise.
10905 (write_ipa_call_summary): Likewise.
10906 (ipa_fn_summary_write): Likewise.
10907 (ipa_free_fn_summary): Likewise.
10908 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10909 (struct ipa_call_summary): Likewise.
10910 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10912 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10913 (estimate_size_after_inlining): Likewise.
10914 (estimate_growth): Likewise.
10915 (growth_likely_positive): Likewise.
10916 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10917 (inline_call): Likewise.
10918 * ipa-inline.c (caller_growth_limits): Likewise.
10919 (can_inline_edge_p): Likewise.
10920 (can_inline_edge_by_limits_p): Likewise.
10921 (compute_uninlined_call_time): Likewise.
10922 (compute_inlined_call_time): Likewise.
10923 (want_inline_small_function_p): Likewise.
10924 (edge_badness): Likewise.
10925 (update_caller_keys): Likewise.
10926 (update_callee_keys): Likewise.
10927 (inline_small_functions): Likewise.
10928 (inline_to_all_callers_1): Likewise.
10929 (dump_overall_stats): Likewise.
10930 (early_inline_small_functions): Likewise.
10931 (early_inliner): Likewise.
10932 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10933 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10934 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10935 * ipa-split.c (execute_split_functions): Likewise.
10936 * symbol-summary.h: Likewise.
10937 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10939 2018-06-19 Richard Biener <rguenther@suse.de>
10941 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10942 (vectorize_loops): ... here. Fix dbgcnt handling.
10943 (try_vectorize_loop): Wrap try_vectorize_loop_1.
10945 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
10948 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10949 ieee128 argument takes up only one (vector) register, not two (floating
10952 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10954 * gimplify.c (gimplify_init_constructor): Really never clear for an
10955 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10957 2018-06-19 Richard Biener <rguenther@suse.de>
10959 PR tree-optimization/86179
10960 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10961 after failed recognition.
10963 2018-06-18 Martin Sebor <msebor@redhat.com>
10965 PR middle-end/85602
10966 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10967 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10968 Handle integer subtraction.
10969 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10970 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10972 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10974 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10975 param from rtx to rtx_insn *.
10976 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10978 (frv_ifcvt_modify_insn): Likwise.
10979 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10980 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
10981 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10982 as_a <rtx_insn *> cast to local "unprotected_region" once
10983 it's been established that it's not NULL or pc_rtx.
10984 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10985 param "sethi" from rtx to rtx_insn *.
10986 (nds32_group_float_insns): Likewise for param "insn".
10987 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10989 (vax_output_int_subtract): Likewise.
10990 * config/vax/vax.c (vax_output_int_add): Likewise for param
10992 (vax_output_int_subtract): Likewise.
10993 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10994 (emit_pattern_after): Likewise for param "after".
10995 (emit_insn_after): Likewise.
10996 (emit_jump_insn_after): Likewise.
10997 (emit_call_insn_after): Likewise.
10998 (emit_debug_insn_after): Likewise.
10999 (emit_pattern_before): Likewise for param "before".
11000 (emit_insn_before): Likewise.
11001 (emit_jump_insn_before): Likewise.
11002 * final.c (get_insn_template): Likewise for param "insn", removing
11004 * output.h (get_insn_template): Likewise for 2nd param.
11005 * rtl.h (emit_insn_before): Likewise.
11006 (emit_jump_insn_before): Likewise.
11007 (emit_debug_insn_before_noloc): Likewise.
11008 (emit_insn_after): Likewise.
11009 (emit_jump_insn_after): Likewise.
11010 (emit_call_insn_after): Likewise.
11011 (emit_debug_insn_after): Likewise.
11012 (set_insn_deleted): Likewise for param.
11014 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
11017 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11018 floating point modes, so that IFmode is numerically greater than
11019 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11020 to declare the ordering. This prevents IFmode from being
11021 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11022 machine. Include rs6000-modes.h to share the fractional values
11023 between genmodes* and the rest of the compiler.
11024 (IFmode): Likewise.
11025 (KFmode): Likewise.
11026 (TFmode): Likewise.
11027 * config/rs6000/rs6000-modes.h: New file.
11028 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11029 meaning of rs6000_long_double_size so that 126..128 selects an
11030 appropriate 128-bit floating point type.
11031 (rs6000_option_override_internal): Likewise.
11032 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11033 (TARGET_LONG_DOUBLE_128): Change the meaning of
11034 rs6000_long_double_size so that 126..128 selects an appropriate
11035 128-bit floating point type.
11036 (LONG_DOUBLE_TYPE_SIZE): Update comment.
11037 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11038 source and destination to match the standard usage.
11039 (truncifkf2): Likewise.
11040 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11041 ISA 2.07 to use an explicit clobber, instead of passing in a
11043 (copysign<mode>3_soft): Likewise.
11045 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11047 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11048 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11049 (vect_slp_analyze_instance_dependence): Likewise.
11050 (vect_enhance_data_refs_alignment): Likewise.
11051 (vect_analyze_data_refs_alignment): Likewise.
11052 (vect_slp_analyze_and_verify_instance_alignment
11053 (vect_analyze_data_ref_accesses): Likewise.
11054 (vect_prune_runtime_alias_test_list): Likewise.
11055 (vect_analyze_data_refs): Likewise.
11056 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11057 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11058 (vect_analyze_scalar_cycles_1): Likewise.
11059 (vect_get_loop_niters): Likewise.
11060 (vect_analyze_loop_form_1): Likewise.
11061 (vect_update_vf_for_slp): Likewise.
11062 (vect_analyze_loop_operations): Likewise.
11063 (vect_analyze_loop): Likewise.
11064 (vectorizable_induction): Likewise.
11065 (vect_transform_loop): Likewise.
11066 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11067 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11068 (vect_make_slp_decision): Likewise.
11069 (vect_detect_hybrid_slp): Likewise.
11070 (vect_slp_analyze_operations): Likewise.
11071 (vect_slp_bb): Likewise.
11072 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11073 (vectorizable_bswap): Likewise.
11074 (vectorizable_call): Likewise.
11075 (vectorizable_simd_clone_call): Likewise.
11076 (vectorizable_conversion): Likewise.
11077 (vectorizable_assignment): Likewise.
11078 (vectorizable_shift): Likewise.
11079 (vectorizable_operation): Likewise.
11080 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11082 2018-06-18 Martin Sebor <msebor@redhat.com>
11084 PR tree-optimization/81384
11085 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11086 * builtins.c (expand_builtin_strnlen): New function.
11087 (expand_builtin): Call it.
11088 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11089 * builtins.def (BUILT_IN_STRNLEN): New.
11090 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11091 Warn for bounds in excess of maximum object size.
11092 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11093 single-value ranges. Handle strnlen.
11094 (handle_builtin_strlen): Handle strnlen.
11095 (strlen_check_and_optimize_stmt): Same.
11096 * doc/extend.texi (Other Builtins): Document strnlen.
11098 2018-06-18 Maya Rashish <coypu@sdf.org>
11100 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11101 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11102 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11104 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11106 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11108 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11110 * tree.c (escaped_string::escape): Replace cast to char * by
11111 const_cast<char *> (unescaped).
11113 2018-06-18 Nick Clifton <nickc@redhat.com>
11116 * tree.c (escaped_string): New class. Converts an unescaped
11117 string into its escaped equivalent.
11118 (warn_deprecated_use): Use the new class to convert the
11119 deprecation message, if present.
11120 (test_escaped_strings): New self test.
11121 (test_c_tests): Add test_escaped_strings.
11122 * doc/extend.texi (deprecated): Add a note that the
11123 deprecation message is affected by the -fmessage-length
11124 option, and that control characters will be escaped.
11125 (#pragma GCC error): Document this pragma.
11126 (#pragma GCC warning): Likewise.
11127 * doc/invoke.texi (-fmessage-length): Document this option's
11128 effect on the #warning and #error preprocessor directives and
11129 the deprecated attribute.
11131 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
11133 * tree.c (decl_value_expr_lookup): Revert latest change.
11134 (decl_value_expr_insert): Likewise.
11136 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
11138 * gimplify.c (nonlocal_vlas): Delete.
11139 (nonlocal_vla_vars): Likewise.
11140 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11142 (gimplify_body): Do not create and destroy nonlocal_vlas.
11143 * tree-nested.c: Include diagnostic.h.
11144 (use_pointer_in_frame): Tweak.
11145 (lookup_field_for_decl): Add assertion and declare the transformation.
11146 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11147 internal error when the reference is in a wrong context. Do not
11148 create a debug decl by default.
11149 (note_nonlocal_block_vlas): Delete.
11150 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11151 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
11152 create a debug decl by default.
11153 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11154 call is in a wrong context.
11155 (fixup_vla_decls): New function.
11156 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
11157 debug variables were created.
11158 * tree.c (decl_value_expr_lookup): Add checking assertion.
11159 (decl_value_expr_insert): Likewise.
11161 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11163 PR middle-end/82479
11164 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11165 * tree-scalar-evolution.c (interpret_expr): Likewise.
11166 (expression_expensive_p): Likewise.
11167 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11168 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11169 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11170 (ssa_defined_by_minus_one_stmt_p): New.
11172 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11174 PR middle-end/64946
11175 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11176 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11177 * dojump.c (do_jump): Likewise.
11178 * expr.c (expand_expr_real_2): Check operand type's sign.
11179 * fold-const.c (const_unop): Handle ABSU_EXPR.
11180 (fold_abs_const): Likewise.
11181 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11182 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11183 (strip_sign_op_1): Likesise.
11184 * match.pd: Add new pattern to generate ABSU_EXPR.
11185 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11186 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11187 * tree-eh.c (operation_could_trap_helper_p): Likewise.
11188 * tree-inline.c (estimate_operator_cost): Likewise.
11189 * tree-pretty-print.c (dump_generic_node): Likewise.
11190 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11191 * tree.def (ABSU_EXPR): New.
11193 2018-06-16 Jakub Jelinek <jakub@redhat.com>
11195 PR middle-end/86095
11196 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11197 documented as preserved for backward compatibility only.
11198 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11200 PR rtl-optimization/86108
11201 * bb-reorder.c (create_forwarder_block): Renamed to ...
11202 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
11203 jump from new landing pad to the second part.
11204 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11207 2018-06-15 Jakub Jelinek <jakub@redhat.com>
11209 PR middle-end/85878
11210 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11211 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11212 Only call store_expr for halves if the mode is the same.
11214 PR middle-end/86123
11215 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11216 Fix up comment formatting.
11218 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
11220 * typed-splay-tree.h (typed_splay_tree::remove): New function.
11221 (typed_splay_tree::closure,
11222 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11223 (typed_splay_tree::typed_splay_tree,
11224 typed_splay_tree::operator =): Declared private.
11225 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11226 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11227 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11228 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11229 typed_splay_tree::splay_tree_splay,
11230 typed_splay_tree::splay_tree_foreach_helper,
11231 typed_splay_tree::splay_tree_insert,
11232 typed_splay_tree::splay_tree_remove,
11233 typed_splay_tree::splay_tree_lookup,
11234 typed_splay_tree::splay_tree_predecessor,
11235 typed_splay_tree::splay_tree_successor,
11236 typed_splay_tree::splay_tree_min,
11237 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11238 (typed_splay_tree::root, typed_splay_tree::comp,
11239 typed_splay_tree::delete_key,
11240 typed_splay_tree::delete_value): New data members.
11241 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11242 typed_splay_tree::remove.
11244 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
11246 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11247 -mginv and -mno-ginv to the assembler.
11248 * config/mips/mips.opt (-mcrc): New option.
11249 (-mginv): Likewise.
11250 * doc/invoke.text (-mcrc): Document.
11251 (-mginv): Likewise.
11253 2018-06-15 Nick Clifton <nickc@redhat.com>
11256 * tree.c (escaped_string): New class. Converts an unescaped
11257 string into its escaped equivalent.
11258 (warn_deprecated_use): Use the new class to convert the
11259 deprecation message, if present.
11260 (test_escaped_strings): New self test.
11261 (test_c_tests): Add test_escaped_strings.
11262 * doc/extend.texi (deprecated): Add a note that the
11263 deprecation message is affected by the -fmessage-length
11264 option, and that control characters will be escaped.
11265 (#pragma GCC error): Document this pragma.
11266 (#pragma GCC warning): Likewise.
11267 * doc/invoke.texi (-fmessage-length): Document this option's
11268 effect on the #warning and #error preprocessor directives and
11269 the deprecated attribute.
11271 2018-06-15 Richard Biener <rguenther@suse.de>
11273 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11274 here, also noting vector size used.
11275 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
11276 size used in MSG_OPTIMIZED_LOCATIONS dump.
11277 (pass_slp_vectorize::execute): Adjust.
11279 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
11282 * config/arc/arc.c (arc_return_address_register): Fix
11285 2018-06-15 Richard Biener <rguenther@suse.de>
11287 PR middle-end/86159
11288 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11289 leave useless conversion stripping to force_gimple_operand_gsi.
11290 (gimplify_build2): Likewise.
11291 (gimplify_build1): Likewise.
11293 2018-06-15 Richard Biener <rguenther@suse.de>
11295 PR middle-end/86076
11296 * tree-cfg.c (move_stmt_op): unshare invariant addresses
11297 before adjusting their block.
11299 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
11301 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11302 multilibs for *-*-rtems*.
11303 * config/riscv/t-rtems: New file.
11305 2018-06-14 Jakub Jelinek <jakub@redhat.com>
11307 PR middle-end/86122
11308 * match.pd ((A +- CST1) +- CST2): Punt if last resort
11309 unsigned_type_for returns NULL.
11312 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11313 subregs of multi-word pseudos unless the float mode has word size.
11315 2018-06-14 Richard Biener <rguenther@suse.de>
11317 PR middle-end/86139
11318 * tree-vect-generic.c (build_word_mode_vector_type): Remove
11319 duplicate and harmful type_hash_canon.
11320 * tree.c (type_hash_canon): Assert we didn't find ourselves.
11322 2018-06-14 Richard Biener <rguenther@suse.de>
11325 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11328 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
11330 * config/rtems.h (STDINT_LONG32): Define.
11332 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
11333 Prachi Godbole <prachi.godbole@imgtec.com>
11335 * config/mips/mips-cpus.def: Define P6600.
11336 * config/mips/mips-tables.opt: Regenerate.
11337 * config/mips/mips.c (mips_ucbranch_type): New enum.
11338 (mips_rtx_cost_data): Add support for P6600.
11339 (mips_issue_rate): Likewise.
11340 (mips_multipass_dfa_lookahead): Likewise.
11341 (mips_avoid_hazard): Likewise.
11342 (mips_reorg_process_insns): Likewise.
11343 (mips_classify_branch_p6600): New function.
11344 * config/mips/mips.h (TUNE_P6600): New define.
11345 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11346 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11347 * config/mips/mips.md: Include p6600.md.
11348 (processor): Add p6600.
11349 * config/mips/p6600.md: New file.
11350 * doc/invoke.texi: Add p6600 to supported architectures.
11352 2018-06-13 Martin Sebor <msebor@redhat.com>
11354 PR tree-optimization/86114
11355 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11357 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11359 2018-06-13 Richard Biener <rguenther@suse.de>
11361 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11362 Properly set vector type of the intermediate stmt.
11363 * tree-vect-stmts.c (vectorizable_operation): The destination
11364 var always has vectype_out type.
11366 2018-06-13 Jeff Law <law@redhat.com>
11368 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11369 integer 0 for argument to print_rtl_with_bb.
11370 (rl78_reorg): Likewise.
11372 2018-06-13 David Malcolm <dmalcolm@redhat.com>
11374 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11375 from rtx to rtx_insn *.
11376 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11378 (add_sched_insns_for_speculation): Likewise for local "target",
11379 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11380 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11381 from rtx_insn ** to rtx_code_label **.
11382 (reorg_emit_nops): Likewise.
11383 (c6x_reorg): Likewise for local "call_labels".
11384 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11386 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11387 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11388 the loops over LABEL_REFS.
11389 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11391 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11392 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11393 (split_branches): Strengthen local "olabel" from rtx to
11394 rtx_insn *, adding a safe_as_a cast.
11395 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11397 (add_insn_after): Likewise for first two params.
11398 (add_insn_before): Likewise.
11399 (remove_insn): Likewise for param.
11400 (emit_pattern_before_noloc): Likewise for second and third params.
11401 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11402 (emit_call_insn_before_noloc): Likewise.
11403 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11405 (emit_barrier_before): Likewise.
11406 (emit_label_before): Strengthen "label" param from "rtx" to
11407 "rtx_code_label *". Strengthen "before" param from "rtx" to
11409 (emit_insn_after_1): Strengthen "after" param from "rtx" to
11411 (emit_pattern_after_noloc): Likewise.
11412 (emit_insn_after_noloc): Likewise.
11413 (emit_jump_insn_after_noloc): Likewise.
11414 (emit_call_insn_after_noloc): Likewise.
11415 (emit_debug_insn_after_noloc): Likewise.
11416 (emit_barrier_after): Likewise.
11417 (emit_label_after): Likewise for both params.
11418 (emit_pattern_after_setloc): Likewise for "after" param. Convert
11419 "loc" param from "int" to "location_t".
11420 (emit_insn_after_setloc): Likewise.
11421 (emit_jump_insn_after_setloc): Likewise.
11422 (emit_call_insn_after_setloc): Likewise.
11423 (emit_debug_insn_after_setloc): Likewise.
11424 (emit_pattern_before_setloc): Likewise for "before" param. Convert
11425 "loc" param from "int" to "location_t".
11426 (emit_pattern_before): Convert NULL_RTX to NULL.
11427 (emit_insn_before_setloc): Convert "loc" param from "int" to
11429 (emit_jump_insn_before_setloc): Likewise.
11430 (emit_call_insn_before_setloc): Likewise.
11431 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11432 rtx_insn *. Convert "loc" param from "int" to "location_t".
11433 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11434 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11435 Convert 3rd param from "int" to "location_t".
11436 (emit_barrier_before, emit_barrier_after, next_real_insn):
11437 Strengthen param from rtx to rtx_insn *.
11438 (emit_label_before): Strengthen 1st param from "rtx" to
11439 "rtx_code_label *". Strengthen 2nd param from "rtx" to
11441 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11442 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11443 Strengthen 2nd param from "rtx" to "rtx_insn *".
11444 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11445 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11446 Likewise. Convert 3rd param from "int" to "location_t".
11447 (emit_label_after): Strengthen 1st param from "rtx" to
11448 "rtx_code_label *".
11449 (next_real_insn, remove_insn): Strengthen param from "rtx" to
11451 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11452 from "rtx" to "rtx_insn *".
11454 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
11456 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11457 bodies streamed in with -Q.
11458 * dumpfile.c (dump_files): Add lto-stream-out dump file.
11459 * dumpfile.h (tree_dump_index): Add lto_stream_out.
11460 * gimple-streamer-out.c: Include gimple-pretty-print.h
11461 (output_bb): Dump stmts streamed.
11462 * lto-section-out.c: Include print-tree.h
11463 (lto_begin_section): Dump sections created.
11464 (lto_output_decl_index): Dump decl encoded.
11465 * lto-streamer-out.c: Include print-tree.h
11466 (create_output_block): Dump output block created.
11467 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11468 (output_function): Dump function output.
11469 (output_constructor): Dump constructor streamed.
11470 (write_global_stream): Output indexes encoded.
11471 (produce_asm_for_decls): Dump streams encoded.
11472 * lto-streamer.c (streamer_dump_file): New global var.
11473 * lto-streamer.h (streamer_dump_file): Declare.
11474 * passes.c (ipa_write_summaries): Initialize streamer dump.
11475 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11478 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
11481 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11482 offsets for register save directives. Emit a second batch of save
11483 directives, if need be, when the function accesses prior frames.
11485 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11487 * config/arc/fpu.md (fmasf4): Force operand to register.
11488 (fnmasf4): Likewise.
11490 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11492 * config/arc/arc-protos.h (arc_pad_return): Remove.
11493 * config/arc/arc.c (machine_function): Remove force_short_suffix
11495 (arc_print_operand): Adjust printing of '&'.
11496 (arc_verify_short): Remove conditional printing of short suffix.
11497 (arc_final_prescan_insn): Remove reference to size_reason.
11498 (pad_return): New function.
11499 (arc_reorg): Call pad_return.
11500 (arc_pad_return): Remove.
11501 (arc_init_machine_status): Remove reference to force_short_suffix.
11502 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11503 (attr length): When attribute iscompact is true force to 2
11504 regardless; in the case of maybe check if we want to force the
11505 instruction to have 4 bytes length.
11506 (nopv): Change it to generate 4 byte long nop as well.
11507 (blockage): New pattern.
11508 (simple_return): Remove call to arc_pad_return.
11509 (p_return_i): Likewise.
11511 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11513 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11515 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11517 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11520 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11522 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11523 for ARC700 and ARCv2.
11525 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
11528 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11529 operands[2] instead of operands[1].
11532 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11534 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11535 case, check whether the outer register overlaps an unallocatable
11536 register, not just whether it fits the required class.
11538 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11540 * poly-int.h (can_div_trunc_p): Add new overload in which all values
11542 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11543 (memrefs_conflict_p): Likewise.
11544 (init_alias_analysis): Likewise.
11545 * cfgexpand.c (expand_debug_expr): Likewise.
11546 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11547 * cse.c (fold_rtx): Likewise.
11548 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11549 * expr.c (emit_block_move_hints): Likewise.
11550 (clear_storage_hints, push_block, emit_push_insn): Likewise.
11551 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11552 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11553 (emit_group_store): Likewise.
11554 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
11555 to read the PRE/POST_MODIFY increment.
11556 * calls.c (store_one_arg): Use strip_offset.
11557 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11559 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11561 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11562 (simplify_binary_operation_1): Extend CONST_INT handling to
11564 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11565 than a HOST_WIDE_INT.
11566 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11568 (adjust_mems, add_stores): Update accodingly.
11569 (vt_canonicalize_addr): Track polynomial offsets.
11570 (emit_note_insn_var_location): Likewise.
11571 (vt_add_function_parameter): Likewise.
11572 (vt_initialize): Likewise.
11574 2018-06-12 Jeff Law <law@redhat.com>
11576 * config.gcc (alpha*-*-freebsd*): Remove.
11577 * config/alpha/freebsd.h: Remove.
11579 2018-06-12 David Malcolm <dmalcolm@redhat.com>
11582 * spellcheck-tree.c (levenshtein_distance): Rename to...
11583 (get_edit_distance): ...this, and update for underlying renaming.
11584 * spellcheck-tree.h (levenshtein_distance): Rename to...
11585 (get_edit_distance): ...this.
11586 * spellcheck.c (levenshtein_distance): Rename to...
11587 (get_edit_distance): ...this. Convert from Levenshtein distance
11588 to Damerau-Levenshtein distance by supporting transpositions of
11589 adjacent characters. Rename "v1" to "v_next" and "v0" to
11591 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11592 (selftest::test_edit_distance_unit_test_oneway): ...this, and
11593 update for underlying renaming.
11594 (selftest::levenshtein_distance_unit_test): Rename to...
11595 (selftest::test_get_edit_distance_unit): ...this, and update for
11596 underlying renaming.
11597 (selftest::test_find_closest_string): Add example from PR 69968
11598 where transposition helps
11599 (selftest::test_metric_conditions): Update for renaming.
11600 (selftest::test_metric_conditions): Likewise.
11601 (selftest::spellcheck_c_tests): Likewise.
11602 * spellcheck.h (levenshtein_distance): Rename both overloads to...
11603 (get_edit_distance): ...this.
11604 (best_match::consider): Update for renaming.
11606 2018-06-12 Martin Sebor <msebor@redhat.com>
11608 PR tree-optimization/85259
11609 * builtins.c (compute_objsize): Handle constant offsets.
11610 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11611 true iff a warning has been issued.
11612 * gimple.h (gimple_nonartificial_location): New function.
11613 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11614 gimple_nonartificial_location and handle -Wno-system-headers.
11615 (handle_builtin_stxncpy): Same.
11617 2018-06-12 Martin Sebor <msebor@redhat.com>
11620 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11622 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
11624 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11625 BUILTIN_VEC_XST entries for pointer to double and long long.
11627 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
11630 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11632 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11635 2018-06-12 Martin Liska <mliska@suse.cz>
11637 * doc/options.texi: Document IntegerRange.
11639 2018-06-12 Martin Liska <mliska@suse.cz>
11641 * config/i386/i386.opt: Make MPX-related options as Deprecated.
11642 * opt-functions.awk: Handle Deprecated flag.
11643 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11645 (read_cmdline_option): Report warning for a deprecated option.
11646 * opts.h (struct cl_option): Add new field cl_deprecated.
11647 (CL_ERR_DEPRECATED): New.
11649 2018-06-12 Martin Liska <mliska@suse.cz>
11651 * doc/options.texi: Document Deprecated option flag.
11653 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11655 * config/arc/arc-arch.h (arc_extras): New enum.
11656 (arc_cpu_t):Add field extra.
11657 (arc_cpu_types): Consider the extras.
11658 * config/arc/arc-cpus.def: Add extras info.
11659 * config/arc/arc-opts.h (processor_type): Consider extra field.
11660 * config/arc/arc.c (arc_override_options): Handle extra field.
11662 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11664 * config/arc/arc-arch.h: Update ARC_OPTX macro.
11665 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11667 * config/arc/arc.c (arc_init): Update pic warning.
11668 (irq_range): Update irq range parsing warnings.
11669 (arc_override_options): Update various warning messages.
11670 (arc_handle_aux_attribute): Likewise.
11672 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
11674 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11676 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11678 * doc/sourcebuild.texi: Document usage of line number 0 in verify
11679 compiler messages directives.
11681 2018-06-12 Matthew Fortune <mfortune@gmail.com>
11683 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11684 * config/mips/mips-tables.opt: Regenerate.
11685 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11687 * doc/invoke.texi: Document -march=i6500.
11689 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
11691 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11692 (i6400_gpmul): Add cpu_unit.
11693 (i6400_gpdiv): Likewise.
11694 (i6400_msa_add_d): Update reservations.
11695 (i6400_msa_int_add) Likewise.
11696 (i6400_msa_short_logic3) Likewise.
11697 (i6400_msa_short_logic2) Likewise.
11698 (i6400_msa_short_logic) Likewise.
11699 (i6400_msa_move) Likewise.
11700 (i6400_msa_cmp) Likewise.
11701 (i6400_msa_short_float2) Likewise.
11702 (i6400_msa_div_d) Likewise.
11703 (i6400_msa_long_logic1) Likewise.
11704 (i6400_msa_long_logic2) Likewise.
11705 (i6400_msa_mult) Likewise.
11706 (i6400_msa_long_float2) Likewise.
11707 (i6400_msa_long_float4) Likewise.
11708 (i6400_msa_long_float5) Likewise.
11709 (i6400_msa_long_float8) Likewise.
11710 (i6400_fpu_fadd): Include frint type.
11711 (i6400_fpu_store): New define_insn_reservation.
11712 (i6400_fpu_load): Likewise.
11713 (i6400_fpu_move): Likewise.
11714 (i6400_fpu_fcmp): Likewise.
11715 (i6400_fpu_fmadd): Likewise.
11716 (i6400_int_mult): Include imul3nc type and update reservation.
11717 (i6400_int_div): Include idiv3 type and update reservation.
11718 (i6400_int_load): Update to check type not move_type.
11719 (i6400_int_store): Likewise.
11720 (i6400_int_prefetch): Set zero latency.
11722 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
11724 * gcc.c: Document new %@{...} sequence.
11725 (LINK_COMMAND_SPEC): Use it for the -L switches.
11726 (cpp_unique_options): Use it for the -I switches.
11727 (at_file_argbuf): New global variable.
11728 (in_at_file): Likewise.
11729 (alloc_args): Create at_file_argbuf.
11730 (clear_args): Truncate at_file_argbuf.
11731 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11732 (open_at_file): New function.
11733 (close_at_file): Likewise.
11734 (create_at_file): Delete.
11735 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11738 (validate_switches_from_spec): Deal with %@{...} sequence.
11739 (validate_switches): Likewise.
11740 (driver::finalize): Call clear_args.
11742 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
11744 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11746 2018-06-11 Martin Sebor <msebor@redhat.com>
11748 * doc/invoke.texi (-Wall): List -Wc++17-compat.
11749 (Wno-class-memaccess): Add @opindex.
11750 (Wno-templates, Wno-multiple-inheritance): Same.
11751 (Wno-virtual-inheritance, Wno-namespaces): Same.
11752 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11753 (Wno-format-overflow, Wno-format-truncation): Same.
11754 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11755 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11756 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11757 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11758 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11759 (Wno-misspelled-isr): Same.
11761 2018-06-11 Martin Sebor <msebor@redhat.com>
11763 * PR tree-optimization/86083
11764 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11766 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11768 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11770 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11773 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11774 on the correct operand.
11775 (*movdi_internal64): Ditto.
11777 2018-06-11 Martin Liska <mliska@suse.cz>
11779 PR tree-optimization/86089
11780 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11782 2018-06-11 Julia Koval <julia.koval@intel.com>
11784 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11785 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11786 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11788 2018-06-11 Olivier Hainque <hainque@adacore.com>
11790 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11791 for Ada with strict dwarf2.
11793 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11796 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11799 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11801 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11803 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
11805 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11808 2018-06-08 Martin Liska <mliska@suse.cz>
11810 * tree-cfg.h (debug_function): Fix argument type to match
11813 2018-06-08 Martin Liska <mliska@suse.cz>
11815 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11816 Remove usage of MPX-related (and removed) fields.
11817 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11819 2018-06-08 David Malcolm <dmalcolm@redhat.com>
11821 * cfg.c (debug): Use TDF_NONE rather than 0.
11822 * cfghooks.c (debug): Likewise.
11823 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11824 (struct dump_option_value_info): Convert to...
11825 (struct kv_pair): ...this template type.
11826 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11828 (optinfo_verbosity_options): Likewise.
11829 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11831 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11832 than int for "optgroup_flags" param.
11833 (dump_generic_expr_loc): Use dump_flags_t rather than int for
11835 (dump_dec): Likewise.
11836 (dump_finish): Use TDF_NONE rather than 0.
11837 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11838 rather than int for "optgroup_flags" param. Use TDF_NONE rather
11839 than 0. Update for change to option_ptr.
11840 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11841 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
11842 0. Update for changes to optinfo_verbosity_options and
11844 (opt_info_switch_p): Convert optgroup_flags from int to
11846 (dump_basic_block): Use dump_flags_t rather than int
11847 for "dump_kind" param.
11848 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11849 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11850 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11851 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11852 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11853 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11854 TDF_NONE): Convert from macros to...
11855 (enum dump_flag): ...this new enum.
11856 (dump_flags_t): Update to use enum.
11857 (operator|, operator&, operator~, operator|=, operator&=):
11858 Implement for dump_flags_t.
11859 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11860 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11861 Convert from macros to...
11862 (enum optgroup_flag): ...this new enum.
11863 (optgroup_flags_t): New typedef.
11864 (operator|, operator|=): Implement for optgroup_flags_t.
11865 (struct dump_file_info): Convert field "alt_flags" to
11866 dump_flags_t. Convert field "optgroup_flags" to
11868 (dump_basic_block): Use dump_flags_t rather than int for param.
11869 (dump_generic_expr_loc): Likewise.
11870 (dump_dec): Likewise.
11871 (dump_register): Convert param "optgroup_flags" to
11873 (opt_info_enable_passes): Likewise.
11874 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11876 * gimple-pretty-print.c (debug): Likewise.
11877 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11878 (merged_store_group::apply_stores): Likewise.
11879 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11880 * gimple.c (verify_gimple_pp): Likewise.
11881 * graphite-poly.c (print_pbb_body): Likewise.
11882 * passes.c (pass_manager::register_one_dump_file): Convert
11883 local "optgroup_flags" to optgroup_flags_t.
11884 * print-tree.c (print_node): Use TDF_NONE rather than 0.
11886 (debug_body): Likewise.
11887 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11888 to optgroup_flags_t.
11889 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11891 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11892 (convert_mult_to_fma): Likewise.
11893 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11894 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11895 * tree-vect-data-refs.c (dump_lower_bound): Convert param
11896 "dump_kind" to dump_flags_t.
11898 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11900 * config/rs6000/rs6000.c (min, max): Delete.
11902 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11904 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11905 -mabi=spe and -mabi=no-spe.
11907 2018-06-08 Martin Liska <mliska@suse.cz>
11909 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11910 where we expect an existing summary.
11912 2018-06-08 Martin Liska <mliska@suse.cz>
11914 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11915 * ipa-inline.h (estimate_edge_growth): Likewise.
11917 2018-06-08 Martin Liska <mliska@suse.cz>
11919 * cgraph.c (function_version_hasher::hash): Use
11920 cgraph_node::get_uid ().
11921 (function_version_hasher::equal):
11922 * cgraph.h (cgraph_node::get_uid): New method.
11923 * ipa-inline.c (update_caller_keys): Use
11924 cgraph_node::get_uid ().
11925 (update_callee_keys): Likewise.
11926 * ipa-utils.c (searchc): Likewise.
11927 (ipa_reduced_postorder): Likewise.
11928 * lto-cgraph.c (input_node): Likewise.
11929 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11930 * symbol-summary.h (symtab_insertion): Likewise.
11931 (symtab_removal): Likewise.
11932 (symtab_duplication): Likewise.
11933 * tree-pretty-print.c (dump_function_header): Likewise.
11934 * tree-sra.c (convert_callers_for_node): Likewise.
11936 2018-06-08 Martin Liska <mliska@suse.cz>
11938 * cgraph.c (symbol_table::create_edge): Always assign a new
11940 (symbol_table::free_edge): Do not recycle numbers.
11941 * cgraph.h (cgraph_edge::get): New method.
11942 * symbol-summary.h (symtab_removal): Use it.
11943 (symtab_duplication): Likewise.
11944 (call_summary::hashable_uid): Remove.
11946 2018-06-08 Martin Liska <mliska@suse.cz>
11948 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11949 (initialize_growth_caches): Remove.
11950 (free_growth_caches): Likewise.
11951 (do_estimate_edge_time): Use edge_growth_cache.
11952 (do_estimate_edge_size): Likewise.
11953 (do_estimate_edge_hints): Likewise.
11954 * ipa-inline.c (reset_edge_caches): Likewise.
11955 (recursive_inlining): Likewise.
11956 (inline_small_functions): Likewise.
11957 * ipa-inline.h (initialize_growth_caches): Remove.
11958 (estimate_edge_size): Likewise.
11959 (estimate_edge_time): Likewise.
11960 (estimate_edge_hints): Likewise.
11961 (reset_edge_growth_cache): Likewise.
11962 * symbol-summary.h (call_summary::remove): New method.
11964 2018-06-08 Martin Liska <mliska@suse.cz>
11966 * ipa-cp.c (class edge_clone_summary): New summary.
11967 (grow_edge_clone_vectors): Remove.
11968 (ipcp_edge_duplication_hook): Remove.
11969 (class edge_clone_summary_t): New call_summary class.
11970 (ipcp_edge_removal_hook): Remove.
11971 (edge_clone_summary_t::duplicate): New function.
11972 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11973 (create_specialized_node): Likewise.
11974 (ipcp_driver): Initialize edge_clone_summaries and do not
11977 2018-06-08 Martin Liska <mliska@suse.cz>
11979 * symbol-summary.h (get): New function.
11980 (call_summary::m_initialize_when_cloning): New class member.
11982 2018-06-08 Martin Liska <mliska@suse.cz>
11984 * cgraph.c (cgraph_node::remove): Do not recycle uid.
11985 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11986 (symbol_table::allocate_cgraph_symbol): Do not set uid.
11987 * passes.c (uid_hash_t): Record removed_nodes by their uids.
11988 (remove_cgraph_node_from_order): Use the removed_nodes set.
11989 (do_per_function_toporder): Likwise.
11990 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11991 instead of summary_uid.
11992 (symtab_removal): Likewise.
11993 (symtab_duplication): Likewise.
11995 2018-06-08 Martin Liska <mliska@suse.cz>
11997 * ipa-cp.c (ipcp_store_bits_results): Use
11998 ipcp_transformation_sum.
11999 (ipcp_store_vr_results): Likewise.
12000 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12002 (ipcp_transformation_initialize): ... this.
12003 (ipa_set_node_agg_value_chain):
12004 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12005 (write_ipcp_transformation_info): Likewise.
12006 (read_ipcp_transformation_info): Likewise.
12007 (ipcp_update_bits): Likewise.
12008 (ipcp_update_vr): Likewise.
12009 (ipcp_transform_function): Likewise.
12010 * ipa-prop.h: Rename ipcp_transformation_summary to
12011 ipcp_transformation.
12012 (class ipcp_transformation_t): New function summary.
12013 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12014 (ipa_get_agg_replacements_for_node): Likewise.
12016 2018-06-08 Martin Liska <mliska@suse.cz>
12018 * ipa-pure-const.c (struct funct_state_d): Do it class instead
12020 (class funct_state_summary_t): New function_summary class.
12021 (has_function_state): Remove.
12022 (get_function_state): Likewise.
12023 (set_function_state): Likewise.
12024 (add_new_function): Likewise.
12025 (funct_state_summary_t::insert): New function.
12026 (duplicate_node_data): Remove.
12027 (remove_node_data): Remove.
12028 (funct_state_summary_t::duplicate): New function.
12029 (register_hooks): Create new funct_state_summaries.
12030 (pure_const_generate_summary): Use it.
12031 (pure_const_write_summary): Likewise.
12032 (pure_const_read_summary): Likewise.
12033 (propagate_pure_const): Likewise.
12034 (propagate_nothrow): Likewise.
12035 (dump_malloc_lattice): Likewise.
12036 (propagate_malloc): Likewise.
12037 (execute): Do not register hooks, just remove summary
12039 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12042 2018-06-08 Martin Liska <mliska@suse.cz>
12044 * ipa-reference.c (remove_node_data): Remove.
12045 (duplicate_node_data): Likewise.
12046 (class ipa_ref_var_info_summary_t): New class.
12047 (class ipa_ref_opt_summary_t): Likewise.
12048 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12049 (get_reference_optimization_summary): Use
12050 ipa_ref_opt_sum_summaries.
12051 (set_reference_vars_info): Remove.
12052 (set_reference_optimization_summary): Likewise.
12053 (ipa_init): Create summaries.
12054 (init_function_info): Use function summary.
12055 (ipa_ref_opt_summary_t::duplicate): New function.
12056 (ipa_ref_opt_summary_t::remove): New function.
12057 (get_read_write_all_from_node): Fix GNU coding style.
12058 (propagate): Use function summary.
12059 (write_node_summary_p): Fix GNU coding style.
12060 (stream_out_bitmap): Likewise.
12061 (ipa_reference_read_optimization_summary): Use function summary.
12062 (ipa_reference_c_finalize): Do not release hooks.
12064 2018-06-08 Martin Liska <mliska@suse.cz>
12066 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12067 (analyze_function_body): Extract multiple calls of get_create.
12068 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12069 * ipa-inline.c (recursive_inlining): Use ::get method.
12070 * ipa-inline.h (estimate_edge_growth): Likewise.
12072 2018-06-08 Martin Liska <mliska@suse.cz>
12074 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12076 (hsa_function_summary::hsa_function_summary): Use the new enum
12078 (hsa_gpu_implementation_p): Use hsa_summaries::get.
12079 * hsa-gen.c (hsa_get_host_function): Likewise.
12080 (get_brig_function_name): Likewise.
12081 * ipa-hsa.c (process_hsa_functions): Likewise.
12082 (ipa_hsa_write_summary): Likewise.
12083 * symbol-summary.h (symtab_duplication): Use ::get function/
12084 (get): New function.
12086 2018-06-08 Martin Liska <mliska@suse.cz>
12088 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12090 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12091 (hsa_register_kernel): Likewise.
12092 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12093 * hsa-gen.c (hsa_get_host_function): Likewise.
12094 (get_brig_function_name): Likewise.
12095 (generate_hsa): Likewise.
12096 (pass_gen_hsail::execute): Likewise.
12097 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12098 (devirtualization_time_bonus): Likewise.
12099 (ipcp_propagate_stage): Likewise.
12100 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12101 (edge_set_predicate): Likewise.
12102 (evaluate_conditions_for_known_args): Likewise.
12103 (evaluate_properties_for_edge): Likewise.
12104 (ipa_fn_summary::reset): Likewise.
12105 (ipa_fn_summary_t::duplicate): Likewise.
12106 (dump_ipa_call_summary): Likewise.
12107 (ipa_dump_fn_summary): Likewise.
12108 (analyze_function_body): Likewise.
12109 (compute_fn_summary): Likewise.
12110 (estimate_edge_devirt_benefit): Likewise.
12111 (estimate_edge_size_and_time): Likewise.
12112 (estimate_calls_size_and_time): Likewise.
12113 (estimate_node_size_and_time): Likewise.
12114 (inline_update_callee_summaries): Likewise.
12115 (remap_edge_change_prob): Likewise.
12116 (remap_edge_summaries): Likewise.
12117 (ipa_merge_fn_summary_after_inlining): Likewise.
12118 (ipa_update_overall_fn_summary): Likewise.
12119 (read_ipa_call_summary): Likewise.
12120 (inline_read_section): Likewise.
12121 (write_ipa_call_summary): Likewise.
12122 (ipa_fn_summary_write): Likewise.
12123 (ipa_free_fn_summary): Likewise.
12124 * ipa-hsa.c (process_hsa_functions): Likewise.
12125 (ipa_hsa_write_summary): Likewise.
12126 (ipa_hsa_read_section): Likewise.
12127 * ipa-icf.c (sem_function::merge): Likewise.
12128 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12129 (do_estimate_edge_time): Likewise.
12130 (estimate_size_after_inlining): Likewise.
12131 (estimate_growth): Likewise.
12132 (growth_likely_positive): Likewise.
12133 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12134 (inline_call): Likewise.
12135 * ipa-inline.c (caller_growth_limits): Likewise.
12136 (can_inline_edge_p): Likewise.
12137 (can_inline_edge_by_limits_p): Likewise.
12138 (compute_uninlined_call_time): Likewise.
12139 (compute_inlined_call_time): Likewise.
12140 (want_inline_small_function_p): Likewise.
12141 (edge_badness): Likewise.
12142 (update_caller_keys): Likewise.
12143 (update_callee_keys): Likewise.
12144 (recursive_inlining): Likewise.
12145 (inline_small_functions): Likewise.
12146 (inline_to_all_callers_1): Likewise.
12147 (dump_overall_stats): Likewise.
12148 (early_inline_small_functions): Likewise.
12149 (early_inliner): Likewise.
12150 * ipa-inline.h (estimate_edge_growth): Likewise.
12151 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12152 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12153 * ipa-prop.h (IPA_NODE_REF): Likewise.
12154 (IPA_EDGE_REF): Likewise.
12155 * ipa-pure-const.c (malloc_candidate_p): Likewise.
12156 (propagate_malloc): Likewise.
12157 * ipa-split.c (execute_split_functions): Likewise.
12158 * symbol-summary.h: Rename get to get_create.
12160 (get_create): Likewise.
12161 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12163 2018-06-08 Martin Liska <mliska@suse.cz>
12165 * symbol-summary.h (release): Move definition out of class
12167 (symtab_removal): Likewise.
12168 (symtab_duplication): Likewise.
12170 2018-06-08 Martin Liska <mliska@suse.cz>
12172 * symbol-summary.h (function_summary): Move constructor
12173 implementation out of class declaration.
12174 (release): Likewise.
12175 (symtab_insertion): Likewise.
12176 (symtab_removal): Likewise.
12177 (symtab_duplication): Likewise.
12180 2018-06-08 Martin Liska <mliska@suse.cz>
12182 * Makefile.in: Remove support for MPX (macros, related functions,
12183 fields in cgraph_node, ...).
12184 * builtin-types.def (BT_BND): Likewise.
12185 (BT_FN_BND_CONST_PTR): Likewise.
12186 (BT_FN_CONST_PTR_BND): Likewise.
12187 (BT_FN_VOID_PTR_BND): Likewise.
12188 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12189 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12190 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12191 (expand_builtin_mempcpy_with_bounds): Likewise.
12192 (expand_builtin_memset_with_bounds): Likewise.
12193 (expand_builtin_memset_args): Likewise.
12194 (std_expand_builtin_va_start): Likewise.
12195 (expand_builtin): Likewise.
12196 (expand_builtin_with_bounds): Likewise.
12197 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12198 (DEF_LIB_BUILTIN_CHKP): Likewise.
12199 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12200 (DEF_CHKP_BUILTIN): Likewise.
12201 (BUILT_IN_MEMCPY): Likewise.
12202 (BUILT_IN_MEMMOVE): Likewise.
12203 (BUILT_IN_MEMPCPY): Likewise.
12204 (BUILT_IN_MEMSET): Likewise.
12205 (BUILT_IN_STPCPY): Likewise.
12206 (BUILT_IN_STRCAT): Likewise.
12207 (BUILT_IN_STRCHR): Likewise.
12208 (BUILT_IN_STRCPY): Likewise.
12209 (BUILT_IN_STRLEN): Likewise.
12210 (BUILT_IN_MEMCPY_CHK): Likewise.
12211 (BUILT_IN_MEMMOVE_CHK): Likewise.
12212 (BUILT_IN_MEMPCPY_CHK): Likewise.
12213 (BUILT_IN_MEMSET_CHK): Likewise.
12214 (BUILT_IN_STPCPY_CHK): Likewise.
12215 (BUILT_IN_STRCAT_CHK): Likewise.
12216 (BUILT_IN_STRCPY_CHK): Likewise.
12217 * calls.c (store_bounds): Likewise.
12218 (emit_call_1): Likewise.
12219 (special_function_p): Likewise.
12220 (maybe_warn_nonstring_arg): Likewise.
12221 (initialize_argument_information): Likewise.
12222 (finalize_must_preallocate): Likewise.
12223 (compute_argument_addresses): Likewise.
12224 (expand_call): Likewise.
12225 * cfgexpand.c (expand_call_stmt): Likewise.
12226 (expand_return): Likewise.
12227 (expand_gimple_stmt_1): Likewise.
12228 (pass_expand::execute): Likewise.
12229 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12230 (cgraph_node::remove): Likewise.
12231 (cgraph_node::dump): Likewise.
12232 (cgraph_node::verify_node): Likewise.
12233 * cgraph.h (chkp_function_instrumented_p): Likewise.
12234 (symtab_node::get_alias_target): Likewise.
12235 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12236 (cgraph_local_p): Likewise.
12237 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12238 (cgraph_edge::rebuild_references): Likewise.
12239 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12240 (walk_polymorphic_call_targets): Likewise.
12241 (cgraph_node::expand_thunk): Likewise.
12242 (symbol_table::output_weakrefs): Likewise.
12243 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12244 (ix86_handle_option): Likewise.
12245 * config/i386/constraints.md: Likewise.
12246 * config/i386/i386-builtin-types.def (BND): Likewise.
12250 * config/i386/i386-builtin.def (BDESC_END): Likewise.
12251 (BDESC_FIRST): Likewise.
12253 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12254 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12255 * config/i386/i386.c (enum reg_class): Likewise.
12256 (ix86_target_string): Likewise.
12257 (ix86_option_override_internal): Likewise.
12258 (ix86_conditional_register_usage): Likewise.
12259 (ix86_valid_target_attribute_inner_p): Likewise.
12260 (ix86_set_indirect_branch_type): Likewise.
12261 (ix86_set_current_function): Likewise.
12262 (ix86_function_arg_regno_p): Likewise.
12263 (init_cumulative_args): Likewise.
12264 (ix86_function_arg_advance): Likewise.
12265 (ix86_function_arg): Likewise.
12266 (ix86_pass_by_reference): Likewise.
12267 (ix86_function_value_regno_p): Likewise.
12268 (ix86_function_value_1): Likewise.
12269 (ix86_function_value_bounds): Likewise.
12270 (ix86_return_in_memory): Likewise.
12271 (ix86_setup_incoming_vararg_bounds): Likewise.
12272 (ix86_va_start): Likewise.
12273 (indirect_thunk_need_prefix): Likewise.
12274 (print_reg): Likewise.
12275 (ix86_print_operand): Likewise.
12276 (ix86_expand_call): Likewise.
12277 (ix86_output_function_return): Likewise.
12278 (reg_encoded_number): Likewise.
12279 (BDESC_VERIFYS): Likewise.
12280 (ix86_init_mpx_builtins): Likewise.
12281 (ix86_init_builtins): Likewise.
12282 (ix86_emit_cmove): Likewise.
12283 (ix86_emit_move_max): Likewise.
12284 (ix86_expand_builtin): Likewise.
12285 (ix86_builtin_mpx_function): Likewise.
12286 (ix86_get_arg_address_for_bt): Likewise.
12287 (ix86_load_bounds): Likewise.
12288 (ix86_store_bounds): Likewise.
12289 (ix86_load_returned_bounds): Likewise.
12290 (ix86_store_returned_bounds): Likewise.
12291 (ix86_class_likely_spilled_p): Likewise.
12292 (ix86_hard_regno_mode_ok): Likewise.
12293 (x86_order_regs_for_local_alloc): Likewise.
12294 (ix86_mitigate_rop): Likewise.
12295 (ix86_bnd_prefixed_insn_p): Likewise.
12296 (ix86_mpx_bound_mode): Likewise.
12297 (ix86_make_bounds_constant): Likewise.
12298 (ix86_initialize_bounds): Likewise.
12299 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12300 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12301 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12302 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12303 (TARGET_CHKP_BOUND_MODE): Likewise.
12304 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12305 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12306 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12307 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12308 * config/i386/i386.h (TARGET_MPX): Likewise.
12309 (TARGET_MPX_P): Likewise.
12310 (VALID_BND_REG_MODE): Likewise.
12311 (FIRST_BND_REG): Likewise.
12312 (LAST_BND_REG): Likewise.
12313 (enum reg_class): Likewise.
12314 (BND_REG_P): Likewise.
12315 (BND_REGNO_P): Likewise.
12316 (BNDmode): Likewise.
12317 (ADJUST_INSN_LENGTH): Likewise.
12318 * config/i386/i386.md: Likewise.
12319 * config/i386/i386.opt: Likewise.
12320 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12321 (defined): Likewise.
12322 (LINK_MPX): Likewise.
12323 (MPX_SPEC): Likewise.
12324 (LIBMPX_SPEC): Likewise.
12325 (LIBMPXWRAPPERS_SPEC): Likewise.
12326 (CHKP_SPEC): Likewise.
12327 * config/i386/predicates.md: Likewise.
12328 * dbxout.c (dbxout_type): Likewise.
12329 * doc/extend.texi: Likewise.
12330 * doc/invoke.texi: Likewise.
12331 * doc/md.texi: Likewise.
12332 * doc/tm.texi: Likewise.
12333 * doc/tm.texi.in: Likewise.
12334 * dwarf2out.c (is_base_type): Likewise.
12335 (gen_formal_types_die): Likewise.
12336 (gen_subprogram_die): Likewise.
12337 (gen_type_die_with_usage): Likewise.
12338 (gen_decl_die): Likewise.
12339 (dwarf2out_late_global_decl): Likewise.
12340 * expr.c (expand_assignment): Likewise.
12341 (emit_storent_insn): Likewise.
12342 (store_expr_with_bounds): Likewise.
12343 (store_expr): Likewise.
12344 (expand_expr_real_1): Likewise.
12345 * expr.h (store_expr_with_bounds): Likewise.
12346 * function.c (use_register_for_decl): Likewise.
12347 (struct bounds_parm_data): Likewise.
12348 (assign_parms_augmented_arg_list): Likewise.
12349 (assign_parm_find_entry_rtl): Likewise.
12350 (assign_parm_is_stack_parm): Likewise.
12351 (assign_parm_load_bounds): Likewise.
12352 (assign_bounds): Likewise.
12353 (assign_parms): Likewise.
12354 (expand_function_start): Likewise.
12355 * gcc.c (CHKP_SPEC): Likewise.
12356 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12357 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12358 (wrestrict_dom_walker::check_call): Likewise.
12359 * gimple.c (gimple_build_call_from_tree): Likewise.
12360 * gimple.h (enum gf_mask): Likewise.
12361 (gimple_call_with_bounds_p): Likewise.
12362 (gimple_call_set_with_bounds): Likewise.
12363 * gimplify.c (gimplify_init_constructor): Likewise.
12364 * ipa-cp.c (initialize_node_lattices): Likewise.
12365 (propagate_constants_across_call): Likewise.
12366 (find_more_scalar_values_for_callers_subset): Likewise.
12367 * ipa-hsa.c (process_hsa_functions): Likewise.
12368 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12369 * ipa-icf.c (sem_function::merge): Likewise.
12370 * ipa-inline.c (early_inliner): Likewise.
12371 * ipa-pure-const.c (warn_function_noreturn): Likewise.
12372 (warn_function_cold): Likewise.
12373 (propagate_pure_const): Likewise.
12374 * ipa-ref.h (enum GTY): Likewise.
12375 * ipa-split.c (find_retbnd): Likewise.
12376 (consider_split): Likewise.
12377 (split_function): Likewise.
12378 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12379 * ipa.c (walk_polymorphic_call_targets): Likewise.
12380 (symbol_table::remove_unreachable_nodes): Likewise.
12381 (process_references): Likewise.
12382 (cgraph_build_static_cdtor_1): Likewise.
12383 * lto-cgraph.c (lto_output_node): Likewise.
12384 (output_refs): Likewise.
12385 (compute_ltrans_boundary): Likewise.
12386 (input_overwrite_node): Likewise.
12387 (input_node): Likewise.
12388 (input_cgraph_1): Likewise.
12389 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12390 * passes.c (pass_manager::execute_early_local_passes): Likewise.
12391 (class pass_chkp_instrumentation_passes): Likewise.
12392 (make_pass_chkp_instrumentation_passes): Likewise.
12393 * passes.def: Likewise.
12394 * rtl.h (struct GTY): Likewise.
12395 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12396 * stor-layout.c (layout_type): Likewise.
12397 * symtab.c: Likewise.
12398 * target.def: Likewise.
12399 * targhooks.c (default_chkp_bound_type): Likewise.
12400 (default_chkp_bound_mode): Likewise.
12401 (default_builtin_chkp_function): Likewise.
12402 (default_chkp_function_value_bounds): Likewise.
12403 (default_chkp_make_bounds_constant): Likewise.
12404 (default_chkp_initialize_bounds): Likewise.
12405 * targhooks.h (default_chkp_bound_type): Likewise.
12406 (default_chkp_bound_mode): Likewise.
12407 (default_builtin_chkp_function): Likewise.
12408 (default_chkp_function_value_bounds): Likewise.
12409 (default_chkp_make_bounds_constant): Likewise.
12410 (default_chkp_initialize_bounds): Likewise.
12411 * toplev.c (compile_file): Likewise.
12412 (process_options): Likewise.
12413 * tree-core.h (DEF_BUILTIN): Likewise.
12414 (DEF_BUILTIN_CHKP): Likewise.
12415 * tree-inline.c (declare_return_variable): Likewise.
12416 (remap_gimple_stmt): Likewise.
12417 (copy_bb): Likewise.
12418 (initialize_inlined_parameters): Likewise.
12419 (expand_call_inline): Likewise.
12420 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12421 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12422 (make_pass_ipa_chkp_produce_thunks): Likewise.
12423 (make_pass_chkp): Likewise.
12424 (make_pass_chkp_opt): Likewise.
12425 (make_pass_chkp_instrumentation_passes): Likewise.
12426 * tree-pretty-print.c (dump_generic_node): Likewise.
12427 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12428 * tree-ssa-dce.c (propagate_necessity): Likewise.
12429 (eliminate_unnecessary_stmts): Likewise.
12430 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12431 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12432 * tree-ssa-sccvn.h: Likewise.
12433 * tree-ssa-strlen.c (get_string_length): Likewise.
12434 (valid_builtin_call): Likewise.
12435 (adjust_last_stmt): Likewise.
12436 (handle_builtin_strchr): Likewise.
12437 (handle_builtin_strcpy): Likewise.
12438 (handle_builtin_stxncpy): Likewise.
12439 (handle_builtin_memcpy): Likewise.
12440 (handle_builtin_strcat): Likewise.
12441 (strlen_check_and_optimize_stmt): Likewise.
12442 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12443 * tree-streamer-in.c: Likewise.
12444 * tree-streamer.c (record_common_node): Likewise.
12445 * tree.c (tree_code_size): Likewise.
12446 (wide_int_to_tree_1): Likewise.
12447 (type_contains_placeholder_1): Likewise.
12448 (build_common_tree_nodes): Likewise.
12449 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12450 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12451 (POINTER_BOUNDS_P): Likewise.
12452 (BOUNDED_TYPE_P): Likewise.
12453 (BOUNDED_P): Likewise.
12454 (CALL_WITH_BOUNDS_P): Likewise.
12455 (pointer_bounds_type_node): Likewise.
12456 * value-prof.c (gimple_ic): Likewise.
12457 * var-tracking.c (vt_add_function_parameters): Likewise.
12458 * varasm.c (make_decl_rtl): Likewise.
12459 (assemble_start_function): Likewise.
12460 (output_constant): Likewise.
12461 (maybe_assemble_visibility): Likewise.
12462 * varpool.c (ctor_for_folding): Likewise.
12463 * chkp-builtins.def: Remove.
12464 * ipa-chkp.c: Remove.
12465 * ipa-chkp.h: Remove.
12466 * rtl-chkp.c: Remove.
12467 * rtl-chkp.h: Remove.
12468 * tree-chkp-opt.c: Remove.
12469 * tree-chkp.c: Remove.
12470 * tree-chkp.h: Remove.
12472 2018-06-07 Carl Love <cel@us.ibm.com>
12474 * config/rs6000/vsx.md (vextract_fp_from_shorth,
12475 vextract_fp_from_shortl): Add BE support.
12477 2018-06-07 Paul Koning <ni1d@arrl.net>
12479 * compare-elim.c (try_merge_compare): Don't merge compare if
12480 address contains a side effect.
12481 (try_eliminate_compare): Likewise.
12483 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
12485 * config.gcc: Support "tremont".
12486 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12487 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12489 * config/i386/i386.c (m_TREMONT): Define.
12490 (processor_target_table): Add "tremont".
12491 (PTA_TREMONT): Define.
12492 (ix86_lea_outperforms): Add TARGET_TREMONT.
12493 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12494 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12495 and M_INTEL_GOLDMONT_PLUS.
12496 (fold_builtin_cpu): Add "tremont".
12497 (ix86_add_stmt_cost): Add TARGET_TREMONT.
12498 (ix86_option_override_internal): Add "tremont".
12499 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12500 (processor_type): Add PROCESSOR_TREMONT.
12501 * config/i386/x86-tune.def: Add m_TREMONT.
12502 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12504 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12506 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12507 symbol defined for msp430i* devices to be lower case.
12509 2018-06-07 Richard Biener <rguenther@suse.de>
12511 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12512 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
12513 Properly wrap signed arithmetic if overflow wraps.
12515 2018-06-07 Jakub Jelinek <jakub@redhat.com>
12517 PR tree-optimization/69615
12518 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12519 of a cast from a same precision integral SSA_NAME in a bb dominated
12520 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
12521 cast to utype if rhs2 has already a compatible type.
12523 2018-06-07 Richard Biener <rguenther@suse.de>
12525 PR tree-optimization/85935
12526 * graphite-scop-detection.c (find_params_in_bb): Analyze
12527 condition operands with respect to the correct loop. Assert
12528 the analysis doesn't fail.
12530 2018-06-04 Carl Love <cel@us.ibm.com>
12532 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12533 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
12534 as it is slightly cheaper.
12535 (first_match_or_eos_index_<mode>):
12536 Calculate index using natural element order.
12537 (first_match_index_<mode>):
12538 Calculate index using natural element order.
12539 (first_match_or_eos_index_<mode>):
12540 Calculate index using natural order.
12541 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12542 for BE and LE modes.
12543 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12544 P9V_BUILTIN_VCLZLSBB_V16QI.
12545 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12548 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12550 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12551 indentation and line wrap for many prototypes. Add missing
12552 @smallexample directives around block of prototypes for vec_xl and
12555 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
12557 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12558 track if we pass or return IEEE 128-bit floating point.
12559 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12560 C++ mangling that is compatible with GCC 8.1.
12561 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12562 (init_cumulative_args): Note if we pass or return IEEE 128-bit
12563 floating point types.
12564 (rs6000_function_arg_advance_1): Likewise.
12565 (rs6000_mangle_type): Optionally generate mangled names that match
12566 what GCC 8.1 generated for IEEE 128-bit floating point types.
12567 (rs6000_globalize_decl_name): If we have an external function that
12568 passes or returns IEEE 128-bit types, generate a weak reference
12569 from the mangled name used in GCC 8.1 to the current mangled
12571 (rs6000_init_builtins): Make __ibm128 use the long double type if
12572 long double is IBM extended double. Make __float128 use the long
12573 double type if long double is IEEE 128-bit.
12576 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12577 macro for __ibm128 built-in functions.
12578 (PACK_IF): Add __ibm128 pack/unpack functions.
12579 (UNPACK_IF): Likewise.
12580 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12581 enable long double built-in functions if long double is IEEE
12582 128-bit floating point.
12583 (rs6000_invalid_builtin): Update long double built-in function
12585 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12586 functions, adjust the built-in function to use the long double
12587 built-in function if __ibm128 and long double are the same type.
12588 * doc/extend.texi (PowerPC builtins): Update documention for
12589 __builtin_{,un}pack_longdouble. Add documentation for
12590 __builtin_{,un}pack_ibm128.
12592 2018-06-06 Jim Wilson <jimw@sifive.com>
12594 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12595 (struct machine_function): New field interrupt_mode.
12596 (riscv_handle_type_attribute): New function. Add forward declaration.
12597 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12598 (riscv_expand_epilogue): Check interrupt_mode field.
12599 (riscv_set_current_function): Check interrupt attribute args and
12600 set interrupt_mode field.
12601 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12602 (riscv_sret, riscv_uret): New.
12603 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12604 new arguments to interrupt attribute.
12606 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
12609 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12610 Don't handle -mcpu=power8 if -mpower9-vector is also used.
12612 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12614 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12615 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12616 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12617 several redundant entries.
12619 2018-06-06 David Malcolm <dmalcolm@redhat.com>
12621 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12622 type from "rtx" to "rtx_insn *".
12623 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12624 for local "call_insn", removing cast.
12625 (ix86_expand_call): Likewise, introducing a "call_insn" local.
12627 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
12629 PR tree-optimization/86066
12630 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12631 for BIT_INSERT_EXPR stores.
12633 2018-06-06 Richard Biener <rguenther@suse.de>
12635 PR tree-optimization/86062
12636 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12637 component refs ontop
12638 of to be offsetted base.
12640 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12642 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12643 to be static and remove check on interrupt attribute name.
12645 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
12647 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12648 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12650 2018-06-05 Steve Ellcey <sellcey@cavium.com>
12653 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12655 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12656 Remove second argument, change how error is called.
12657 (aarch64_layout_arg): Remove second argument from
12658 aarch64_err_no_fpadvsimd call.
12659 (aarch64_init_cumulative_args): Ditto.
12660 (aarch64_gimplify_va_arg_expr): Ditto.
12661 * config/aarch64/aarch64.md (mov<mode>): Ditto.
12663 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
12665 * config/i386/i386.md (simple_return_indirect_internal): New expander.
12666 (*simple_return_indirect_internal<mode>): Rename from
12667 simple_return_indirect_internal. Use W mode iterator.
12668 (rstorssp): New expander.
12669 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
12670 (clrssbsy): New expander.
12671 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
12673 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
12675 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12677 (cmse_check_pointed_object): Likewise.
12679 2018-06-05 Martin Liska <mliska@suse.cz>
12681 PR gcov-profile/47618
12682 * doc/invoke.texi: Document how -fprofile-dir format
12685 2018-06-05 Richard Biener <rguenther@suse.de>
12687 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12688 removal pretend DOM info isn't available so we do not update
12689 it and only remove edges, not dominated blocks. Actually free
12690 DOM info in case we removed something. Remove unreachable blocks.
12691 (mfb_keep_latches): Work with either DOM info or marked backedges.
12692 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12693 first. Mark backedges if DOM info isn't available.
12694 (Re-)compute DOM info after cleanup_control_flow_pre.
12696 2018-06-05 Richard Biener <rguenther@suse.de>
12698 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12699 (locus_discrim_hasher::hash): Adjust.
12700 (locus_discrim_hasher::equal): Likewise.
12701 (next_discriminator_for_locus): Work on line directly.
12702 (same_line_p): Pass in expanded locus1 as well.
12703 (assign_discriminators): Avoid redundant location expansions.
12705 2018-06-05 Richard Biener <rguenther@suse.de>
12707 PR tree-optimization/86046
12708 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12709 if required after clearing TREE_ADDRESSABLE.
12711 2018-06-05 Richard Biener <rguenther@suse.de>
12713 PR tree-optimization/86047
12714 * tree-ssa-loop.c (for_each_index): Glob handling of all
12715 decls and constants and really handle all of them.
12717 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12720 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12721 qualifier_void_pointer and qualifier_const_void_pointer.
12722 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12723 (arm_init_builtins): Handle the above.
12724 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12725 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12726 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12729 2018-06-05 Martin Liska <mliska@suse.cz>
12731 * auto-profile.c (read_autofdo_file): Do not use
12732 gcov_ctr_summary struct.
12733 (afdo_callsite_hot_enough_for_early_inline): Likewise.
12734 * coverage.c (struct counts_entry): Likewise.
12735 (read_counts_file): Read just single summary entry.
12736 (get_coverage_counts): Use gcov_summary struct.
12737 * coverage.h (get_coverage_counts): Likewise.
12738 * gcov-dump.c (dump_working_sets): Likewise.
12739 (tag_summary): Dump just single summary.
12740 * gcov-io.c (gcov_write_summary): Write just histogram
12742 (gcov_read_summary): Read just single summary.
12743 (compute_working_sets): Use gcov_summary struct.
12744 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12745 of GCOV_COUNTERS_SUMMABLE.
12746 (GCOV_COUNTERS_SUMMABLE): Remove.
12747 (GCOV_FIRST_VALUE_COUNTER): Replace with
12748 GCOV_COUNTER_V_INTERVAL.
12749 (struct gcov_ctr_summary): Remove.
12750 (struct gcov_summary): Directly use fields of former
12752 (compute_working_sets): Use gcov_summary struct.
12753 * gcov.c (read_count_file): Do not use ctrs fields.
12754 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12756 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12758 * profile.c: Likewise.
12759 * profile.h: Likewise.
12761 2018-06-05 Martin Liska <mliska@suse.cz>
12763 PR gcov-profile/84846
12764 * gcov.c (output_lines): Print working directory only
12765 in intermediate format.
12767 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12769 * config/s390/s390-builtin-types.def: Add void function type.
12770 * config/s390/s390-builtins.def: Use the function type for the
12773 2018-06-04 Jim Wilson <jimw@sifive.com>
12775 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12777 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12778 and maybe_eh_return. Change regno to unsigned int. Use new args to
12779 handle EH_RETURN_DATA_REGNO registers properly.
12780 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12781 (riscv_expand_epilogue): Update comment. Change argument name and
12782 type. Update code to use new name and type. Pass new args to
12783 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12785 * config/riscv/riscv.md (NORMAL_RETURN): New.
12786 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12787 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12788 (eh_return): Call gen_eh_return_internal and emit barrier.
12789 (eh_return_internal): Call riscv_expand_epilogue.
12791 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12793 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12794 bit_insertion field and declare can_be_merged_into method.
12795 (merged_store_group::can_be_merged_into): New method.
12796 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12797 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12798 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12800 2018-06-04 Richard Biener <rguenther@suse.de>
12802 PR tree-optimization/85955
12803 * builtins.c (fold_builtin_sincos): Convert pointers to
12804 destination to appropriate type before dereferencing.
12806 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
12808 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12810 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
12812 * expr.c (expand_expr_real_1): Force the operand into memory if
12813 its TYPE_MODE is BLKmode and if there is no integer mode for
12814 the number of bits being extracted.
12816 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12820 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12821 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12823 2018-06-04 Richard Biener <rguenther@suse.de>
12825 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12826 (cleanup_tree_cfg_noloop): ... single caller. Do
12827 start_recording_case_labels later.
12829 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
12831 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12832 to _IMMINTRIN_H_INCLUDED.
12833 * config/i386/pconfigintrin.h: Ditto.
12834 * config/i386/waitpkgintrin.h: Ditto.
12835 * config/i386/immintrin.h: Add includes for sgxintrin.h,
12836 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12837 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12838 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12839 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12840 waitpkgintrin.h and cldemoteintrin.h.
12842 2018-06-04 Richard Biener <rguenther@suse.de>
12844 PR tree-optimization/86038
12845 * tracer.c (find_best_successor): Check probability for
12846 being initialized, bail out if not.
12848 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
12851 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12852 of bits to ignore when comparing architectures.
12854 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12856 PR tree-optimization/69615
12857 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12858 maximum or minimum of the type, try to merge it also as if
12859 range1 is + [-, x - 1] or + [x + 1, -].
12862 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12864 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
12866 PR tree-optimization/86034
12867 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12868 the unsigned bitfield type in a bit insertion sequence if it does not
12869 have a larger precision than the bitfield size.
12870 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12872 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
12874 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12876 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12878 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12879 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12880 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12881 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12883 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12885 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12886 Disable -fdelete-null-pointer-checks for ELF toolchain.
12888 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12889 Kito Cheng <kito.cheng@gmail.com>
12891 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12892 (nds32le-*-*, nds32be-*-*): Integrate checking process.
12893 (nds32*-*-*): Add glibc and uclibc conditions.
12894 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12895 (TARGET_EXCEPT_UNWIND_INFO): Define.
12896 * config/nds32/elf.h: New file.
12897 * config/nds32/linux.h: New file.
12898 * config/nds32/nds32-elf.opt: New file.
12899 * config/nds32/nds32-linux.opt: New file.
12900 * config/nds32/nds32-fp-as-gp.c
12901 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12902 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12904 (nds32_asm_file_end): Ditto.
12905 (nds32_print_operand): Ditto.
12906 (nds32_insert_attributes): Ditto.
12907 (nds32_init_libfuncs): New function.
12908 (TARGET_HAVE_TLS): Define.
12909 (TARGET_INIT_LIBFUNCS): Define.
12910 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12912 (TARGET_ELF): Apply different mcmodel setting.
12913 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12914 been migrated into elf.h and linux.h files.
12915 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12916 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12917 (mcmodel): The content has been migrated into nds32-elf.opt and
12918 nds32-linux.opt files.
12919 * config/nds32/t-elf: New file.
12920 * config/nds32/t-linux: New file.
12922 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12923 Shiva Chen <shiva0217@gmail.com>
12925 * config/nds32/constants.md (unspec_volatile_element): Add
12926 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12927 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12929 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12930 (make_pass_nds32_fp_as_gp): Declare.
12931 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12933 (nds32_asm_function_end_prologue): Remove unused asm output.
12934 (nds32_asm_function_begin_epilogue): Remove unused asm output.
12935 (nds32_asm_file_start): Output necessary fp_as_gp information.
12936 (nds32_option_override): Adjust register usage.
12937 (nds32_expand_prologue): Consider fp_as_gp situation.
12938 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12939 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12942 (simple_return): Ditto.
12943 (omit_fp_begin): Output special directive for fp_as_gp.
12944 (omit_fp_end): Output special directive for fp_as_gp.
12945 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12946 mforbid-fp-as-gp): New options.
12948 2018-06-01 Mark Wielaard <mark@klomp.org>
12950 * dwarf2out.c (dwarf2out_finish): Remove generation of
12951 DW_AT_loclists_base.
12953 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
12955 * gimple-ssa-store-merging.c: Include gimple-fold.h.
12956 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12957 (struct merged_store_group): Add bit_insertion field.
12958 (dump_char_array): Use standard hexadecimal format.
12959 (merged_store_group::merged_store_group): Set bit_insertion to false.
12960 (merged_store_group::apply_stores): Use optimal buffer size. Deal
12961 with BIT_INSERT_EXPR stores. Move up code updating the mask and
12962 also print the mask in the dump file.
12963 (pass_store_merging::gate): Minor tweak.
12964 (imm_store_chain_info::coalesce_immediate): Fix wrong association
12965 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
12966 stores with INTEGER_CST stores.
12967 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12968 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12969 and use it throughout. Generate bit insertion sequences if need be.
12970 (pass_store_merging::process_store): Remove redundant condition.
12971 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12973 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
12975 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12976 the 128-bit floating point types. Fix function comment.
12978 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12980 * config/aarch64/aarch64-simd.md
12981 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12983 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12986 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12988 PR tree-optimization/85989
12989 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12991 (backprop::intersect_uses): Check it when deciding whether this
12992 is a backedge reference.
12993 (backprop::process_block): Add each phi to m_visited_phis
12994 after visiting it, then clear it at the end.
12996 2018-06-01 Richard Biener <rguenther@suse.de>
12998 * tree-vectorizer.h (vect_dr_stmt): New function.
12999 (vect_get_load_cost): Adjust.
13000 (vect_get_store_cost): Likewise.
13001 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13002 Use vect_dr_stmt instead of DR_SMTT.
13003 (vect_record_base_alignments): Likewise.
13004 (vect_calculate_target_alignment): Likewise.
13005 (vect_compute_data_ref_alignment): Likewise and make static.
13006 (vect_update_misalignment_for_peel): Likewise.
13007 (vect_verify_datarefs_alignment): Likewise.
13008 (vector_alignment_reachable_p): Likewise.
13009 (vect_get_data_access_cost): Likewise. Pass down
13010 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13011 (vect_get_peeling_costs_all_drs): Likewise.
13012 (vect_peeling_hash_get_lowest_cost): Likewise.
13013 (vect_enhance_data_refs_alignment): Likewise.
13014 (vect_find_same_alignment_drs): Likewise.
13015 (vect_analyze_data_refs_alignment): Likewise.
13016 (vect_analyze_group_access_1): Likewise.
13017 (vect_analyze_group_access): Likewise.
13018 (vect_analyze_data_ref_access): Likewise.
13019 (vect_analyze_data_ref_accesses): Likewise.
13020 (vect_vfa_segment_size): Likewise.
13021 (vect_small_gap_p): Likewise.
13022 (vectorizable_with_step_bound_p): Likewise.
13023 (vect_prune_runtime_alias_test_list): Likewise.
13024 (vect_analyze_data_refs): Likewise.
13025 (vect_supportable_dr_alignment): Likewise.
13026 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13027 (vect_gen_prolog_loop_niters): Likewise.
13028 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13029 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13031 (vect_recog_mask_conversion_pattern): Likewise.
13032 (vect_try_gather_scatter_pattern): Likewise.
13033 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13034 to vect_get_store_cost.
13035 (vect_get_store_cost): Get stmt_info instead of DR.
13036 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13037 (vect_get_load_cost): Get stmt_info instead of DR.
13039 2018-06-01 Richard Biener <rguenther@suse.de>
13041 PR middle-end/86017
13042 * gimple-fold.c (var_decl_component_p): Also allow offsetted
13043 vars wrapped in MEM_REFs.
13045 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13047 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13048 Fix subreg tests so that we only return a choice between
13049 GENERAL_REGS and FP_REGS if the original classes included both.
13051 2018-06-01 Richard Biener <rguenther@suse.de>
13054 * tree-ssa-structalias.c (get_function_part_constraint):
13055 Handle NULL fi->decl.
13056 (find_func_aliases_for_call): Properly handle indirect
13057 fi from direct call.
13058 (find_func_clobbers): Likewise.
13059 (ipa_pta_execute): Likewise.
13060 (create_variable_info_for): For functions that are ifunc_resolver
13061 resolve to a varinfo that contains the result of the resolver call.
13062 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13065 2018-05-31 Michael Collison <michael.collison@arm.com>
13067 * config/aarch64/aarch64.md:
13068 (*fix_to_zero_extenddfdi2): New pattern.
13069 * gcc.target/aarch64/fix_extend1.c: New testcase.
13071 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
13073 PR middle-end/78809
13074 PR middle-end/83026
13075 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13076 and BUILT_IN_STRNCMP_EQ.
13077 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13078 BUILT_IN_STRNCMP_EQ.
13079 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13080 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13081 (gimple_fold_builtin): Likewise.
13082 * tree-ssa-strlen.c (compute_string_length): New function.
13083 (determine_min_obsize): New function.
13084 (handle_builtin_string_cmp): New function to handle calls to
13085 string compare functions.
13086 (strlen_optimize_stmt): Add handling to builtin string compare
13088 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13089 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13090 * tree.c (build_common_builtin_nodes): Add new defines of
13091 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13093 2018-05-31 Jakub Jelinek <jakub@redhat.com>
13096 * bb-reorder.c (pass_partition_blocks::gate): Return false for
13097 functions with naked attribute.
13099 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
13101 * config/i386/sse.md (avx_vec_concat<mode>):
13102 Substitute concat_tg_mode mode attribute with xtg_mode.
13103 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13104 (concat_tg_mode): Remove mode attribute.
13106 2018-05-31 Martin Sebor <msebor@redhat.com>
13109 * calls.c (alloc_max_size): Correct a logic error/typo.
13110 Treat excessive arguments as infinite. Warn for invalid arguments.
13111 * doc/invoke.texi (-Walloc-size-larger-than): Update.
13113 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
13116 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13117 and movx for Haswell.
13119 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
13120 Cesar Philippidis <cesar@codesourcery.com>
13122 PR middle-end/85879
13123 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13124 when emitting error on private/firstprivate reductions.
13125 * omp-low.c (lower_omp_target): Avoid reference-type processing
13126 on pointers for firstprivate clause.
13128 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13130 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13133 * config/aarch64/aarch64-simd.md
13134 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13135 (aarch64_ld1_x3_<mode>): Likewise
13136 (aarch64_st1x2<VALLDIF:mode>): Likewise
13137 (aarch64_st1_x2_<mode>): Likewise
13138 (aarch64_st1x3<VALLDIF:mode>): Likewise
13139 (aarch64_st1_x3_<mode>): Likewise
13140 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13141 (vld1_s8_x3): Likewise.
13142 (vld1_u16_x3): Likewise.
13143 (vld1_s16_x3): Likewise.
13144 (vld1_u32_x3): Likewise.
13145 (vld1_s32_x3): Likewise.
13146 (vld1_u64_x3): Likewise.
13147 (vld1_s64_x3): Likewise.
13148 (vld1_f16_x3): Likewise.
13149 (vld1_f32_x3): Likewise.
13150 (vld1_f64_x3): Likewise.
13151 (vld1_p8_x3): Likewise.
13152 (vld1_p16_x3): Likewise.
13153 (vld1_p64_x3): Likewise.
13154 (vld1q_u8_x3): Likewise.
13155 (vld1q_s8_x3): Likewise.
13156 (vld1q_u16_x3): Likewise.
13157 (vld1q_s16_x3): Likewise.
13158 (vld1q_u32_x3): Likewise.
13159 (vld1q_s32_x3): Likewise.
13160 (vld1q_u64_x3): Likewise.
13161 (vld1q_s64_x3): Likewise.
13162 (vld1q_f16_x3): Likewise.
13163 (vld1q_f32_x3): Likewise.
13164 (vld1q_f64_x3): Likewise.
13165 (vld1q_p8_x3): Likewise.
13166 (vld1q_p16_x3): Likewise.
13167 (vld1q_p64_x3): Likewise.
13168 (vst1_s64_x2): Likewise.
13169 (vst1_u64_x2): Likewise.
13170 (vst1_f64_x2): Likewise.
13171 (vst1_s8_x2): Likewise.
13172 (vst1_p8_x2): Likewise.
13173 (vst1_s16_x2): Likewise.
13174 (vst1_p16_x2): Likewise.
13175 (vst1_s32_x2): Likewise.
13176 (vst1_u8_x2): Likewise.
13177 (vst1_u16_x2): Likewise.
13178 (vst1_u32_x2): Likewise.
13179 (vst1_f16_x2): Likewise.
13180 (vst1_f32_x2): Likewise.
13181 (vst1_p64_x2): Likewise.
13182 (vst1q_s8_x2): Likewise.
13183 (vst1q_p8_x2): Likewise.
13184 (vst1q_s16_x2): Likewise.
13185 (vst1q_p16_x2): Likewise.
13186 (vst1q_s32_x2): Likewise.
13187 (vst1q_s64_x2): Likewise.
13188 (vst1q_u8_x2): Likewise.
13189 (vst1q_u16_x2): Likewise.
13190 (vst1q_u32_x2): Likewise.
13191 (vst1q_u64_x2): Likewise.
13192 (vst1q_f16_x2): Likewise.
13193 (vst1q_f32_x2): Likewise.
13194 (vst1q_f64_x2): Likewise.
13195 (vst1q_p64_x2): Likewise.
13196 (vst1_s64_x3): Likewise.
13197 (vst1_u64_x3): Likewise.
13198 (vst1_f64_x3): Likewise.
13199 (vst1_s8_x3): Likewise.
13200 (vst1_p8_x3): Likewise.
13201 (vst1_s16_x3): Likewise.
13202 (vst1_p16_x3): Likewise.
13203 (vst1_s32_x3): Likewise.
13204 (vst1_u8_x3): Likewise.
13205 (vst1_u16_x3): Likewise.
13206 (vst1_u32_x3): Likewise.
13207 (vst1_f16_x3): Likewise.
13208 (vst1_f32_x3): Likewise.
13209 (vst1_p64_x3): Likewise.
13210 (vst1q_s8_x3): Likewise.
13211 (vst1q_p8_x3): Likewise.
13212 (vst1q_s16_x3): Likewise.
13213 (vst1q_p16_x3): Likewise.
13214 (vst1q_s32_x3): Likewise.
13215 (vst1q_s64_x3): Likewise.
13216 (vst1q_u8_x3): Likewise.
13217 (vst1q_u16_x3): Likewise.
13218 (vst1q_u32_x3): Likewise.
13219 (vst1q_u64_x3): Likewise.
13220 (vst1q_f16_x3): Likewise.
13221 (vst1q_f32_x3): Likewise.
13222 (vst1q_f64_x3): Likewise.
13223 (vst1q_p64_x3): Likewise.
13225 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13227 * config/msp430/msp430.c (msp430_output_labelref): Prepend
13228 user_label_prefix to name.
13230 * tree-core.h: Update comment about the format of NAME string
13231 passed to handler in attribute_spec.
13233 * config/msp430/msp430.md: Remove erroneous subreg expression from
13234 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13235 zero_extend{q,h}isi2.
13237 2018-05-30 Borislav Petkov <bp@suse.de>
13239 * doc/extend.texi: Document some architecture specific
13240 constraints and sort entries.
13242 2018-05-30 Martin Sebor <msebor@redhat.com>
13244 PR middle-end/85369
13245 * builtins.c (expand_builtin_stpcpy_1): New function.
13246 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13247 only if the former succeeds.
13249 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13251 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13254 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13256 * doc/invoke.texi (-flinker-output): Document
13258 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13260 * passes.c (ipa_write_summaries): Only modify statements if body
13262 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13263 incrementally linking.
13264 (ipa_passes): Likewise.
13265 * lto-cgraph.c (lto_output_node): When incrementally linking do not
13266 pass down resolution info.
13267 * common.opt (flag_incremental_link): Update info.
13268 * gcc.c (plugin specs): Turn flinker-output=* to
13269 -plugin-opt=-linker-output-known
13270 * toplev.c (compile_file): Also cut compilation when doing incremental
13272 * flag-types. (enum lto_partition_model): Add
13273 LTO_LINKER_OUTPUT_NOLTOREL.
13274 (invoke.texi): Add -flinker-output docs.
13275 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13276 link same way as WPA; do not stream in dead initializers.
13278 * dwarf2out.c (dwarf2out_die_ref_for_decl,
13279 darf2out_register_external_decl): Support incremental link.
13281 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13283 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13285 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13287 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13288 it down to simple_object_copy_lto_debug_sections.
13289 (run_gcc): Determine incremental LTO link time and configure
13290 lto1 into non-wpa mode, disable renaming of debug sections.
13292 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
13294 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13295 descriptions of various incorrectly documented functions.
13297 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
13300 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13301 address check not strict.
13303 2018-05-30 Richard Biener <rguenther@suse.de>
13305 PR tree-optimization/85964
13306 * tracer.c (better_p): Drop initialized count check, we only
13307 call the function with initialized counts now.
13308 (find_best_successor): Do find a best edge if one
13309 has uninitialized count.
13310 (find_best_predecessor): Likewise. Do BB frequency check only
13311 if count is initialized.
13313 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
13315 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13316 (aarch64_ldrstr_offset_compare): New.
13317 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13318 load/store orderings.
13319 (aarch64_gen_adjusted_ldpstp): Likewise.
13321 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
13323 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13324 Check for subset of GENERAL_REGS and FP_REGS.
13325 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13328 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
13330 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13331 and wi::to_poly_offset. Add the current offset and then check
13332 whether the sum fits, rather than using an unchecked addition of
13333 a checked term. Check for a shwi rather than a uhwi.
13334 * expr.c (get_bit_range): Use tree_to_poly_uint64.
13335 (store_constructor): Use poly_int_tree_p.
13336 (expand_expr_real_1): Likewise.
13337 * function.c (assign_temp): Likewise.
13338 * fold-const.c (const_binop): Use poly_int_tree_p and
13339 wi::to_poly_offset.
13340 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
13342 * ipa-icf-gimple.c (func_checker::compare_operand): Use
13343 to_poly_offset for MEM offsets.
13344 * ipa-icf.c (sem_variable::equals): Likewise.
13345 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13346 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13347 wi::to_poly_offset for BIT_FIELD_REF offsets.
13348 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13349 wi::to_poly_offset.
13350 * var-tracking.c (emit_note_insn_var_location): Use
13351 tree_to_poly_uint64.
13353 2018-05-29 Jim Wilson <jimw@sifive.com>
13355 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13357 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
13360 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13361 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13363 (sse4_1_round<mode>2): Use nonimmediate_operand
13364 for operand 1 predicate.
13366 2018-05-29 Martin Sebor <msebor@redhat.com>
13367 Richard Biener <rguenther@suse.de>
13370 * calls.c (get_size_range): Call determine_value_range instead
13371 of get_value_range..
13372 * tree-vrp.h (determine_value_range): Declared new function.
13373 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13375 2018-05-29 Richard Biener <rguenther@suse.de>
13377 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13378 sure to use non-pattern stmts for get_earlier_stmt arguments.
13379 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13380 called on pattern stmts.
13381 (get_later_stmt): Likewise.
13383 2018-05-29 Martin Liska <mliska@suse.cz>
13385 PR gcov-profile/85759
13386 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13389 2018-05-29 Jakub Jelinek <jakub@redhat.com>
13391 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13393 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13397 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13398 VEC_PACK_FLOAT_EXPR): New tree codes.
13399 * tree-pretty-print.c (op_code_prio): Handle
13400 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13401 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13402 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13403 * tree-inline.c (estimate_operator_cost): Likewise.
13404 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13405 * fold-const.c (const_binop): Likewise.
13406 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13407 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13408 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13409 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13410 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13411 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13412 * expr.c (expand_expr_real_2): Likewise.
13413 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13414 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13415 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13417 * optabs.c (expand_widen_pattern_expr): For
13418 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13419 sign from result type rather than operand's type.
13420 (expand_binop_directly): For vec_packu_float_optab and
13421 vec_packs_float_optab allow result type to be different from operand's
13423 * optabs-tree.c (optab_for_tree_code): Handle
13424 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13425 VEC_PACK_FLOAT_EXPR. Formatting fixes.
13426 * tree-vect-generic.c (expand_vector_operations_1): Handle
13427 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13428 VEC_PACK_FLOAT_EXPR.
13429 * tree-vect-stmts.c (supportable_widening_operation): Handle
13431 (supportable_narrowing_operation): Handle FLOAT_EXPR.
13432 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13433 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13434 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
13435 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13437 (vec_pack<floatprefix>_float_<mode>): New expander.
13438 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13440 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13441 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13442 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13443 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13444 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13446 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13447 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13448 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13449 VEC_PACK_FLOAT_EXPR): Document.
13451 2018-05-29 Richard Biener <rguenther@suse.de>
13453 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13455 (stmt_vec_info_vec): Make pointer.
13456 (init_stmt_vec_info_vec): Remove.
13457 (free_stmt_vec_info_vec): Likewise.
13458 (set_stmt_vec_info_vec): New function.
13459 (free_stmt_vec_infos): Likewise.
13460 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13461 (set_vinfo_for_stmt): Likewise.
13462 (get_earlier_stmt): Likewise.
13463 (get_later_stmt): Likewise.
13464 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13465 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13466 (vec_info::~vec_info): Free stmt_vec_infos.
13467 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13468 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13469 (pass_slp_vectorize::execute): Likewise.
13470 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13471 (free_stmt_vec_info_vec): Likewise.
13472 (set_stmt_vec_info_vec): New function.
13473 (free_stmt_vec_infos): Likewise.
13474 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13475 the global stmt_vec_info_vec.
13476 * tree-parloops.c (gather_scalar_reductions): Use
13477 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13480 2018-05-29 Richard Biener <rguenther@suse.de>
13482 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13484 2018-05-29 Martin Liska <mliska@suse.cz>
13485 David Malcolm <dmalcolm@redhat.com>
13487 * vec.c (test_reverse): New.
13488 (vec_c_tests): Add new test.
13489 * vec.h (vl_ptr>::reverse): New function.
13491 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
13493 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13495 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13498 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13500 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13502 2018-05-28 Richard Biener <rguenther@suse.de>
13504 PR tree-optimization/85933
13505 * tree-vect-data-refs.c (vect_record_base_alignments): Only
13506 look at stmts marked as vectorizable.
13508 2018-05-28 Richard Biener <rguenther@suse.de>
13510 PR tree-optimization/85934
13511 * tree-vect-generic.c (expand_vector_operations_1): Hoist
13512 vector boolean check before scalar optimization.
13514 2018-05-28 Jakub Jelinek <jakub@redhat.com>
13516 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13519 2018-05-28 Mark Wielaard <mark@klomp.org>
13521 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13522 if it is an expression containing a minus sign.
13524 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
13526 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13528 2018-05-27 Paul Koning <ni1d@arrl.net>
13530 * config/pdp11/pdp11.md (truncsihi2): Remove.
13532 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
13533 Chung-Ju Wu <jasonwucj@gmail.com>
13535 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13537 (unaligned_store_dw): Ditto.
13538 * config/nds32/nds32-memory-manipulation.c
13539 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13540 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13541 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13542 (emit_setmem_word_loop): Rename to ...
13543 (emit_setmem_doubleword_loop): ... this.
13544 (nds32_gen_dup_4_byte_to_word_value): New function.
13545 (nds32_gen_dup_8_byte_to_double_word_value): New function.
13546 (nds32_expand_setmem_loop): Refine implementation.
13547 (nds32_expand_setmem_loop_v3m): Ditto.
13548 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13551 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13553 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13555 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13557 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13558 (nds32_init_machine_status): Initialize machine->attr_naked_p and
13559 machine->attr_no_prologue_p.
13560 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13561 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13562 (nds32_expand_epilogue): Consider attr_naked_p.
13563 (nds32_expand_epilogue_v3pop): Likewise.
13564 (nds32_can_use_return_insn): Likewise.
13565 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13566 attr_no_prologue_p fields.
13567 * config/nds32/nds32.opt (mret-in-naked-func): New option.
13569 2018-05-27 Jakub Jelinek <jakub@redhat.com>
13572 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13574 * config/i386/sse.md
13575 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13577 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13579 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13581 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13583 (*<floatsuffix>floatv2div2sf2): Rename to ...
13584 (*float<floatunssuffix>v2div2sf2): ... this.
13585 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13586 (float<floatunssuffix>v2div2sf2_mask): ... this.
13587 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13588 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13589 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13591 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13593 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13595 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13597 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13599 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13601 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13602 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13603 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13604 gen_ufix_truncv8dfv8si2.
13605 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13606 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13607 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13608 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13609 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13610 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13611 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13612 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13614 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13618 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13620 2018-05-25 Jim Wilson <jimw@sifive.com>
13622 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13623 * config/riscv/riscv.c (struct machine_function): Add
13624 interrupt_handler_p and attribute_checked_p fields.
13625 (riscv_attribute_table): Add interrupt.
13626 (riscv_interrupt_type_p): New.
13627 (riscv_save_reg_p): Save extra regs for interrupt handler.
13628 (riscv_use_save_libcall): Return false for interrupt handler.
13629 (riscv_first_stack_step): Add forward declaration.
13630 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
13631 for interrupt handler with large frame. Use it for saved reg list.
13632 (riscv_expand_prologue): Move flag_stack_usage_info support to
13633 eliminate duplication.
13634 (riscv_expand_epilogue): Generate mret for interrupt handler.
13635 (riscv_epilogue_uses): New.
13636 (riscv_can_use_return_insn): Return false for interrupt handler.
13637 (riscv_function_ok_for_sibcall): Likewise.
13638 (riscv_set_current_function): Add interrupt handler support.
13639 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13640 * config/riscv/riscv.md (UNSPECV_MRET): New.
13642 (riscv_frflags, riscv_fsflags): Use tab after opcode.
13644 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13646 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
13648 PR tree-optimization/85712
13649 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13650 this candidate has already been replaced in-situ by a copy.
13652 2018-05-25 Jason Merrill <jason@redhat.com>
13654 PR c++/80485 - inline function non-zero address.
13655 * symtab.c (nonzero_address): Check DECL_COMDAT.
13657 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
13660 * config/alpha/alpha.md (ashlsi3): New insn pattern.
13661 (*ashlsi_se): Rename from *ashldi_se. Define as sign
13662 extension of SImode operation. Use const123_operand predicate.
13663 (*saddsi_1): Remove.
13664 (*saddl_se_1): Ditto.
13665 (*ssubsi_1): Ditto.
13666 (*ssubl_se_1): Ditto.
13667 * config/alpha/predicates.md (const123_operand): New predicate.
13668 * config/alpha/constraints.md (P): Use IN_RANGE.
13670 2018-05-25 Richard Biener <rguenther@suse.de>
13672 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13674 (ref_maybe_used_by_stmt_p): Likewise.
13675 (stmt_may_clobber_ref_p): Likewise.
13676 (stmt_may_clobber_ref_p_1): Likewise.
13677 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13679 (ref_maybe_used_by_stmt_p): Likewise.
13680 (stmt_may_clobber_ref_p): Likewise.
13681 (stmt_may_clobber_ref_p_1): Likewise.
13682 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13683 the alias oracle to disambiguate DRs with stmts DR analysis
13685 (vect_analyze_data_refs): Do not give up on not analyzable
13686 DRs for BB vectorization. Remove code truncating the dataref
13689 2018-05-25 Jakub Jelinek <jakub@redhat.com>
13692 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13693 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
13694 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13696 2018-05-25 Richard Biener <rguenther@suse.de>
13698 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13699 function, combining stmt data ref gathering and fatal analysis
13701 (vect_analyze_data_refs): Remove now redudnant code and simplify.
13702 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13703 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13704 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13705 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13707 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13709 PR tree-optimization/85720
13710 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13711 SCC if all partitions are builtins.
13712 (version_loop_by_alias_check): New parameter. Generate cancelable
13713 runtime alias check if all partitions are builtins.
13714 (distribute_loop): Update call to above function.
13716 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13718 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13719 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13720 (parm_default_def_partition_arg): Ditto.
13721 (set_parm_default_def_partition): Ditto.
13722 (get_parm_default_def_partitions): Ditto and make it static.
13723 (get_undefined_value_partitions): Ditto and make it static.
13724 (remove_ssa_form): Refactor call to init_var_map here.
13725 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13726 computation for loop region.
13727 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13728 (register_default_def): Delete.
13729 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13730 (parm_default_def_partition_arg): Ditto.
13731 (set_parm_default_def_partition): Ditto.
13732 (get_parm_default_def_partitions): Ditto and make it static.
13733 (get_undefined_value_partitions): Ditto and make it static.
13734 (coalesce_with_default, coalesce_with_default): Update comment.
13735 (create_coalesce_list_for_region): New func factored out from
13736 create_outofssa_var_map.
13737 (populate_coalesce_list_for_outofssa): New func factored out from
13738 create_outofssa_var_map and coalesce_ssa_name.
13739 (create_outofssa_var_map): Delete.
13740 (coalesce_ssa_name): Refactor to support live range computation.
13741 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13742 (get_parm_default_def_partitions): Delete.
13743 (get_undefined_value_partitions): Ditto.
13744 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13745 computation for loop region.
13746 (new_tree_live_info, loe_visit_block): Ditto.
13747 (live_worklist, set_var_live_on_entry): Ditto.
13748 (calculate_live_on_exit, verify_live_on_entry): Ditto.
13749 * tree-ssa-live.h (struct _var_map): New fields.
13750 (init_var_map): Change decl.
13751 (region_contains_p): New.
13753 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13755 * tree-ssa-live.h (live_merge_and_clear): Delete.
13757 2018-05-25 Richard Biener <rguenther@suse.de>
13760 * tree-dump.c (dequeue_and_dump): Remove access to removed
13761 operand 2 of a SWITCH_EXPR.
13763 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13765 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13766 multiplication and division.
13767 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13768 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13769 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13770 (cond_udiv_optab, cond_umod_optab): New optabs.
13771 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13772 (IFN_COND_RDIV): New internal functions.
13773 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13774 TRUNC_MOD_EXPR and RDIV_EXPR.
13775 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13776 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13778 (SVE_INT_BINARY): Include mult.
13779 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13780 (optab, sve_int_op): Handle mult.
13781 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13783 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13784 for SVE_INT_BINARY_SD.
13786 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13788 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13789 (optab, sve_int_op): Handle div and udiv.
13790 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13791 for SVE_INT_BINARY_SD.
13792 (*<optab><mode>3): New insn for the same.
13794 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13796 * tree-vect-patterns.c: Include predict.h.
13797 (vect_recog_divmod_pattern): Restrict check for division support
13798 to when optimizing for size.
13800 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13802 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13803 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13804 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13805 (gimple_match_op::set_op): Likewise.
13806 (gimple_resimplify4): Declare.
13807 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13808 (expr::gen_transform): Likewise.
13809 (decision_tree::gen): Generate a simplification routine for 4 operands.
13810 * gimple-match-head.c (gimple_simplify): Add an overload for
13811 4 operands. In the top-level function, handle up to 4 call
13812 arguments and call gimple_resimplify4.
13813 (gimple_resimplify4): New function.
13814 (build_call_internal): Pass a fourth operand.
13815 (maybe_push_to_seq): Likewise.
13816 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13817 Fold VEC_COND_EXPRs of an operation and a default value into
13818 an IFN_COND_* function if possible.
13819 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13821 (SVE_COND_FP_BINARY): Include them.
13822 (optab, sve_fp_op): Handle them.
13823 (SVE_INT_BINARY_REV): New code iterator.
13824 (SVE_COND_FP_BINARY_REV): New int iterator.
13825 (commutative): New int attribute.
13826 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13828 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13830 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13831 (*cond_<optab><mode>): New patterns for reversed operands.
13833 2018-05-25 Richard Biener <rguenther@suse.de>
13835 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13836 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13837 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13838 (STMT_VINFO_GROUPED_ACCESS): Adjust.
13839 * tree-vect-data-refs.c (everywhere): Adjust users.
13840 * tree-vect-loop.c (everywhere): Likewise.
13841 * tree-vect-slp.c (everywhere): Likewise.
13842 * tree-vect-stmts.c (everywhere): Likewise.
13843 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13845 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13847 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13849 (gcc_cv_as_section_exclude): ... this.
13850 Try Solaris as #exclude syntax.
13851 * configure: Regenerate.
13852 * config.in: Regenerate.
13853 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13855 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13856 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13858 * varasm.c (default_elf_asm_named_section): Don't check if
13859 HAVE_GAS_SECTION_EXCLUDE is defined.
13861 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13863 * doc/md.texi: Update the documentation of the cond_* optabs
13864 to mention the new final operand. Fix GET_MODE_NUNITS call.
13865 Describe the scalar case too.
13866 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13867 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13869 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13870 (get_conditional_internal_fn): Update comment.
13871 * tree-vect-loop.c (vectorizable_reduction): Pass the original
13872 accumulator value as a final argument to conditional functions.
13873 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13874 a define_expand and add an "else" operand. Assert for now that
13875 the else operand is equal to operand 2. Use SVE_INT_BINARY and
13876 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13877 (*cond_<optab><mode>): New patterns.
13878 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13879 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13880 (UNSPEC_COND_EOR): Delete.
13881 (optab): Remove associated mappings.
13882 (SVE_INT_BINARY): New code iterator.
13883 (sve_int_op): Remove int attribute and add "minus" to the code
13885 (SVE_COND_INT_OP): Delete.
13886 (SVE_COND_FP_OP): Rename to...
13887 (SVE_COND_FP_BINARY): ...this.
13889 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13891 * optabs.c (can_reuse_operands_p): New function.
13892 (maybe_legitimize_operands): Try to reuse the results for
13895 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13897 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13898 Add {q} suffix to insn mnemonic.
13900 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13902 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13903 (msp430_warn_func_return): New.
13905 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
13907 * fold-const.c (tree_nonzero_bits): New function.
13908 * fold-const.h (tree_nonzero_bits): Likewise.
13909 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13910 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13912 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13916 * varasm.c (assemble_alias): Check ifunc_resolver only on
13919 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13922 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13923 when memory input operand is handled.
13925 2018-05-24 Luis Machado <luis.machado@linaro.org>
13927 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13929 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13931 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13933 * match.pd: Delay FMA folds until after vectorization.
13935 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
13938 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13939 address check not strict.
13941 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13943 * gimple-match.h (gimple_match_op): New class.
13944 (mprts_hook): Replace parameters with a gimple_match_op *.
13945 (maybe_build_generic_op): Likewise.
13946 (gimple_simplified_result_is_gimple_val): Replace parameters with
13947 a const gimple_match_op *.
13948 (gimple_simplify): Replace code_helper * and tree * parameters with
13949 a gimple_match_op * parameter.
13950 (gimple_resimplify1): Replace code_helper *, tree and tree *
13951 parameters with a gimple_match_op * parameter.
13952 (gimple_resimplify2): Likewise.
13953 (gimple_resimplify3): Likewise.
13954 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13955 parameters with a gimple_match_op * parameter.
13956 * gimple-match-head.c (gimple_simplify): Change prototypes of
13957 auto-generated functions to take a gimple_match_op * instead of
13958 separate code_helper * and tree * parameters. Make the same
13959 change in the top-level overload and update calls to the
13960 gimple_resimplify routines. Update calls to the auto-generated
13961 functions and to maybe_push_res_to_seq in the publicly-facing
13962 operation-specific gimple_simplify overloads.
13963 (gimple_match_op::MAX_NUM_OPS): Define.
13964 (gimple_resimplify1): Replace rcode and ops with a single res_op
13965 parameter. Update call to gimple_simplify.
13966 (gimple_resimplify2): Likewise.
13967 (gimple_resimplify3): Likewise.
13968 (mprts_hook): Replace parameters with a gimple_match_op *.
13969 (maybe_build_generic_op): Likewise.
13970 (build_call_internal): Replace type, nargs and ops with
13971 a gimple_match_op *.
13972 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13973 with a single gimple_match_op *. Update calls to mprts_hook,
13974 build_call_internal and gimple_simplified_result_is_gimple_val.
13975 Factor out code that is common to the tree_code and combined_fn cases.
13976 * genmatch.c (expr::gen_transform): Replace tem_code and
13977 tem_ops with a gimple_match_op called tem_op. Update calls
13978 to the gimple_resimplify functions and maybe_push_res_to_seq.
13979 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13980 res_ops. Update call to the gimple_resimplify functions.
13981 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13982 (decision_tree::gen): Make the functions take a gimple_match_op *
13983 called res_op instead of separate res_code and res_ops parameters.
13984 Update call accordingly.
13985 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13986 and ops with a single res_op parameter. Update calls to
13987 maybe_build_generic_op and maybe_push_res_to_seq.
13988 (fold_stmt_1): Update calls to gimple_simplify and
13989 replace_stmt_with_simplification.
13990 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13991 and gimple_simplified_result_is_gimple_val.
13992 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13994 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13995 with a gimple_match_op *.
13996 (vn_nary_build_or_lookup): Likewise. Update call to
13997 vn_nary_build_or_lookup_1.
13998 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13999 gimple_match_op *. Update calls to the gimple_resimplify routines
14000 and to gimple_simplified_result_is_gimple_val.
14001 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14002 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14003 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14004 (visit_nary_op): Likewise.
14005 (visit_reference_op_load): Likewise.
14007 2018-05-23 Luis Machado <luis.machado@linaro.org>
14009 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14010 modifier for printing the step amount.
14012 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
14015 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14018 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
14020 * doc/sourcebuild.texi (Endianness): New subsubsection.
14022 2018-05-23 Luis Machado <luis.machado@linaro.org>
14024 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14025 <prefetch_dynamic_strides>: New const bool field.
14026 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14027 prefetch_dynamic_strides.
14028 (exynosm1_prefetch_tune): Likewise.
14029 (thunderxt88_prefetch_tune): Likewise.
14030 (thunderx_prefetch_tune): Likewise.
14031 (thunderx2t99_prefetch_tune): Likewise.
14032 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14034 (aarch64_override_options_internal): Update to set
14035 PARAM_PREFETCH_DYNAMIC_STRIDES.
14036 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14037 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14038 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14039 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14040 prefetch-dynamic-strides setting.
14042 2018-05-23 Luis Machado <luis.machado@linaro.org>
14044 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14045 <minimum_stride>: New const int field.
14046 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14047 minimum_stride field defaulting to -1.
14048 (exynosm1_prefetch_tune): Likewise.
14049 (thunderxt88_prefetch_tune): Likewise.
14050 (thunderx_prefetch_tune): Likewise.
14051 (thunderx2t99_prefetch_tune): Likewise.
14052 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14053 <default_opt_level>: Set to 3.
14054 (aarch64_override_options_internal): Update to set
14055 PARAM_PREFETCH_MINIMUM_STRIDE.
14056 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14057 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14058 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14059 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14060 stride is constant and is below the minimum stride threshold.
14062 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14064 * config/arm/arm-cpus.in (mode26): Delete.
14065 (armv4): Delete mode26 reference.
14066 * config/arm/arm.c (arm_configure_build_target): Delete use of
14069 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14071 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14073 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14074 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
14076 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
14077 (floatunsdidf2): Ditto.
14079 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14081 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14082 (fixuns_trunc<mode>si2_avx512f): Ditto.
14083 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14084 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14085 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14087 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14089 PR rtl-optimization/79985
14090 * df-scan.c (df_insn_refs_collect): Remove special case for
14091 global registers and asm statements.
14093 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14095 * extend.texi (Global Register Variables): Rewrite the bullet list.
14096 Note that the register is available for allocation. Note that access
14097 via inline asm must use constraints. Add note about async-signal
14098 handlers. Remove paragraph about automagic register selection.
14100 2018-05-23 Richard Biener <rguenther@suse.de>
14102 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14103 of fixed offset from memset VN.
14105 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
14107 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14108 first_interp field.
14109 (alloc_cand_and_find_basis): Initialize first_interp field.
14110 (slsr_process_mul): Modify first_interp field.
14111 (slsr_process_add): Likewise.
14112 (slsr_process_cast): Modify first_interp field for each new
14114 (slsr_process_copy): Likewise.
14115 (dump_candidate): Dump first_interp field.
14116 (replace_mult_candidate): Process all interpretations, not just
14118 (replace_rhs_if_not_dup): Likewise.
14119 (replace_one_candidate): Likewise.
14121 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
14123 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14125 (aarch64_needs_frame_chain): New function.
14126 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14128 2018-05-23 Sudakshina Das <sudi.das@arm.com>
14131 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14132 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14133 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14134 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14135 as true for strict-align.
14136 (aarch64_can_inline_p): Perform checks even when callee has no
14137 attributes to check for strict alignment.
14138 * doc/extend.texi (AArch64 Function Attributes): Document
14140 * doc/invoke.texi: (AArch64 Options): Likewise.
14142 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
14144 PR tree-optimization/85853
14145 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14146 the handling of the root of the node to...
14147 (vect_slp_analyze_node_operations_1): ...this new function,
14148 and run the whole thing with the child nodes' def types
14149 set according to their SLP node's def type.
14151 2018-05-23 Richard Biener <rguenther@suse.de>
14153 PR middle-end/85874
14154 * tree-data-ref.c (create_runtime_alias_checks): Defer
14155 and ignore overflow warnings.
14157 2018-05-23 Yury Gribov <tetra2005@gmail.com>
14159 PR tree-optimization/85822
14160 * tree-vrp.c (is_masked_range_test): Fix handling of negative
14163 2018-05-23 Richard Biener <rguenther@suse.de>
14165 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14166 memset constants via native_interpret_expr.
14168 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
14171 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14173 (cgraph_node::create_alias): Likewise.
14174 (cgraph_node::get_availability): Check ifunc_resolver instead
14175 of looking up ifunc attribute.
14176 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14177 * varasm.c (do_assemble_alias): Likewise.
14178 (assemble_alias): Likewise.
14179 (default_binds_local_p_3): Likewise.
14180 * cgraph.h (cgraph_node): Add ifunc_resolver.
14181 (cgraph_node::only_called_directly_or_aliased_p): Return false
14182 for IFUNC resolver.
14183 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14185 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14186 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14187 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14188 instead of looking up ifunc attribute.
14190 2018-05-22 Luis Machado <luis.machado@linaro.org>
14192 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14194 2018-05-22 Martin Sebor <msebor@redhat.com>
14196 PR middle-end/85359
14197 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14198 only when expasion succeeds.
14199 (expand_builtin_strcmp): Same.
14200 (expand_builtin_strncmp): Same.
14202 2018-05-22 Martin Sebor <msebor@redhat.com>
14204 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14206 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14207 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14209 * config/aarch64/aarch64-ldpstp.md: Replace uses of
14210 aarch64_mem_pair_operand with memory_operand and delete operand swapping
14212 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14213 Add check for legitimate_address.
14214 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14215 (aarch64_swap_ldrstr_operands): New.
14216 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14219 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14220 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14222 * config/aarch64/aarch64.md: New patterns to generate stp
14224 (store_pair_sw, store_pair_dw): New patterns to generate stp for
14225 single words and double words.
14226 (load_pair_sw, load_pair_dw): Likewise.
14227 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14229 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14231 * config/aarch64/aarch64-ldpstp.md: Modify peephole
14232 for different mode ldpstp and add peephole for merged zero stores.
14233 Likewise for loads.
14234 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14236 (aarch64_gen_store_pair): Rename calls to match new patterns.
14237 (aarch64_gen_load_pair): Rename calls to match new patterns.
14238 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14239 (load_pair<DREG:mode><DREG2:mode>): ... This.
14240 (store_pair<mode>): Rename to...
14241 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14242 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14243 New mode iterators.
14244 (V_INT_EQUIV): Handle SImode.
14245 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14248 2018-05-22 Martin Sebor <msebor@redhat.com>
14251 * calls.c (maybe_warn_nonstring_arg): Use string length to set
14252 or ajust the presumed bound on an operation to avoid unnecessary
14255 2018-05-22 Martin Sebor <msebor@redhat.com>
14257 PR tree-optimization/85826
14258 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14259 assuming that a DECL necesarily has a constant size.
14261 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
14263 PR middle-end/85862
14264 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14266 2018-05-22 Richard Biener <rguenther@suse.de>
14268 PR tree-optimization/85834
14269 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14270 non-constant and non-zero memset arguments.
14272 2018-05-22 Martin Liska <mliska@suse.cz>
14275 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14277 2018-05-22 Richard Biener <rguenther@suse.de>
14279 PR tree-optimization/85863
14280 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14281 comparisons when vectype is specified.
14282 (vectorizable_condition): Do not specify vectype for
14283 vect_is_simple_cond when SLP vectorizing.
14285 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
14288 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14289 define __ibm128 as long double.
14290 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14291 as a distinct type when IEEE 128-bit support is enabled.
14292 (init_float128_ieee): Fix up conversions between IFmode and IEEE
14293 128-bit types to use the correct functions.
14294 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14295 convert between 128-bit floating point types that have different
14296 modes but the same representation, instead of using gen_lowpart to
14298 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14300 (IFKF_reg): New attributes to give the register constraints for
14302 (extend<mode>tf2_internal): New insns to mark an explicit
14303 conversion between 128-bit floating point types that have a
14304 different mode but share the same representation.
14306 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
14308 PR tree-optimization/85814
14309 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14310 a null return from get_strinfo when unsharing the next
14311 strinfo in the chain.
14313 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14316 * varasm.c (weak_finish): Clean up weak_decls.
14318 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14320 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14321 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14322 UNSPEC_UADALP values.
14323 * config/aarch64/iterators.md (ABAL): New int iterator.
14326 (sur): Add mappings for the above.
14327 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14329 (aarch64_<sur>abal<mode>_4): Likewise.
14330 (aarch64_<sur>adalp<mode>_3): Likewise.
14331 (<sur>sadv16qi): New define_expand.
14333 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
14335 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14336 (*movdf_internal): Ditto.
14337 (*rcpsf2_sse): Ditto.
14338 (*rsqrtsf2_sse): Ditto.
14339 (*sqrt<mode>2_sse): Ditto.
14341 2018-05-21 Tamar Christina <tamar.christina@arm.com>
14343 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14345 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14346 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14347 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14348 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14350 * config/aarch64/arm_neon.h: Likewise.
14351 * config/aarch64/iterators.md (VQ_I): New.
14353 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
14355 * config.gcc: Add arc/t-multilib-linux to tmake_file for
14357 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14360 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14362 * config/nds32/constraints.md (S): New constraint.
14363 * config/nds32/nds32.md (call_internal): Use constraint S.
14364 (call_value_internal): Likewise.
14365 (sibcall_internal): Likewise.
14366 (sibcall_value_internal): Likewise.
14368 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14369 Chung-Ju Wu <jasonwucj@gmail.com>
14371 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14372 into consideration.
14374 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14375 Chung-Ju Wu <jasonwucj@gmail.com>
14377 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14378 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14379 (nds32_rtx_costs_impl): Simplify.
14380 (nds32_address_cost_impl): Simplify.
14381 (nds32_init_rtx_costs): New function.
14382 (nds32_rtx_costs_speed_prefer): Likewise.
14383 (nds32_rtx_costs_size_prefer): Likewise.
14384 (nds32_address_cost_speed_prefer): Likewise.
14385 (nds32_address_cost_speed_fwprop): Likewise.
14386 (nds32_address_cost_size_prefer): Likewise.
14387 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14388 * config/nds32/nds32.c (nds32_option_override): Use
14389 nds32_init_rtx_costs function.
14391 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14393 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14394 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14395 (TARGET_PIPELINE_N8): Likewise.
14396 (TARGET_PIPELINE_N10): Likewise.
14397 (TARGET_PIPELINE_N13): Likewise.
14398 (TARGET_PIPELINE_GRAYWOLF): Likewise.
14400 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
14402 * config/nds32/nds32-fpu.md: Update copyright year.
14404 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14406 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14408 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14410 * config/nds32/nds32.c
14411 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14412 * config/nds32/nds32.opt (minline-asm-r15): New option.
14414 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14416 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14418 * config/nds32/nds32.md (abssi2): New pattern.
14420 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
14422 * config/i386/i386.md (rex64namesuffix): New mode attribute.
14423 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14424 Merge insn pattern from sse_cvtsi2ss<round_name> and
14425 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14426 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14427 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14428 using SWI48 mode iterator.
14429 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14430 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14431 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14432 pattern from sse_cvttss2si<round_saeonly_name>
14433 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14434 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14435 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14436 using SWI48 mode iterator.
14437 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14438 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14439 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14440 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14441 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14442 using SWI48 mode iterator.
14443 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14444 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14445 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14446 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14447 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14448 SWI48 mode iterator.
14449 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14450 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14451 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14452 pattern from sse_cvttsd2si<round_saeonly_name>
14453 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14455 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14457 * config/nds32/nds32-md-auxiliary.c
14458 (nds32_valid_smw_lwm_base_p): Refine.
14459 (nds32_output_smw_single_word): Refine.
14460 (nds32_output_smw_double_word): New.
14461 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14463 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14465 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14466 (nds32_output_stack_pop): Refine.
14467 (nds32_expand_unaligned_load): Refine.
14468 (nds32_expand_unaligned_store): Refine.
14470 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14471 Chung-Ju Wu <jasonwucj@gmail.com>
14473 * config/nds32/constants.md: Add TP_REGNUM constant.
14474 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14475 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14477 * config/nds32/nds32-doubleword.md: Consider flag_pic.
14478 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14479 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14480 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14481 and PIC code generation.
14482 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14484 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14486 * config/nds32/nds32.md: Support TLS and PIC.
14487 * config/nds32/nds32.c: Support TLS and PIC.
14488 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14489 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14492 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14494 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14495 mode with E_ prefix.
14497 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14498 Chung-Ju Wu <jasonwucj@gmail.com>
14500 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14501 * config/nds32/nds32-md-auxiliary.c
14502 (symbolic_reference_mentioned_p): New.
14503 (nds32_legitimize_ict_address): New.
14504 (nds32_expand_ict_move): New.
14505 (nds32_indirect_call_referenced_p): New.
14506 (nds32_symbol_binds_local_p): Delete.
14507 (nds32_long_call_p): Modify.
14508 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14509 * config/nds32/nds32-protos.h
14510 (symbolic_reference_mentioned_p): Declare.
14511 (nds32_legitimize_ict_address): Declare.
14512 (nds32_expand_ict_move): Declare.
14513 (nds32_indirect_call_referenced_p): Declare.
14514 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14515 (nds32_relax_group): Use nds32_ict_const_p as condition.
14516 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14517 (nds32_asm_file_start): Output ict_model directive in asm code.
14518 (nds32_legitimate_address_p): Consider indirect call.
14519 (nds32_print_operand): Consider indirect call.
14520 (nds32_print_operand_address): Consider indirect call.
14521 (nds32_insert_attributes): Handle "indirect_call" attribute.
14522 (TARGET_LEGITIMATE_ADDRESS_P): Define.
14523 (TARGET_LEGITIMATE_CONSTANT_P): Define.
14524 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14525 (TARGET_DELEGITIMIZE_ADDRESS): Define.
14526 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14527 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14528 (TARGET_ICT_MODEL_SMALL): Define.
14529 (TARGET_ICT_MODEL_LARGE): Define.
14530 * config/nds32/nds32.md (movsi): Consider ict model.
14531 (call, call_value): Consider ict model.
14532 (sibcall, sibcall_value): Consider ict model.
14533 * config/nds32/nds32.opt (mict-model): New option.
14534 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14537 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
14538 Monk Chiang <sh.chiang04@gmail.com>
14539 Jim Wilson <jimw@sifive.com>
14541 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14542 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
14543 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14544 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14545 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
14546 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14547 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14548 compute save_libcall_adjustment properly.
14549 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14550 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14551 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14552 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14553 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14554 (ABI_SPEC): Handle mabi=ilp32e.
14555 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14556 (RVE): Add RVE mask.
14557 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14558 <-march>: Add rv32e as an example.
14560 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
14563 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14564 (intra_create_variable_infos): Handle C++ constructors.
14566 2018-05-18 Martin Liska <mliska@suse.cz>
14568 * passes.def: Remove a redundant pass.
14570 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
14573 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14575 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14577 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14579 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14581 (armv2, armv2a, armv3, armv3m): Delete architectures.
14582 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14583 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14584 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14586 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14587 (*mulsidi3adddi): Likewise.
14588 (mulsidi3): Likewise.
14589 (*mulsidi3_nov6): Likewise.
14590 (umulsidi3): Likewise.
14591 (umulsidi3_nov6): Likewise.
14592 (umaddsidi4): Likewise.
14593 (*umulsidi3adddi): Likewise.
14594 (smulsi3_highpart): Likewise.
14595 (*smulsi3_highpart_nov6): Likewise.
14596 (umulsi3_highpart): Likewise.
14597 (*umulsi3_highpart_nov6): Likewise.
14598 * config/arm/arm.h (arm_arch3m): Delete.
14599 * config/arm/arm.c (arm_arch3m): Delete.
14600 (arm_option_override_internal): Update armv3-related comment.
14601 (arm_configure_build_target): Delete use of isa_bit_mode32.
14602 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14603 (arm_rtx_costs_internal): Delete check of arm_arch3m.
14604 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14605 (mulsa3): Likewise.
14606 (mulusa3): Likewise.
14607 * config/arm/arm-protos.h (arm_arch3m): Delete.
14608 * config/arm/arm-tables.opt: Regenerate.
14609 * config/arm/arm-tune.md: Likewise.
14610 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14611 deleted architectures.
14613 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14615 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14616 (armv5t, armv5te): New features.
14617 (ARMv5, ARMv5e): Delete fgroups.
14618 (ARMv5t, ARMv5te): Adjust for above changes.
14619 (ARMv6m): Likewise.
14620 (armv5, armv5e): Delete arches.
14621 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14623 (*call_reg_arm): Likewise.
14624 (*call_value_reg_armv5): Likewise.
14625 (*call_value_reg_arm): Likewise.
14626 (*call_symbol): Likewise.
14627 (*call_value_symbol): Likewise.
14628 (*sibcall_insn): Likewise.
14629 (*sibcall_value_insn): Likewise.
14630 (clzsi2): Likewise.
14631 (prefetch): Likewise.
14632 (define_split and define_peephole2 dependent on arm_arch5):
14634 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14636 (TARGET_ARM_QBIT): Likewise.
14637 (TARGET_DSP_MULTIPLY): Likewise.
14638 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14639 (arm_arch5, arm_arch5e): Delete.
14640 (arm_arch5t, arm_arch5te): Declare.
14641 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14642 (arm_arch5t): Declare.
14643 (arm_option_reconfigure_globals): Update for the above.
14644 (arm_options_perform_arch_sanity_checks): Update comment, replace
14645 use of arm_arch5 with arm_arch5t.
14646 (use_return_insn): Likewise.
14647 (arm_emit_call_insn): Likewise.
14648 (output_return_instruction): Likewise.
14649 (arm_final_prescan_insn): Likewise.
14650 (arm_coproc_builtin_available): Likewise.
14651 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14652 arm_arch5e with arm_arch5t and arm_arch5te.
14653 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14654 (arm_arch5t, arm_arch5te): Declare.
14655 * config/arm/arm-tables.opt: Regenerate.
14656 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14657 * config/arm/t-multilib: Likewise.
14658 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14659 instead of arm_arch5.
14660 (*call_reg_thumb1): Likewise.
14661 (*call_value_reg_thumb1_v5): Likewise.
14662 (*call_value_reg_thumb1): Likewise.
14663 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14665 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14667 2018-05-18 Martin Liska <mliska@suse.cz>
14669 PR gcov-profile/84846
14670 * doc/gcov.texi: Document -t option of gcov tool.
14672 2018-05-18 Martin Liska <mliska@suse.cz>
14674 PR gcov-profile/84846
14675 * gcov.c (print_usage): Add new -t option.
14676 (process_args): Handle the option.
14677 (generate_results): Use stdout as output when requested by
14680 2018-05-18 Martin Liska <mliska@suse.cz>
14682 PR gcov-profile/84846
14683 * coverage.c (coverage_init): Write PWD to .gcno file.
14684 * doc/gcov.texi: Document how working directory is printed.
14685 * gcov-dump.c (dump_gcov_file): Print PWD.
14686 * gcov.c (output_intermediate_file): Likewise.
14687 (read_graph_file): Read PWD string.
14688 (output_lines): Print PWD.
14690 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14692 PR middle-end/85817
14693 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14694 for retval and return false if all args to phi are zero.
14696 2018-05-18 Richard Biener <rguenther@suse.de>
14698 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14700 (evrp_dom_walker::before_dom_children): Call it.
14702 2018-05-18 Richard Biener <rguenther@suse.de>
14704 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14705 results when processing array refs with variable index.
14707 2018-05-18 Toon Moene <toon@moene.org>
14709 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14710 directly after that of -floop-interchange. Indicate that both
14711 options are enabled by default when specifying -O3.
14713 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14715 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14716 iterator. Delete separate integer-mode vec_set<mode> expander.
14717 (aarch64_simd_vec_setv2di): Delete.
14718 (vec_setv2di): Delete.
14719 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14720 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
14721 the "w, r" alternative.
14723 2018-05-18 Martin Liska <mliska@suse.cz>
14725 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14726 * tree-pass.h (make_pass_lower_switch_O0): New function.
14727 * tree-switch-conversion.c (node_has_low_bound): Remove.
14728 (node_has_high_bound): Likewise.
14729 (node_is_bounded): Likewise.
14730 (class pass_lower_switch): Make it a template type and create
14732 (pass_lower_switch::execute): Add template argument.
14733 (make_pass_lower_switch): New function.
14734 (make_pass_lower_switch_O0): New function.
14735 (do_jump_if_equal): Remove.
14736 (emit_case_nodes): Simplify to just handle all 3 cases and leave
14737 all the hard work to tree optimization passes.
14739 2018-05-18 Martin Liska <mliska@suse.cz>
14741 * dbgcnt.c (limit_low): Renamed from limit.
14742 (limit_high): New variable.
14743 (dbg_cnt_is_enabled): Check for upper limit.
14744 (dbg_cnt): Adjust dumping.
14745 (dbg_cnt_set_limit_by_index): Add new argument for high
14747 (dbg_cnt_set_limit_by_name): Likewise.
14748 (dbg_cnt_process_single_pair): Parse new format.
14749 (dbg_cnt_process_opt): Use strtok.
14750 (dbg_cnt_list_all_counters): Remove 'value' and add
14752 * doc/invoke.texi: Document changes.
14754 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
14756 * doc/sourcebuild.texi (scalar_all_fma): Document.
14757 * tree.def (FMA_EXPR): Delete.
14758 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14759 * internal-fn.c (ternary_direct): New macro.
14760 (expand_ternary_optab_fn): Likewise.
14761 (direct_ternary_optab_supported_p): Likewise.
14762 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14763 * builtins.c (fold_builtin_fma): Delete.
14764 (fold_builtin_3): Don't call it.
14765 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14766 * expr.c (expand_expr_real_2): Likewise.
14767 * fold-const.c (operand_equal_p): Likewise.
14768 (fold_ternary_loc): Likewise.
14769 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14770 * gimple.c (DEFTREECODE): Likewise.
14771 * gimplify.c (gimplify_expr): Likewise.
14772 * optabs-tree.c (optab_for_tree_code): Likewise.
14773 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14774 * tree-eh.c (operation_could_trap_p): Likewise.
14775 (stmt_could_throw_1_p): Likewise.
14776 * tree-inline.c (estimate_operator_cost): Likewise.
14777 * tree-pretty-print.c (dump_generic_node): Likewise.
14778 (op_code_prio): Likewise.
14779 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14780 * tree-ssa-operands.c (get_expr_operands): Likewise.
14781 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14782 * fold-const-call.h (fold_fma): Delete.
14783 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14784 CFN_FNMA and CFN_FNMS.
14785 (fold_fma): Delete.
14786 * genmatch.c (combined_fn): New enum.
14787 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14788 (commutative_op): New function.
14789 (commutate): Use it. Handle more than 2 operands.
14790 (dt_operand::gen_gimple_expr): Use commutative_op.
14791 (parser::parse_expr): Allow :c to be used with non-binary
14792 operators if the commutative operand is known.
14793 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14794 CFN_FMS, CFN_FNMA and CFN_FNMS.
14795 (backprop::process_assign_use): Remove FMA_EXPR handling.
14796 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14797 (gen_hsa_fma): New function.
14798 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14799 IFN_FNMA and IFN_FNMS.
14800 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14801 * gimple-fold.h (follow_all_ssa_edges): Declare.
14802 * gimple-fold.c (follow_all_ssa_edges): New function.
14803 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14804 gimple_build interface and use follow_all_ssa_edges to fold the result.
14805 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14806 instead of checking for optabs directly.
14807 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14808 rather than FMA_EXPRs.
14809 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14810 call to IFN_FMA instead of an FMA_EXPR.
14812 2018-05-17 Jim Wilson <jimw@sifive.com>
14814 * expr.c (do_tablejump): When converting index to Pmode, if we have a
14815 sign extended promoted subreg, and the range does not have the sign bit
14816 set, then do a sign extend.
14818 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14819 test, check for sign extended subreg and/or constant operands, and
14820 do a sign extend in that case.
14822 2018-05-17 Steve Ellcey <sellcey@cavium.com>
14824 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14825 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14827 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14828 Change logics_shift_reg to logics_shift_imm.
14829 (thunderx2t99_fp_loadpair_basic): Delete.
14830 (thunderx2t99_fp_storepair_basic): Delete.
14831 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14832 (thunderx2t99_asimd_polynomial): Delete.
14833 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14834 and neon_fp_mul_d_scalar_q.
14835 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14836 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14837 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14838 (thunderx2t99_asimd_lut): Add missing tbl types.
14839 (thunderx2t99_asimd_ext): Delete.
14840 (thunderx2t99_asimd_load1_1_mult): Delete.
14841 (thunderx2t99_asimd_load1_2_mult): Delete.
14842 (thunderx2t99_asimd_load1_ldp): New.
14843 (thunderx2t99_asimd_load1): New.
14844 (thunderx2t99_asimd_load2): Add missing *load2* types.
14845 (thunderx2t99_asimd_load3): New.
14846 (thunderx2t99_asimd_load4): New.
14847 (thunderx2t99_asimd_store1_1_mult): Delete.
14848 (thunderx2t99_asimd_store1_2_mult): Delete.
14849 (thunderx2t99_asimd_store2_mult): Delete.
14850 (thunderx2t99_asimd_store2_onelane): Delete.
14851 (thunderx2t99_asimd_store_stp): New.
14852 (thunderx2t99_asimd_store1): New.
14853 (thunderx2t99_asimd_store2): New.
14854 (thunderx2t99_asimd_store3): New.
14855 (thunderx2t99_asimd_store4): New.
14857 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
14859 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14860 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
14862 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
14863 Segher Boessenkool <segher@kernel.crashing.org>
14866 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14869 2018-05-17 Richard Biener <rguenther@suse.de>
14871 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14872 for pruning loop and prune defs feeding only already visited PHIs.
14874 2018-05-17 Richard Biener <rguenther@suse.de>
14876 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14878 2018-05-17 Bin Cheng <bin.cheng@arm.com>
14879 Richard Biener <rguenther@suse.de>
14881 PR tree-optimization/85793
14882 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14883 for VMAT_ELEMENTWISE.
14885 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14887 * internal-fn.h (lookup_internal_fn): Declare
14888 * internal-fn.c (lookup_internal_fn): New function.
14889 * gimple.c (gimple_build_call_from_tree): Handle calls to
14890 internal functions.
14891 * gimple-pretty-print.c (dump_gimple_call): Print "." before
14892 internal function names.
14893 * tree-pretty-print.c (dump_generic_node): Likewise.
14894 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14896 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14898 * gimple-fold.h (gimple_build): Make the function forms take
14899 combined_fn rather than built_in_function.
14900 (gimple_simplify): Likewise.
14901 * gimple-match-head.c (gimple_simplify): Likewise.
14902 * gimple-fold.c (gimple_build): Likewise.
14903 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14904 rather than gimple_build_call_internal.
14905 (get_initial_defs_for_reduction): Likewise.
14906 (vect_create_epilog_for_reduction): Likewise.
14907 (vectorizable_live_operation): Likewise.
14909 2018-05-17 Martin Liska <mliska@suse.cz>
14911 * gimple-ssa-sprintf.c (format_directive): Do not use
14912 space in between 'G_' and '('.
14914 2018-05-17 Jakub Jelinek <jakub@redhat.com>
14917 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14918 even if the mask is not all ones.
14921 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14923 (ix86_gimple_fold_builtin): Likewise.
14926 * config/i386/i386.c: Include tree-vector-builder.h.
14927 (ix86_vector_shift_count): New function.
14928 (ix86_fold_builtin): Fold shift builtins by scalar count.
14929 (ix86_gimple_fold_builtin): Likewise.
14931 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14932 _mm512_setzero): New intrinsics.
14934 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
14935 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14937 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14938 code generation for cases where splatting a value is not useful.
14939 * simplify-rtx.c (simplify_ternary_operation): Simplify
14940 vec_merge across a vec_duplicate and a paradoxical subreg forming
14941 a vector mode to a vec_concat.
14943 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
14945 * config.gcc: Support "goldmont-plus".
14946 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14948 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14949 PROCESSOR_GOLDMONT_PLUS.
14950 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14951 (processor_target_table): Add "goldmont-plus".
14952 (PTA_GOLDMONT_PLUS): Define.
14953 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14954 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14955 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14956 (fold_builtin_cpu): Add "goldmont-plus".
14957 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14958 (ix86_option_override_internal): Add "goldmont-plus".
14959 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14960 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14961 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14962 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14964 2018-05-17 Richard Biener <rguenther@suse.de>
14966 PR tree-optimization/85757
14967 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14968 remove defs that only feed that PHI from further processing.
14970 2018-05-16 Jim Wilson <jimw@sifive.com>
14972 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14974 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14976 2018-05-16 Mark Wielaard <mark@klomp.org>
14978 * dwarf2out.c (count_index_strings): New function.
14979 (output_indirect_strings): Call count_index_strings and generate
14980 header for dwarf_version >= 5.
14982 2018-05-16 Mark Wielaard <mark@klomp.org>
14984 * dwarf2out.c (dwarf_FORM): New function.
14985 (set_indirect_string): Use dwarf_FORM.
14986 (reset_indirect_string): Likewise.
14987 (size_of_die): Likewise.
14988 (value_format): Likewise.
14989 (output_die): Likewise.
14990 (add_skeleton_AT_string): Likewise.
14991 (output_macinfo_op): Likewise.
14992 (index_string): Likewise.
14993 (output_index_string_offset): Likewise.
14994 (output_index_string): Likewise.
14995 (count_index_strings): Likewise.
14997 2018-05-16 Carl Love <cel@us.ibm.com>
14999 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15000 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15002 2018-05-16 Martin Jambor <mjambor@suse.cz>
15004 * ipa-prop.c (ipa_free_all_edge_args): Remove.
15005 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15007 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
15009 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15010 (fnma<mode>4): Likewise.
15011 (fms<mode>4): Likewise.
15012 (fnms<mode>4): Likewise.
15013 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15014 (aarch64_fnma<mode>4): Likewise.
15015 (aarch64_fms<mode>4): Likewise.
15016 (aarch64_fnms<mode>4): Likewise.
15017 (aarch64_fnmadd<mode>4): Likewise.
15019 2018-05-16 Jason Merrill <jason@redhat.com>
15021 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
15023 2018-05-16 Richard Biener <rguenther@suse.de>
15025 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15026 (dump_stmt_cost): Declare.
15027 (add_stmt_cost): Dump cost we add.
15028 (add_stmt_costs): New function.
15029 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15030 No longer exported.
15031 (vect_analyze_stmt): Adjust prototype.
15032 (vectorizable_condition): Likewise.
15033 (vectorizable_live_operation): Likewise.
15034 (vectorizable_reduction): Likewise.
15035 (vectorizable_induction): Likewise.
15036 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15037 cost vector to pass to vectorizable_ and record afterwards.
15038 (vect_model_reduction_cost): Take cost vector argument and adjust.
15039 (vect_model_induction_cost): Likewise.
15040 (vectorizable_reduction): Likewise.
15041 (vectorizable_induction): Likewise.
15042 (vectorizable_live_operation): Likewise.
15043 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15044 SLP_TREE_NUMBER_OF_VEC_STMTS.
15045 (vect_analyze_slp_cost_1): Remove.
15046 (vect_analyze_slp_cost): Likewise.
15047 (vect_slp_analyze_node_operations): Take visited args and
15048 a target cost vector. Avoid processing already visited stmt sets.
15049 (vect_slp_analyze_operations): Use a local cost vector to gather
15050 costs and register those of non-discarded instances.
15051 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15052 (vect_schedule_slp_instance): Remove copying of
15053 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
15055 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15056 adding cost. Record cost entry location.
15057 (vect_prologue_cost_for_slp_op): Function to compute cost of
15058 a constant or invariant generated for SLP vect in the prologue,
15059 split out from vect_analyze_slp_cost_1.
15060 (vect_model_simple_cost): Make static. Adjust for SLP costing.
15061 (vect_model_promotion_demotion_cost): Likewise.
15062 (vect_model_store_cost): Likewise, make static.
15063 (vect_model_load_cost): Likewise.
15064 (vectorizable_bswap): Add cost vector arg and adjust.
15065 (vectorizable_call): Likewise.
15066 (vectorizable_simd_clone_call): Likewise.
15067 (vectorizable_conversion): Likewise.
15068 (vectorizable_assignment): Likewise.
15069 (vectorizable_shift): Likewise.
15070 (vectorizable_operation): Likewise.
15071 (vectorizable_store): Likewise.
15072 (vectorizable_load): Likewise.
15073 (vectorizable_condition): Likewise.
15074 (vectorizable_comparison): Likewise.
15075 (can_vectorize_live_stmts): Likewise.
15076 (vect_analyze_stmt): Likewise.
15077 (vect_transform_stmt): Adjust calls to vectorizable_*.
15078 * tree-vectorizer.c: Include gimple-pretty-print.h.
15079 (dump_stmt_cost): New function.
15081 2018-05-16 Richard Biener <rguenther@suse.de>
15083 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15084 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15085 * tree-ssa-dse.c: Include tree-ssa-loop.h.
15086 (check_name): New callback.
15087 (dse_classify_store): Track cycles via a visited bitmap of PHI
15088 defs and simplify handling of in-loop and across loop dead stores
15089 and properly fail for loop-variant refs. Handle byte-tracking with
15090 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15093 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15095 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15096 (vect_get_mask_type_for_stmt): Likewise.
15097 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15099 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
15100 to determine the statement's vector type and the vector type that
15101 should be used for calculating nunits. Deal with cases in which
15102 the type has to be deferred.
15103 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15104 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15105 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15106 (vect_determine_vf_for_stmt): New functions, split out from...
15107 (vect_determine_vectorization_factor): ...here.
15108 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15109 (vect_get_mask_type_for_stmt): New functions, split out from
15110 vect_determine_vectorization_factor.
15112 2018-05-16 Richard Biener <rguenther@suse.de>
15114 * tree-cfg.c (verify_gimple_assign_ternary): Properly
15115 verify the [VEC_]COND_EXPR embedded comparison.
15117 2018-05-15 Martin Sebor <msebor@redhat.com>
15119 PR tree-optimization/85753
15120 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15121 RECORD_TYPE in addition to ARRAY_TYPE.
15123 2018-05-15 Martin Sebor <msebor@redhat.com>
15125 PR middle-end/85643
15126 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15128 2018-05-15 Richard Biener <rguenther@suse.de>
15130 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15131 add by_clobber_p one. Change algorithm to collect all defs
15132 representing uses we need to walk and try reducing them to
15133 a single one before failing.
15134 (dse_dom_walker::dse_optimize_stmt): Adjust.
15136 2018-05-13 Mark Wielaard <mark@klomp.org>
15138 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15139 (size_of_loc_descr): Likewise.
15140 (output_loc_operands): Likewise.
15141 (output_loc_operands_raw): Likewise.
15142 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15143 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15144 (hash_loc_operands): Likewise.
15145 (compare_loc_operands): Likewise.
15147 2018-05-14 Mark Wielaard <mark@klomp.org>
15149 * dwarf2out.c (count_index_addrs): New function.
15150 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15152 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15154 PR tree-optimization/83648
15155 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15156 return value as malloc candidate.
15158 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15161 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15162 param as true in call to suggest_attribute.
15164 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
15166 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15167 -mreadonly-in-sdata.
15169 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15171 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15173 (aarch64_crypto_aesd_fused): Likewise.
15175 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
15177 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15178 (movsi_aarch64): Likewise.
15179 (load_pairsi): Likewise.
15180 (load_pairdi): Likewise.
15181 (store_pairsi): Likewise.
15182 (store_pairdi): Likewise.
15183 (load_pairsf): Likewise.
15184 (load_pairdf): Likewise.
15185 (store_pairsf): Likewise.
15186 (store_pairdf): Likewise.
15187 (zero_extend): Likewise.
15188 (trunc): Swap alternatives.
15189 (fcvt_target): Add '?' to prefer w over r.
15191 2018-05-14 Jakub Jelinek <jakub@redhat.com>
15194 * config/i386/i386.md: Disallow non-commutative arithmetics in
15195 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15196 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
15197 in the peephole2 before it.
15199 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
15201 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15202 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15203 (ix86_handle_option): Handle -mcldemote.
15204 * config.gcc: New header.
15205 * config/i386/cldemoteintrin.h: New file.
15206 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15207 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15209 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15210 OPTION_MASK_ISA_CLDEMOTE.
15211 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15212 (ix86_valid_target_attribute_inner_p): Ditto.
15213 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15214 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15215 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15216 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15217 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15219 * config/i386/i386.opt: Add -mcldemote.
15220 * config/i386/x86intrin.h: New header.
15221 * doc/invoke.texi: Add -mcldemote.
15223 2018-05-14 Richard Biener <rguenther@suse.de>
15225 * doc/match-and-simplify.texi: Adjust :s documentation.
15227 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
15229 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15230 intended memcpy size.
15231 (REORDER_45): Likewise.
15233 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
15235 * sort.cc: New file.
15236 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15237 * vec.c (qsort_chk): Use gcc_qsort.
15238 * Makefile.in (OBJS-libcommon): Add sort.o.
15239 (build/sort.o): New target. Use it...
15240 (BUILD_RTL): ... here, and...
15241 (build/gencfn-macros): ... here, and...
15242 (build/genmatch): ... here.
15244 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15245 Chung-Ju Wu <jasonwucj@gmail.com>
15247 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15248 * config/nds32/nds32-graywolf.md: New file.
15249 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15250 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15252 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15253 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15254 * config/nds32/nds32.md (pipeline_model): Add graywolf.
15255 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15256 * config/nds32/pipelines.md: Include n15 settings.
15258 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15259 Chung-Ju Wu <jasonwucj@gmail.com>
15261 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15262 * config/nds32/nds32-n13.md: New file.
15263 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15264 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15266 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15267 * config/nds32/nds32.md (pipeline_model): Add n13.
15268 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15269 * config/nds32/pipelines.md: Include n13 settings.
15271 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15272 Chung-Ju Wu <jasonwucj@gmail.com>
15274 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15275 * config/nds32/nds32-n10.md: New file.
15276 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15277 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15279 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15280 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15281 * config/nds32/nds32.md (pipeline_model): Add n10.
15282 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15283 * config/nds32/pipelines.md: Include n10 settings.
15285 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
15286 Kito Cheng <kito.cheng@gmail.com>
15287 Chung-Ju Wu <jasonwucj@gmail.com>
15289 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15290 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15291 Add enum values for DSP extension instructions.
15292 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15294 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15295 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15296 New code iterators.
15297 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15298 * config/nds32/nds32-dspext.md: New file for DSP implementation.
15299 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15300 * config/nds32/nds32-intrinsic.md: Likewise.
15301 * config/nds32/nds32_intrinsic.h: Likewise.
15302 * config/nds32/nds32-md-auxiliary.c: Likewise.
15303 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15304 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15305 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15306 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15307 * config/nds32/nds32-protos.h: New declarations for DSP extension.
15308 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15309 TYPE_DMAC in switch statement.
15310 * config/nds32/nds32.c: New checking and implementation for DSP
15311 extension instructions.
15312 * config/nds32/nds32.h: Likewise.
15313 * config/nds32/nds32.md: Likewise.
15314 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15315 * config/nds32/predicates.md: Implement new predicates for DSP
15318 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15320 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15321 Reformat alternatives and attributes so it is easier to identify
15322 which constraints/attributes go with which instruction.
15323 (mov<mode>_hardfloat32, FMOVE64): Likewise.
15324 (mov<mode>_softfloat32, FMOVE64): Likewise.
15325 (mov<mode>_hardfloat64, FMOVE64): Likewise.
15326 (mov<mode>_softfloat64, FMOVE64): Likewise.
15328 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
15330 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15332 (Basic PowerPC Built-in Functions): The new name of the
15333 subsection previously known as "PowerPC Built-in Functions".
15334 (Basic PowerPC Built-in Functions Available on all Configurations):
15336 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15337 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15338 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15339 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15341 2018-05-11 Martin Jambor <mjambor@suse.cz>
15344 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15347 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15350 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15352 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
15354 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15355 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15356 (ix86_handle_option): Handle -mwaitpkg.
15357 * config.gcc: New header.
15358 * config/i386/cpuid.h (bit_WAITPKG): New bit.
15359 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15360 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15362 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15363 OPTION_MASK_ISA_WAITPKG.
15364 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15365 (ix86_option_override_internal): Add PTA_WAITPKG.
15366 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15367 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15368 IX86_BUILTIN_TPAUSE.
15369 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15370 __builtin_ia32_umwait and __builtin_ia32_tpause.
15371 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15372 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15373 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15374 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15375 UNSPECV_TPAUSE): New.
15376 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15377 * config/i386/i386.opt: Add -mwaitpkg.
15378 * config/i386/waitpkgintrin.h: New file.
15379 * config/i386/x86intrin.h: New header.
15380 * doc/invoke.texi: Add -mwaitpkg.
15382 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15385 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15387 (cortex-m0): Use armv6s-m isa.
15388 (cortex-m0plus): Likewise.
15389 (cortex-m1): Likewise.
15390 (cortex-m0.small-multiply): Likewise.
15391 (cortex-m0plus.small-multiply): Likewise.
15392 (cortex-m1.small-multiply): Likewise.
15394 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
15395 Jakub Jelinek <jakub@redhat.com>
15397 PR tree-optimization/85692
15398 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15399 source permute as well.
15401 2018-05-11 Martin Liska <mliska@suse.cz>
15404 * doc/extend.texi: Document LLVM style format for no_sanitize
15407 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15409 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15410 mode_supports_vsx_dform_quad to mode_supports_dq_form.
15411 (mode_supports_vsx_dform_quad): Likewise.
15412 (mode_supports_vmx_dform): Move these functions to be next to the
15413 other mode_supports functions.
15414 (mode_supports_dq_form): Likewise.
15415 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15416 mode_supports_dq_form.
15417 (reg_offset_addressing_ok_p): Likewise.
15418 (offsettable_ok_by_alignment): Likewise.
15419 (rs6000_legitimate_offset_address_p): Likewise.
15420 (legitimate_lo_sum_address_p): Likewise.
15421 (rs6000_legitimize_address): Likewise.
15422 (rs6000_legitimize_reload_address): Likewise.
15423 (rs6000_secondary_reload_inner): Likewise.
15424 (rs6000_preferred_reload_class): Likewise.
15425 (rs6000_output_move_128bit): Likewise.
15427 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15429 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15430 Generate SImode target register for null target.
15431 <case IX86_BUILTIN_XGETBV>: Ditto.
15432 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15433 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15435 2018-05-10 Carl Love <cel@us.ibm.com>
15437 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15438 dcbtt and dcbtstt if operands[2] is 0.
15440 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15443 * config/i386/sse.md (usadv64qi): New expander.
15445 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
15447 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15448 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15449 -maltivec=be support.
15450 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15451 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15452 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15453 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15454 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15455 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15456 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15457 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15458 altivec_vsumsws): Adjust.
15459 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15460 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15461 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15463 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15464 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15465 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15466 (altivec_lve<VI_char>x): Delete expand.
15467 (*altivec_lve<VI_char>x_internal): Rename to...
15468 (altivec_lve<VI_char>x): ... this.
15469 (altivec_lvxl_<mode>): Delete expand.
15470 (*altivec_lvxl_<mode>_internal): Rename to ...
15471 (altivec_lvxl_<mode>): ... this.
15472 (altivec_stvxl_<mode>): Delete expand.
15473 (*altivec_stvxl_<mode>_internal): Rename to ...
15474 (altivec_stvxl_<mode>): ... this.
15475 (altivec_stve<VI_char>x): Delete expand.
15476 (*altivec_stve<VI_char>x_internal): Rename to ...
15477 (altivec_stve<VI_char>x): ... this.
15478 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15479 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15480 reduc_plus_scal_<mode>): Adjust.
15481 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15483 (rs6000_cpu_cpp_builtins): Adjust.
15484 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15485 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15486 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15487 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15488 -maltivec=be support.
15489 (rs6000_split_vec_extract_var): Adjust.
15490 (rs6000_split_v4si_init): Adjust.
15491 (swap_selector_for_mode): Delete.
15492 (altivec_expand_lvx_be, altivec_expand_stvx_be,
15493 altivec_expand_stvex_be): Delete.
15494 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15495 -maltivec=be support.
15496 (rs6000_gimple_fold_builtin): Ditto.
15497 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15499 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15500 (TARGET_DIRECT_MOVE_64BIT): Adjust.
15501 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15502 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15503 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15504 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15505 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15506 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15507 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15508 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15509 anonymous split): Adjust.
15510 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15511 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15513 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
15515 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15516 when --with-gxx-include-dir is also specified.
15517 * configure: Regenerate.
15519 2018-05-09 Jim Wilson <jimw@sifive.com>
15522 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15523 * config/riscv/t-withmultilib: New.
15524 * config/riscv/withmultilib.h: New.
15525 * doc/install.texi: Document RISC-V --with-multilib-list support.
15527 2018-05-09 Richard Biener <rguenther@suse.de>
15529 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15531 (vect_bb_vectorization_profitable_p): Adjust. Compute
15532 actual scalar cost using the cost vector and the add_stmt_cost
15535 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15537 PR rtl-optimization/85645
15538 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15539 in the REG_CFA_REGISTER note for LR, don't leave it empty.
15541 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15543 PR rtl-optimization/85645
15544 * shrink-wrap.c (spread_components): Return a boolean saying if
15545 anything was changed.
15546 (try_shrink_wrapping_separate): Iterate spread_components until
15547 nothing changes anymore.
15549 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15551 PR rtl-optimization/85645
15552 * regrename.c (build_def_use): Also kill the chains that include the
15553 destination of a REG_CFA_REGISTER note.
15555 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15557 PR rtl-optimization/85645
15558 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15559 insn that has a REG_CFA_REGISTER note.
15561 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
15563 * cfgexpand.c (expand_clobber): New function.
15564 (expand_gimple_stmt_1): Use it.
15565 * tree-vect-stmts.c (vect_clobber_variable): New function,
15567 (vectorizable_simd_clone_call): ...here.
15568 (vectorizable_store): Emit a clobber either side of an
15569 IFN_STORE_LANES sequence.
15570 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15572 2018-05-09 Tom de Vries <tom@codesourcery.com>
15575 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15576 (define_insn "trap_if_false"): Add exit after trap.
15578 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
15580 PR rtl-optimization/85638
15581 * bb-reorder.c: Include common/common-target.h.
15582 (create_forwarder_block): New function extracted from...
15583 (fix_up_crossing_landing_pad): ...here. Rename into...
15584 (dw2_fix_up_crossing_landing_pad): ...this.
15585 (sjlj_fix_up_crossing_landing_pad): New function.
15586 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15587 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15588 from both partitions and exit the loop after one iteration.
15590 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15593 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15595 (Basic PowerPC Built-in Functions): The new name of the
15596 subsection previously known as "PowerPC Built-in Functions".
15597 (Basic PowerPC Built-in Functions Available on all Configurations):
15599 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15601 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15603 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15605 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15608 2018-05-08 Jim Wilson <jimw@sifive.com>
15610 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15611 (LD_EMUL_SUFFIX): New.
15612 (LINK_SPEC): Use it.
15614 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15616 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15618 (Basic PowerPC Built-in Functions): The new name of the
15619 subsection previously known as "PowerPC Built-in Functions".
15620 (Basic PowerPC Built-in Functions Available on all Configurations):
15622 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15624 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15626 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15628 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15631 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15634 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15635 after cmpelim optimization.
15637 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
15639 * config.gcc: Support "goldmont".
15640 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15641 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15642 PROCESSOR_GOLDMONT.
15643 * config/i386/i386.c (m_GOLDMONT): Define.
15644 (processor_target_table): Add "goldmont".
15645 (PTA_GOLDMONT): Define.
15646 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15647 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15648 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15649 (fold_builtin_cpu): Add "goldmont".
15650 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15651 (ix86_option_override_internal): Add "goldmont".
15652 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15653 (processor_type): Add PROCESSOR_GOLDMONT.
15654 * config/i386/i386.md: Add CPU "glm".
15655 * config/i386/glm.md: New file.
15656 * config/i386/x86-tune.def: Add m_GOLDMONT.
15657 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15659 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15662 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15664 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15665 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
15666 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
15669 * config/i386/i386.c (ix86_fold_builtin): Handle
15670 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15673 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15674 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15676 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
15679 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15680 (check_arch): Likewise.
15681 (check_fpu): Return the result rather than printing it.
15682 (end arch): Fix operator precedence.
15683 (end cpu): Likewise.
15684 (END): Print the result from check_fpu.
15686 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15687 Alan Hayward <alan.hayward@arm.com>
15688 David Sherwood <david.sherwood@arm.com>
15690 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15691 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15692 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15693 (*fcmuo<mode>_and): New patterns.
15695 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15697 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15698 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15699 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15700 (cmp_op, sve_imm_con): New code attributes.
15701 (SVE_COND_INT_CMP, imm_con): Delete.
15702 (cmp_op): Remove above unspecs from int attribute.
15703 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15705 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
15706 comparison-specific unspecs.
15707 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15708 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15709 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15710 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15711 (*vec_fcm<cmp_op><mode>): Rename to...
15712 (*fcm<cmp_op><mode>): ...this and adjust likewise.
15713 (*vec_fcmuo<mode>): Rename to...
15714 (*fcmuo<mode>): ...this and adjust likewise.
15715 (*pred_fcm<cmp_op><mode>): New pattern.
15716 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15717 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15719 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15721 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15722 (aarch64_emit_sve_predicated_cond): New function.
15723 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15724 (aarch64_emit_unspec_cond_or): Replace with...
15725 (aarch64_emit_sve_or_conds): ...this new function. Use
15726 aarch64_emit_sve_ptrue_op for the individual comparisons and
15727 aarch64_emit_binop to OR them together.
15728 (aarch64_emit_inverted_unspec_cond): Replace with...
15729 (aarch64_emit_sve_inverted_cond): ...this new function. Use
15730 aarch64_emit_sve_ptrue_op for the comparison and
15731 aarch64_emit_unop to invert the result.
15732 (aarch64_expand_sve_vec_cmp_float): Update after the above
15733 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
15735 2018-05-07 Nathan Sidwell <nathan@acm.org>
15737 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15738 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15739 (Backwards Compatibility): Likewise.
15741 2018-05-07 Luis Machado <luis.machado@linaro.org>
15745 2018-05-07 Luis Machado <luis.machado@linaro.org>
15747 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15748 <prefetch_dynamic_strides>: New const bool field.
15749 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15750 prefetch_dynamic_strides.
15751 (exynosm1_prefetch_tune): Likewise.
15752 (thunderxt88_prefetch_tune): Likewise.
15753 (thunderx_prefetch_tune): Likewise.
15754 (thunderx2t99_prefetch_tune): Likewise.
15755 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15757 (aarch64_override_options_internal): Update to set
15758 PARAM_PREFETCH_DYNAMIC_STRIDES.
15759 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15760 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15761 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15762 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15763 prefetch-dynamic-strides setting.
15765 2018-05-07 Luis Machado <luis.machado@linaro.org>
15767 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15768 <minimum_stride>: New const int field.
15769 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15770 minimum_stride field.
15771 (exynosm1_prefetch_tune): Likewise.
15772 (thunderxt88_prefetch_tune): Likewise.
15773 (thunderx_prefetch_tune): Likewise.
15774 (thunderx2t99_prefetch_tune): Likewise.
15775 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15776 (aarch64_override_options_internal): Update to set
15777 PARAM_PREFETCH_MINIMUM_STRIDE.
15778 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15779 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15780 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15781 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15782 stride is constant and is below the minimum stride threshold.
15784 2018-05-07 Luis Machado <luis.machado@linaro.org>
15786 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15789 2018-05-07 Luis Machado <luis.machado@linaro.org>
15791 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15792 <prefetch_dynamic_strides>: New const bool field.
15793 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15794 prefetch_dynamic_strides.
15795 (exynosm1_prefetch_tune): Likewise.
15796 (thunderxt88_prefetch_tune): Likewise.
15797 (thunderx_prefetch_tune): Likewise.
15798 (thunderx2t99_prefetch_tune): Likewise.
15799 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15801 (aarch64_override_options_internal): Update to set
15802 PARAM_PREFETCH_DYNAMIC_STRIDES.
15803 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15804 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15805 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15806 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15807 prefetch-dynamic-strides setting.
15809 2018-05-07 Luis Machado <luis.machado@linaro.org>
15811 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15812 <minimum_stride>: New const int field.
15813 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15814 minimum_stride field.
15815 (exynosm1_prefetch_tune): Likewise.
15816 (thunderxt88_prefetch_tune): Likewise.
15817 (thunderx_prefetch_tune): Likewise.
15818 (thunderx2t99_prefetch_tune): Likewise.
15819 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15820 (aarch64_override_options_internal): Update to set
15821 PARAM_PREFETCH_MINIMUM_STRIDE.
15822 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15823 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15824 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15825 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15826 stride is constant and is below the minimum stride threshold.
15828 2018-05-06 Jakub Jelinek <jakub@redhat.com>
15831 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15832 the type is addressable. Don't force op into register if it has
15835 2018-05-05 Roland McGrath <mcgrathr@google.com>
15838 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15839 any section for which we don't know a specific type it should have,
15840 regardless of name. Previously this was done only for the exact
15841 names ".init_array", ".fini_array", and ".preinit_array".
15842 (default_elf_asm_named_section): Add comment about
15843 relationship with default_section_type_flags and SECTION_NOTYPE.
15844 (get_section): Don't consider it a type conflict if one side has
15845 SECTION_NOTYPE and the other doesn't, as long as neither has the
15846 SECTION_BSS et al used in the default_section_type_flags logic.
15848 2018-05-05 Tom de Vries <tom@codesourcery.com>
15851 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15852 (workaround_barsyncs): New function.
15853 (nvptx_reorg): Use workaround_barsyncs.
15854 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15855 (define_expand "nvptx_membar_cta"): New define_expand.
15856 (define_insn "*nvptx_membar_cta"): New insn.
15858 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
15860 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15861 To improve optimization opportunities.
15862 * builtin-types.def: The new needed builtin types for the above.
15864 2018-05-04 Richard Biener <rguenther@suse.de>
15866 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15867 * gimple-ssa-store-merging.c
15868 (imm_store_chain_info::output_merged_store): Remove redundant create,
15869 release split_store vector contents on failure.
15870 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15871 scalar stmt vector on cache hit.
15873 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
15875 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15877 * config.gcc (powerpc-xilinx-eabi*): Remove.
15878 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15880 (fusion_addis_mem_combo_load): Ditto.
15881 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15883 (rs6000_cpu_cpp_builtins): Ditto.
15884 * config/rs6000/rs6000-linux.c
15885 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15886 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15887 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15889 (rs6000_setup_reg_addr_masks): Ditto.
15890 (rs6000_init_hard_regno_mode_ok): Ditto.
15891 (rs6000_option_override_internal): Ditto.
15892 (legitimate_lo_sum_address_p): Ditto.
15893 (rs6000_legitimize_address): Ditto.
15894 (rs6000_legitimize_reload_address): Ditto.
15895 (rs6000_legitimate_address_p): Ditto.
15896 (abi_v4_pass_in_fpr): Ditto.
15897 (setup_incoming_varargs): Ditto.
15898 (rs6000_gimplify_va_arg): Ditto.
15899 (rs6000_split_multireg_move): Ditto.
15900 (rs6000_savres_strategy): Ditto.
15901 (rs6000_emit_prologue_components): Ditto.
15902 (rs6000_emit_epilogue_components): Ditto.
15903 (rs6000_emit_prologue): Ditto.
15904 (rs6000_emit_epilogue): Ditto.
15905 (rs6000_elf_file_end): Ditto.
15906 (rs6000_function_value): Ditto.
15907 (rs6000_libcall_value): Ditto.
15908 * config/rs6000/rs6000.h: Ditto.
15909 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15910 (TARGET_MINMAX): ... this. New.
15911 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15912 * config/rs6000/rs6000.md: Remove Xilinx FP support.
15913 (*movsi_internal1_single): Delete.
15914 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15915 mfpu=, mxilinx-fpu): Delete.
15916 * config/rs6000/singlefp.h: Delete.
15917 * config/rs6000/sysv4.h: Remove Xilinx FP support.
15918 * config/rs6000/t-rs6000: Ditto.
15919 * config/rs6000/t-xilinx: Delete.
15920 * config/rs6000/titan.md: Adjust for fp_type removal.
15921 * config/rs6000/vsx.md: Remove Xilinx FP support.
15922 (VStype_simple): Delete.
15923 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15924 * config/rs6000/xfpu.h: Delete.
15925 * config/rs6000/xfpu.md: Delete.
15926 * config/rs6000/xilinx.h: Delete.
15927 * config/rs6000/xilinx.opt: Delete.
15928 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15929 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15931 2018-05-04 Tom de Vries <tom@codesourcery.com>
15934 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15937 2018-05-04 Richard Biener <rguenther@suse.de>
15939 PR middle-end/85627
15940 * tree-complex.c (update_complex_assignment): We are always in SSA form.
15941 (expand_complex_div_wide): Likewise.
15942 (expand_complex_operations_1): Likewise.
15943 (expand_complex_libcall): Preserve EH info of the original stmt.
15944 (tree_lower_complex): Handle removed blocks.
15945 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15946 on complex multiplication and division libcall builtins.
15948 2018-05-04 Richard Biener <rguenther@suse.de>
15950 PR middle-end/85574
15951 * fold-const.c (negate_expr_p): Restrict negation of operand
15952 zero of a division to when we know that can happen without
15954 (fold_negate_expr_1): Likewise.
15956 2018-05-04 Jakub Jelinek <jakub@redhat.com>
15959 * real.h (real_nextafter): Declare.
15960 * real.c (real_nextafter): New function.
15961 * fold-const-call.c (fold_const_nextafter): New function.
15962 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15963 CASE_CFN_NEXTTOWARD.
15964 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15965 even when arg1_mode is different from arg0_mode.
15967 2018-05-03 Nathan Sidwell <nathan@acm.org>
15969 * doc/extend.texi (Deprecated Features): Remove
15970 -ffriend-injection.
15971 (Backwards Compatibility): Likewise.
15972 * doc/invoke.texi (C++ Language Options): Likewise.
15973 (C++ Dialect Options): Likewise.
15975 2018-05-03 Jakub Jelinek <jakub@redhat.com>
15978 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15979 _mm512_mask_mullox_epi64): New intrinsics.
15981 2018-05-03 Tom de Vries <tom@codesourcery.com>
15984 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15985 dump files): Add offload-tree.
15987 2018-05-03 Richard Biener <rguenther@suse.de>
15989 PR tree-optimization/85615
15990 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15991 to loops not nested in BBs loop father to avoid creating multi-entry
15994 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15996 PR tree-optimization/70291
15997 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15998 arguments. Change return type to tree. Emit libcall as a new
15999 statement rather than replacing existing one when inplace_p is true.
16000 (expand_complex_multiplication_components): New function.
16001 (expand_complex_multiplication): Expand floating-point complex
16002 multiplication using the above.
16003 (expand_complex_division): Rename inner_type parameter to type.
16004 Update expand_complex_libcall call-site.
16005 (expand_complex_operations_1): Update expand_complex_multiplication
16006 and expand_complex_division call-sites.
16008 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16011 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16012 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16013 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16014 the highest significant bit of the shift count mask is clear. In
16015 check whether and[sq]i3 is needed verify that all significant bits
16016 of the shift count other than the highest are set.
16018 2018-05-02 Tom de Vries <tom@codesourcery.com>
16021 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16022 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16023 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16024 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16025 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16026 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16027 __builtin_goacc_parlevel_size.
16029 2018-05-02 Richard Biener <rguenther@suse.de>
16031 PR tree-optimization/85597
16032 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16033 do not use split vect_get_vec_defs call but call vect_get_slp_defs
16036 2018-05-02 Tom de Vries <tom@codesourcery.com>
16039 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16040 dump files): Add ltrans-tree.
16042 2018-05-02 Tom de Vries <tom@codesourcery.com>
16045 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16046 dump files): Add wpa-ipa.
16048 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
16050 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
16051 powerpc*-*-linux*paired* target.
16052 * config/rs6000/750cl.h: Delete.
16053 * config/rs6000/paired.h: Delete.
16054 * config/rs6000/paired.md: Delete.
16055 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16057 * config/rs6000/rs6000-builtin.def: Remove paired float support.
16058 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16059 comment. Remove paired float support.
16060 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16061 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16063 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16064 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16066 * config/rs6000/rs6000.c: Remove paired float support.
16067 (paired_expand_vector_init, paired_expand_vector_move,
16068 paired_emit_vector_compare, paired_emit_vector_cond_expr,
16069 (paired_expand_lv_builtin, paired_expand_stv_builtin,
16070 paired_expand_builtin, paired_expand_predicate_builtin,
16071 paired_init_builtins): Delete.
16072 * config/rs6000/rs6000.h: Remove paired float support.
16073 * config/rs6000/rs6000.md: Remove paired float support.
16074 (move_from_CR_ov_bit): Delete.
16075 * config/rs6000/rs6000.opt (mpaired): Delete.
16076 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16077 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16079 2018-05-02 Richard Biener <rguenther@suse.de>
16081 PR middle-end/85567
16082 * gimplify.c (gimplify_save_expr): When in SSA form allow
16083 SAVE_EXPRs to compute to SSA vars.
16085 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16088 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16089 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16090 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16091 clobber operands[2], instead use a new pseudo. Formatting fixes.
16093 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
16095 PR tree-optimization/85586
16096 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16097 exit early for statements in the same group if the accesses are
16100 2018-05-02 Tom de Vries <tom@codesourcery.com>
16103 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16106 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
16108 PR tree-optimization/85143
16109 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16111 2018-05-01 Tom de Vries <tom@codesourcery.com>
16114 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16115 not found" error message.
16117 2018-05-01 Tom de Vries <tom@codesourcery.com>
16120 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16121 * vec.c (test_ordered_remove_if): New function.
16122 (vec_c_tests): Call test_ordered_remove_if.
16123 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16124 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16125 * tree-vect-patterns.c (vect_pattern_recog_1): Use
16126 VEC_ORDERED_REMOVE_IF.
16128 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16130 PR tree-optimization/82665
16131 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16132 pointer subtraction where arguments come from a memchr call.
16134 2018-05-01 Jakub Jelinek <jakub@redhat.com>
16136 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16137 --push-state --as-needed and --pop-state instead of --as-needed and
16138 --no-as-needed if ld supports it.
16139 * configure: Regenerated.
16142 * doc/install.texi2html: Replace _002d with - and _002a with * in
16143 generated html files using sed.
16145 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16148 * gcc-rich-location.c (blank_line_before_p): New function.
16149 (use_new_line): New function.
16150 (gcc_rich_location::add_fixit_insert_formatted): New function.
16151 * gcc-rich-location.h
16152 (gcc_rich_location::add_fixit_insert_formatted): New function.
16154 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16156 * selftest.c (assert_streq): Rename "expected" and "actual" to
16157 "val1" and "val2". Extend NULL-handling to cover both inputs
16158 symmetrically, while still requiring both to be non-NULL for a pass.
16159 * selftest.h (assert_streq): Rename "expected" and "actual" to
16161 (ASSERT_EQ): Likewise.
16162 (ASSERT_EQ_AT): Likewise.
16163 (ASSERT_KNOWN_EQ): Likewise.
16164 (ASSERT_KNOWN_EQ_AT): Likewise.
16165 (ASSERT_NE): Likewise.
16166 (ASSERT_MAYBE_NE): Likewise.
16167 (ASSERT_MAYBE_NE_AT): Likewise.
16168 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
16169 the assertion to pass.
16170 (ASSERT_STREQ_AT): Likewise.
16172 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16174 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16175 interaction with -pie.
16177 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16179 * selftest.h: Fix alphabetization of per-source-file selftest
16182 2018-04-30 Jason Merrill <jason@redhat.com>
16184 PR c++/61982 - dead stores to destroyed objects.
16185 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16188 2018-04-30 Jason Merrill <jason@redhat.com>
16190 * tree.c (build_clobber): New.
16191 * tree.h: Declare it.
16192 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16194 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16196 * diagnostic-show-locus.c (layout::layout): Update for
16197 location_get_source_line returning a char_span.
16198 (struct char_span): Move to input.h.
16199 (struct correction): Update for fields in char_span becoming
16201 (struct source_line): Update for location_get_source_line
16202 returning a char_span.
16203 (layout::print_line): Likewise.
16204 * edit-context.c (edited_file::print_content): Likewise.
16205 (edited_file::print_diff_hunk): Likewise.
16206 (edited_file::print_run_of_changed_lines): Likewise.
16207 (edited_file::get_num_lines): Likewise.
16208 (edited_line::edited_line): Likewise.
16209 * final.c (asm_show_source): Likewise.
16210 * input.c (location_get_source_line): Convert return type
16211 from const char * to char_span, losing the final "line_len"
16213 (dump_location_info): Update for the above.
16214 (get_substring_ranges_for_loc): Likewise. Use a char_span
16215 when handling the literal within the line.
16216 (test_reading_source_line): Update for location_get_source_line
16217 returning a char_span.
16218 * input.h (class char_span): Move here from
16219 diagnostic-show-locus.c, converting from a struct to a class.
16220 Make data members private.
16221 (char_span::operator bool): New.
16222 (char_span::length): New.
16223 (char_span::get_buffer): New.
16224 (char_span::operator[]): New.
16225 (char_span::subspan): Make const.
16226 (char_span::xstrdup): New.
16227 (location_get_source_line): Convert return type from const char *
16228 to char_span, losing the final "line_size" param.
16230 2018-04-30 Jan Hubicka <jh@suse.cz>
16232 * lto-wrapper.c (ltrans_priorities): New static var.
16233 (cmp_priority): New.
16234 (run_gcc): Read priorities and if doing parallel build order
16235 the Makefile by them.
16237 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16239 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16241 2018-04-30 Richard Biener <rguenther@suse.de>
16243 * tree-cfg.c (verify_address): Remove base argument, add
16244 flag whether to check TREE_ADDRESSABLE and do that.
16245 (verify_expr): Remove.
16246 (verify_types_in_gimple_reference): Add pieces from verify_expr.
16247 (verify_gimple_assign_single): Likewise.
16248 (verify_gimple_switch): Likewise.
16249 (verify_expr_location_1): Dereference tp once. Add (disabled)
16250 piece from verify_expr.
16251 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16253 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16255 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16257 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16259 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16260 (small_data_pattern): Likewise.
16261 (arc_rewrite_small_data): Likewise.
16262 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16263 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16264 (get_symbol_alignment): New function.
16265 (legitimate_small_data_address_p): Likewise.
16266 (legitimate_scaled_address): Update, call
16267 legitimate_small_data_address_p.
16268 (output_sdata): New static variable.
16269 (arc_print_operand): Update how we handle small data operands.
16270 (arc_print_operand_address): Likewise.
16271 (arc_legitimate_address_p): Update, use
16272 legitimate_small_data_address_p.
16273 (arc_rewrite_small_data_p): Remove.
16274 (arc_rewrite_small_data_1): Likewise.
16275 (arc_rewrite_small_data): Likewise.
16276 (small_data_pattern): Likewise.
16277 (compact_sda_memory_operand): Update to use
16278 legitimate_small_data_address_p and get_symbol_alignment.
16279 (prepare_move_operands): Don't rewite sdata pattern.
16280 (prepare_extend_operands): Remove.
16281 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16283 (zero_extendqisi2): Likewise.
16284 (zero_extendhisi2): Likewise.
16285 (extendqihi2): Likewise.
16286 (extendqisi2): Likewise.
16287 (extendhisi2): Likewise.
16288 (addsi3): Likewise.
16289 (subsi3): Likewise.
16290 (andsi3): Likewise.
16291 * config/arc/constraints.md (Usd): Change it to memory constraint.
16293 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16295 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16296 as source of std instructions.
16297 * config/arc/arc.md (movsi_insn): Update pattern predicate to
16298 allow 6-bit constants as source for store instructions.
16299 (movdi_insn): Update instruction pattern to allow 6-bit constants
16300 as source for store instructions.
16302 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16304 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16306 2018-04-30 Nathan Sidwell <nathan@acm.org>
16307 Sandra Loosemore <sandra@codesourcery.com>
16309 * dumpfile.c (dump_open): Allow '-' for stdout.
16310 * doc/invoke.texi (Developer Options): Document dump filename
16311 determination early. Document stdin/stdout selection.
16313 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
16315 Microblaze Target: PIC data text relative
16317 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16318 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16320 * config/microblaze/microblaze.h (microblaze_constant_address_p):
16321 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16322 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16323 New addressing mode for data-text relative position indepenedent code.
16324 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16325 'ADDRESS_SYMBOLIC_TXT_REL'.
16326 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16327 (microblaze_legitimate_pic_operand): Exclude function calls from
16328 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16329 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16331 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16332 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16333 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16334 for 'address + offset'.
16335 (microblaze_expand_prologue): Add new function prologue call for
16337 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16338 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16339 table in case of TARGET_PIC_DATA_TEXT_REL.
16340 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16341 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16342 Add new macros 'UNSPEC_TEXT',
16343 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16344 + exclude function calls from 'UNSPEC_PLT' in case of data text
16346 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16347 new target hook for generating address diff vector tables in case of
16349 * doc/tm.texi : Regenerate.
16350 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16351 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16352 of addr diff vector generation.
16353 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16354 target hook definition.
16355 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16356 Add default function for generate_pic_addr_diff_vec -> flag_pic.
16357 * doc/invoke.texi (Add new pic option): Add new microblaze pic
16358 option for data text relative.
16360 2018-04-30 Richard Biener <rguenther@suse.de>
16362 * tree-chrec.h (evolution_function_is_constant_p): Remove
16364 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16366 2018-04-30 Richard Biener <rguenther@suse.de>
16369 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16371 2018-04-30 Richard Biener <rguenther@suse.de>
16373 PR tree-optimization/28364
16374 PR tree-optimization/85275
16375 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16376 copying first exit test.
16378 2018-04-28 Mark Wielaard <mark@klomp.org>
16380 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16381 dwarf_version >= 5.
16382 (dwarf_AT): Handle DW_AT_addr_base.
16383 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16385 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
16388 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16389 (*ashl<dwi>3_doubleword_mask_1): Ditto.
16390 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16391 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16393 2018-04-28 Richard Biener <rguenther@suse.de>
16395 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16396 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16397 to reflect use. Only add interesting stmts.
16399 2018-04-27 Martin Jambor <mjambor@suse.cz>
16402 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16403 the jump function allows for passing through aggregate values.
16405 2018-04-27 David Malcolm <dmalcolm@redhat.com>
16407 * input.h (in_system_header_at): Convert from macro to inline
16409 (from_macro_expansion_at): Likewise.
16410 (from_macro_definition_at): Likewise.
16412 2018-04-27 Jeff Law <law@redhat.com>
16414 * config.gcc: Mark tile* targets as deprecated/obsolete.
16416 2018-04-27 Richard Biener <rguenther@suse.de>
16418 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16421 2018-04-27 Richard Biener <rguenther@suse.de>
16423 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16425 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
16427 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16428 with Yd constraint. Set "preferred_for_speed" attribute from
16429 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16430 with Yd constraint.
16431 (*movdi_internal): Ditto.
16432 (movti_interunit splitters): Remove
16433 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16434 (movdi_interunit splitters): Ditto.
16435 * config/i386/constraints.md (Ye): Remove.
16436 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16438 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16441 * config/aarch64/constraints.md (Usg): Limit to 31.
16442 (Usj): Limit to 63.
16444 2018-04-27 Jakub Jelinek <jakub@redhat.com>
16446 PR tree-optimization/85529
16447 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16448 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16449 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16450 zero extension or masking of the MSB bit.
16451 (optimize_range_tests): Add FIRST_BB argument, pass it through
16452 to optimize_range_tests_var_bound.
16453 (maybe_optimize_range_tests, reassociate_bb): Adjust
16454 optimize_range_tests callers.
16456 2018-04-26 Richard Biener <rguenther@suse.de>
16457 Jakub Jelinek <jakub@redhat.com>
16459 * cgraph.h (symbol_table): Just declare debug method here.
16460 * symtab.c (symbol_table::debug): Define.
16462 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
16464 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16466 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
16468 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16469 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16470 (*movdi_internal): Substitute Yi and Yj constraint with x
16471 and Ym and Yn constraint with y constraint. Update "isa"
16472 attribute and set "preferred_for_speed" attribute from
16473 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16474 (*movsi_internal): Ditto.
16475 (*movdf_internal): Ditto.
16476 (*movsf_internal): Ditto.
16477 (*zero_extendsidi2): Ditto.
16478 * config/i386/sse.md (vec_set<mode>_0): Ditto.
16479 (sse2_loadld): Ditto.
16480 (*vec_extract<ssevecmodelower>_0): Ditto.
16481 (*vec_extractv4si_0_zext_sse4): Ditto.
16482 (vec_concatv2di): Ditto.
16483 (*vec_dup<mode>): Ditto.
16484 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16485 * config/i386/constraints.md (Yi): Remove.
16490 2018-04-26 Nathan Sidwell <nathan@acm.org>
16492 * dumpfile.c (dump_open): New.
16493 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16494 (dump_finish): Detect stdio/stderr by value not name.
16496 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
16498 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16500 2018-04-26 Tom de Vries <tom@codesourcery.com>
16503 * config/nvptx/nvptx.c (verify_neutering_jumps)
16504 (verify_neutering_labels): New function
16505 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16507 2018-04-26 Tom de Vries <tom@codesourcery.com>
16510 * config/nvptx/nvptx.c (needs_neutering_p): New function.
16511 (nvptx_single): Use needs_neutering_p to skip over insns that do not
16514 2018-04-26 Richard Biener <rguenther@suse.de>
16515 Tom de Vries <tom@codesourcery.com>
16518 * lto-streamer-out.c (output_function): Fixup loops if required to match
16519 discovery done in the reader.
16521 2018-04-26 Richard Biener <rguenther@suse.de>
16523 PR tree-optimization/85116
16524 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16525 have a loop exit from the single latch predecessor. Remove
16526 case of header with just condition.
16527 (ch_base::copy_headers): Exclude infinite loops from any
16529 (pass_ch::execute): Record exits.
16531 2018-04-26 Richard Biener <rguenther@suse.de>
16533 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16534 prologue cost vector and pass it to vect_get_load_cost.
16535 (vect_get_peeling_costs_all_drs): Likewise.
16536 (vect_peeling_hash_get_lowest_cost): Likewise.
16537 (vect_enhance_data_refs_alignment): Likewise.
16539 2018-04-26 Richard Biener <rguenther@suse.de>
16541 PR middle-end/85450
16542 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16543 checking of integer<->pointer conversions.
16544 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16545 sign-/zero-extending pointer types.
16546 (expand_omp_for_static_chunk): Likewise.
16548 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
16549 Jean Lee <xiaoyur347@gmail.com>
16551 * config/mips/mips.c (mips_asan_shadow_offset): New function.
16552 (TARGET_ASAN_SHADOW_OFFSET): Define.
16553 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16554 true for -fsanitize=address.
16556 2018-04-25 Mark Wielaard <mark@klomp.org>
16558 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16561 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16563 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16564 than "alu", remove explicit "memory" and "imm_disp" attributes.
16565 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16567 PR middle-end/85414
16568 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16569 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16570 gen_lowpart_no_emit.
16572 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
16575 * config/i386/i386.c (ix86_expand_builtin): Change memory
16576 operand to XI, extend p0 to Pmode.
16577 * config/i386/i386.md: Change unspec volatile and operand
16578 1 mode to XI, change operand 0 mode to P.
16580 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16582 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16583 GET_MODE_MASK before any checking.
16584 (nds32_can_use_bset_p): Likewise.
16585 (nds32_can_use_btgl_p): Likewise.
16587 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16589 * config/nds32/nds32-doubleword.md: New define_split pattern for
16590 illegal register number.
16592 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16594 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16596 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16598 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16600 2018-04-25 Richard Biener <rguenther@suse.de>
16602 * lto-streamer.h (LTO_major_version): Bump to 8.
16604 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16606 * BASE-VER: Set to 9.0.0.
16608 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
16610 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16611 in __abskf2 and __powikf2.
16613 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16616 * config/aarch64/constraints.md (Usg, Usj): New constraints.
16617 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16618 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16619 Use the above on operand 2. Reindent.
16620 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16622 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16625 * common/config/i386/i386-common.c (ix86_handle_option): Don't
16627 * config/i386/i386.opt (mcet): Removed.
16628 * doc/install.texi: Remove -mcet documentation.
16629 * doc/invoke.texi: Likewise.
16631 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16634 * doc/install.texi: Remove -mcet from bootstrap-cet.
16636 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16639 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16640 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16644 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16645 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16646 containing a CONST_VECTOR.
16648 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
16650 * doc/install.texi: Update newlib dependency for nvptx.
16652 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16655 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16656 instead of INTVAL when shifting x left.
16658 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
16660 PR tree-optimization/85478
16661 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16662 vect_grouped_store_supported for single element vectors.
16664 2018-04-24 Richard Biener <rguenther@suse.de>
16667 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16668 load cost increase to the case of non-constant step.
16670 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16673 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16674 destination if any_malformed_asm.
16676 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
16678 PR middle-end/85496
16679 * expr.c (store_field): In the bitfield case, if the value comes from
16680 a function call and is returned in registers by means of a PARALLEL,
16681 do not change the mode of the temporary unless BLKmode and VOIDmode.
16683 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
16685 PR rtl-optimization/85423
16686 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16687 dependencies to debug insns when the previous insn is non-debug.
16689 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16691 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16692 enums into a single definition.
16693 (fls): Fix predicates and printing.
16696 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16698 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16699 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16700 and short u6 immediate.
16701 (check_if_valid_sleep_operand): Remove.
16702 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16704 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16706 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16707 flag_always_save_lp condition.
16708 * config/nds32/nds32.opt (malways-save-lp): New option.
16710 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16712 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16713 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16714 * config/nds32/nds32.h
16715 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16716 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16718 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16720 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16721 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16723 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16724 Chung-Ju Wu <jasonwucj@gmail.com>
16726 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16728 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16729 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16731 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16733 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16735 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16737 * config/nds32/nds32-protos.h (nds32_data_alignment,
16738 nds32_local_alignment): Declare.
16739 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16740 nds32_local_alignment): New functions.
16741 (TARGET_CONSTANT_ALIGNMENT): Define.
16742 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16744 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16746 * config/nds32/nds32.c
16747 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16748 (TARGET_MODES_TIEABLE_P): Likewise.
16750 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16752 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16753 level Ofast and Og.
16755 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
16756 Chung-Ju Wu <jasonwucj@gmail.com>
16758 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16759 for unaligned access.
16760 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16762 * config/nds32/nds32-intrinsic.md: Likewise.
16763 * config/nds32/nds32_intrinsic.h: Likewise.
16764 * config/nds32/nds32.h (nds32_builtins): Likewise.
16765 * config/nds32/nds32.opt (munaligned-access): New option.
16766 * config/nds32/nds32.c (nds32_asm_file_start): Display
16767 flag_unaligned_access status.
16769 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16771 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16772 -mno-relax is present.
16773 * config/riscv/linux.h (LINK_SPEC): Ditto.
16775 2018-04-20 Martin Sebor <msebor@redhat.com>
16778 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16780 (gimple_fold_builtin_stxcpy_chk): Same.
16781 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16783 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16786 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16787 __powikf2 when long double is IEEE 128-bit.
16789 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16791 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16792 step to make sure stack always aligned.
16794 2018-04-20 Carl Love <cel@us.ibm.com>
16797 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16798 size check for arg0.
16800 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16801 Tom de Vries <tom@codesourcery.com>
16804 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16805 Emit insns for calls too.
16806 (nvptx_find_par): Always look for worker-level predecessor insn.
16807 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
16809 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16810 (nvptx_process_pars): Propagate frames for calls.
16812 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
16815 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16817 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16818 (ix86_handle_option): Don't handle OPT_mibt.
16819 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16821 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16823 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16824 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16825 (ix86_target_macros): Define __CET__ with flag_cf_protection
16826 for -fcf-protection.
16827 * config/i386/i386.c (isa2_opts): Remove -mibt.
16828 * config/i386/i386.h (TARGET_IBT): Removed.
16829 (TARGET_IBT_P): Likewise.
16830 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16831 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16832 * config/i386/i386.opt (mcet): Update help message.
16833 (mshstk): Likewise.
16835 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
16836 -mcet as an alias for -mshstk.
16838 2018-04-20 Richard Biener <rguenther@suse.de>
16840 PR middle-end/85475
16841 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16842 complexity by forcing a single use of the multiply operand.
16844 2018-04-20 Martin Jambor <mjambor@suse.cz>
16847 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16848 recursion dependency to only apply to non-clones.
16850 2018-04-20 Martin Jambor <mjambor@suse.cz>
16853 * ipa-cp.c (create_specialized_node): Check that clones of
16854 self-recursive edges exist during IPA-CP.
16856 2018-04-19 Toon Moene <toon@moene.org>
16858 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16861 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16863 PR tree-optimization/85467
16864 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16865 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
16866 VECTOR_CST element to type.
16868 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16871 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16872 * config/i386/i386.md (builtin_setjmp_setup): Removed.
16873 (builtin_longjmp): Likewise.
16874 (save_stack_nonlocal): New pattern.
16875 (restore_stack_nonlocal): Likewise.
16877 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16880 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16881 Replace ASM_OUTPUT_LABEL with fprintf.
16883 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16886 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16887 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16888 * config/i386/i386-c.c (ix86_target_macros_internal): Also
16889 define __IBT__ and __SHSTK__ for -fcf-protection.
16890 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16892 (ix86_trampoline_init): Likewise.
16893 (x86_output_mi_thunk): Likewise.
16894 (ix86_notrack_prefixed_insn_p): Likewise.
16895 (ix86_option_override_internal): Don't disallow -fcf-protection.
16896 * config/i386/i386.md (rdssp<mode>): Also enable for
16898 (incssp<mode>): Likewise.
16899 (nop_endbr): Likewise.
16900 * config/i386/i386.opt (mcet): Change help message to built-in
16903 (mshstk): Likewise.
16904 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16905 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
16906 enable CET built-in functions.
16908 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
16910 * common/config/i386/i386-common.c
16911 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16912 OPTION_MASK_ISA_MOVDIRI_UNSET,
16913 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16914 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16915 * config.gcc (movdirintrin.h): New header.
16916 * config/i386/cpuid.h (bit_MOVDIRI,
16917 bit_MOVDIR64B): New bits.
16918 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16920 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16921 (VOID, PVOID, PCVOID)): New function types.
16922 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16923 __builtin_ia32_directstoreu_u64,
16924 __builtin_ia32_movdir64b): New builtins.
16925 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16926 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16928 (ix86_valid_target_attribute_inner_p): Ditto.
16929 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16930 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16931 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16932 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16933 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16934 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16935 (movdiri<mode>, movdir64b_<mode>): New.
16936 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16937 * config/i386/immintrin.h: Include movdirintrin.h.
16938 * config/i386/movdirintrin.h: New file.
16939 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16941 2018-04-19 Richard Biener <rguenther@suse.de>
16943 PR middle-end/85455
16944 * cfg.c (clear_bb_flags): When loop state says we have
16945 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16947 2018-04-19 Richard Biener <rguenther@suse.de>
16949 PR tree-optimization/84737
16950 * tree-vect-data-refs.c (vect_copy_ref_info): New function
16951 copying restrict info.
16952 (vect_setup_realignment): Use it.
16953 * tree-vectorizer.h (vect_copy_ref_info): Declare.
16954 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16955 the first DR to all generated stores.
16956 (vectorizable_load): Likewise for loads.
16958 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16960 PR tree-optimization/85446
16961 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16962 the integral and pointer types to have the same precision.
16964 * doc/install.texi: Document --disable-cet being the default and
16967 2018-04-18 Martin Liska <mliska@suse.cz>
16969 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16972 2018-04-18 Martin Liska <mliska@suse.cz>
16975 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16978 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16979 arguments if they are comparable.
16981 2018-04-18 Martin Liska <mliska@suse.cz>
16984 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16987 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16990 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
16993 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16994 ENDBR after calling __morestack.
16996 2018-04-18 David Malcolm <dmalcolm@redhat.com>
16999 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17000 by using gcc_base_ver to generate a gcc_driver_version, and use
17001 it when generating GCC_DRIVER_NAME.
17002 * configure: Regenerate.
17004 2018-04-18 Jakub Jelinek <jakub@redhat.com>
17007 * config.gcc: Obsolete powerpc*-*-*spe*.
17009 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17012 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17013 (stabstr_D): Change type of unum from unsigned int to
17014 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
17017 2018-04-17 Jim Wilson <jimw@sifive.com>
17020 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17021 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17022 Set arg_pointer_offset after using pretend_args_size.
17024 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17026 PR rtl-optimization/85431
17027 * dse.c (record_store): Ignore zero width stores.
17030 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
17031 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17032 __builtin_stack_restore rather than after it.
17033 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17034 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17035 argument instead of virtual_dynamic_stack_rtx.
17037 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
17039 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17041 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17042 Add note to error message to explain internal mapping of overloaded
17043 built-in function name to non-overloaded built-in function name.
17044 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17047 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17050 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17051 where the inputs overlap with the output.
17053 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17055 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17056 (=v, v) alternative and explicit "memory" attribute.
17057 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
17058 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17060 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17061 "sselog1" type instead of "sselog".
17062 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17063 "sselog". Remove explicit "memory" attribute.
17064 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17065 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17067 (vec_extract_hi_v32hi): Merge all alternatives into one, use
17068 "sselog1" type instead of "sselog". Remove explicit "memory"
17070 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17071 use "sselog1" type instead of "sselog". Remove explicit "memory"
17073 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17074 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17076 (vec_extract_hi_v64qi): Merge all alternatives into one, use
17077 "sselog1" type instead of "sselog". Remove explicit "memory"
17079 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17080 use "sselog1" type instead of "sselog". Remove explicit "memory"
17084 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17086 PR middle-end/85414
17087 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17090 2018-04-17 Martin Jambor <mjambor@suse.cz>
17093 * ipa-cp.c (create_specialized_node): Call
17094 expand_all_artificial_thunks if necessary.
17096 2018-04-17 Martin Liska <mliska@suse.cz>
17099 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17100 in message, remote space in between '_G' and '('.
17102 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17105 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17106 avx512f_vmcmp<mode>3<round_saeonly_name>,
17107 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17108 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17109 avx512f_rndscale<mode><round_saeonly_name>,
17110 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17111 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17112 Use %<iptr>2 instead of %2 for -masm=intel.
17113 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17114 avx512f_vcvttss2usi<round_saeonly_name>,
17115 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17117 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17118 avx512f_vcvttsd2usi<round_saeonly_name>,
17119 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17120 Use %q1 instead of %1 for -masm=intel.
17121 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17122 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17123 of %3 for -masm=intel.
17124 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17126 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17128 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17130 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17131 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17133 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17135 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17136 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17137 %g1 and one with %0 and %1.
17138 (avx512er_vmrcp28<mode><round_saeonly_name>,
17139 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17140 %1 for -masm=intel.
17141 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17142 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17143 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17144 of %0 and %{%4%} for -masm=intel.
17145 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17146 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17147 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17148 order of %0 and %{%5%}%{z%} for -masm=intel.
17150 2018-04-17 Jan Hubicka <jh@suse.cz>
17153 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17155 2018-04-17 Martin Liska <mliska@suse.cz>
17158 * multiple_target.c (create_dispatcher_calls): Set apostrophes
17159 for target_clone error message. Make default implementation
17160 clone to be a local declaration.
17161 (separate_attrs): Add new argument and check for an empty
17163 (expand_target_clones): Handle it.
17164 (ipa_target_clone): Make redirection just for target_clones
17167 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
17168 Tom de Vries <tom@codesourcery.com>
17170 PR middle-end/84955
17171 * omp-expand.c (expand_oacc_for): Add dummy false branch for
17172 tiled basic blocks without omp continue statements.
17174 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
17177 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17178 vec_extract expression as having side effects to make sure it gets
17181 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
17184 * config/i386/i386.c (get_builtin_code_for_version): Check
17187 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
17190 * config.gcc: Support "skylake".
17191 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17193 * config/i386/i386.c (m_SKYLAKE): Define.
17194 (processor_target_table): Add "skylake".
17195 (ix86_option_override_internal): Add "skylake".
17196 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17197 PROCESSOR_CANNONLAKE.
17198 (get_builtin_code_for_version): Fix priority for
17199 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17200 PROCESSOR_SKYLAKE-AVX512.
17201 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17202 (processor_type): Add PROCESSOR_SKYLAKE.
17204 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
17205 Jason Merrill <jason@redhat.com>
17208 * convert.c (convert_to_integer_1): Use direct recursion for
17209 enumeral types and types with a precision less than the number
17210 of bits in their mode.
17212 2018-04-16 Julia Koval <julia.koval@intel.com>
17215 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17216 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17218 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
17221 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17222 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17223 and -mno-direct-move.
17225 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
17228 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17229 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17230 Ensure negative shifts result in {0}.
17232 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
17234 PR rtl-optimization/79916
17235 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17236 regs (if any) to define how to gnerate SD moves when LRA is in
17239 2018-04-13 Jakub Jelinek <jakub@redhat.com>
17241 PR rtl-optimization/85393
17242 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17243 * except.c (expand_dw2_landing_pad_for_region): Make static.
17244 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17245 a label and unconditional jump to old_bb, rather than
17246 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17249 PR rtl-optimization/85376
17250 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17251 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17252 instead of a specific value.
17254 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17255 Bin Cheng <bin.cheng@arm.com>
17257 PR tree-optimization/82965
17258 PR tree-optimization/83991
17259 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17260 by_profile_only parameter.
17261 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17262 information if the loop was predicted to iterate too many times.
17263 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17265 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17268 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17271 2018-04-13 Martin Liska <mliska@suse.cz>
17272 Jakub Jelinek <jakub@redhat.com>
17274 PR middle-end/81657
17275 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17276 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17277 * builtins.c (expand_builtin_memory_copy_args): Use
17278 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17279 handle dest_addr == pc_rtx.
17281 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
17284 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17285 asked to not generate direct moves.
17286 (fix_trunc<mode>si2_stfiwx): Similar.
17287 (fix_trunc<mode>si2_internal): Similar.
17289 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17292 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17293 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17294 lookup if dest in some wider mode is known to be const0_rtx and
17295 if so, record permanent equivalence for it to be ZERO_EXTEND of
17296 the narrower mode destination.
17298 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17300 * lto-streamer-out.c (output_function): Revert 259346.
17301 * omp-expand.c (expand_oacc_for): Likewise.
17303 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
17305 PR rtl-optimization/85354
17306 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17307 * sel-sched.c (sel_global_init): ... here.
17309 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
17312 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17313 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17314 mode for PE-COFF targets.
17315 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17316 (i386_pe_asm_lto_end): Likewise.
17317 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17318 (TARGET_ASM_LTO_END): Likewise.
17319 * config/i386/winnt.c (saved_debug_info_level): New static variable.
17320 (i386_pe_asm_lto_start): New function.
17321 (i386_pe_asm_lto_end): Likewise.
17323 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17324 Richard Biener <rguenther@suse.de>
17326 PR middle-end/84955
17327 * lto-streamer-out.c (output_function): Fix CFG loop state before
17329 * omp-expand.c (expand_oacc_for): Handle calls to internal
17330 functions like regular functions.
17332 2018-04-12 Richard Biener <rguenther@suse.de>
17335 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17336 for the early LTO debug to properly generate references to it
17337 during DIE emission. Do not re-use that for the skeleton for
17339 (dwarf2out_early_finish): Likewise.
17341 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17344 * config/i386/sse.md
17345 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17346 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17347 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17348 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17349 and output is a reg, avoid creating invalid lowpart subreg, but
17350 instead split into a 512-bit move. Don't split if not AVX512VL,
17351 input is xmm16+ reg and output is a mem.
17352 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17353 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17354 xmm16+ reg and output is a mem.
17356 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17358 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17359 also for flag_dwarf2_cfi_asm.
17361 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17363 PR rtl-optimization/85342
17364 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17365 a bool scalar var inside of the loop instead. Don't try to update
17366 recog_data.operand after failed apply_change_group.
17368 2018-04-12 Tom de Vries <tom@codesourcery.com>
17371 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17372 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
17373 array with flexible array member as array without given dimension.
17374 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17375 argument for undefined param to true.
17377 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
17380 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17381 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17382 from PowerPC section.
17383 * config/rs6000/sysv4.opt (mcall-): Improve help text.
17384 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17385 help text that is too long.
17386 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17387 help text that is too long.
17388 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17389 help text that is too long.
17391 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
17393 * config/alpha/alpha.md (stack_probe_internal): Rename
17394 from "probe_stack". Update all callers.
17396 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17398 PR rtl-optimization/84566
17399 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17400 sched_macro_fuse_insns.
17402 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17405 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17406 (compute_block_dependences): ... from here.
17408 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17410 PR tree-optimization/85331
17411 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17412 from int to HOST_WIDE_INT.
17414 2018-04-11 Martin Jambor <mjambor@suse.cz>
17417 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17418 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17419 not the same as the source val.
17420 (cgraph_edge_brings_value_p): New parameter.
17421 (gather_edges_for_value): Pass destination value to
17422 cgraph_edge_brings_value_p.
17423 (perhaps_add_new_callers): Likewise.
17424 (get_info_about_necessary_edges): Likewise and exclude values brought
17425 only by self-recursive edges.
17426 (create_specialized_node): Redirect only clones of self-calling edges.
17427 (+self_recursive_pass_through_p): New function.
17428 (find_more_scalar_values_for_callers_subset): Use it.
17429 (find_aggregate_values_for_callers_subset): Likewise.
17430 (known_aggs_to_agg_replacement_list): Removed.
17431 (decide_whether_version_node): Re-calculate known constants for all
17432 remaining context clones.
17434 2018-04-11 Richard Biener <rguenther@suse.de>
17437 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17438 from early DWARF output.
17439 (dwarf2out_early_finish): Output line info unconditionally into
17440 early DWARF and add reference to it.
17442 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17445 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17446 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17447 other than V2DFmode using iptr mode attribute.
17448 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17450 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17452 PR rtl-optimization/84659
17453 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17455 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17458 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17460 (output_loc_list): Pass address of a dummy size variable even in the
17461 locview handling loop.
17462 (index_location_lists): Add comment on why skip_loc_list_entry can't
17465 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
17468 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17471 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
17474 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17475 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17476 and -mstring-compare-inline-limit.
17478 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17481 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17482 for stack clash protection in a register whenever we need it to be in
17485 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17487 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17488 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17490 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17493 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17495 (mlong-double-): Ditto.
17496 * config/rs6000/sysv4.opt (msdata=): Ditto.
17497 (mtls-size=): Ditto.
17499 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17501 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17502 erroneous entries for
17503 "vector int vec_ldl (int, long int *)", and
17504 "vector unsigned int vec_ldl (int, unsigned long int *)".
17505 Add comments and entries for
17506 "vector bool char vec_ldl (int, bool char *)",
17507 "vector bool short vec_ldl (int, bool short *)",
17508 "vector bool int vec_ldl (int, bool int *)",
17509 "vector bool long long vec_ldl (int, bool long long *)",
17510 "vector pixel vec_ldl (int, pixel *)",
17511 "vector long long vec_ldl (int, long long *)",
17512 "vector unsigned long long vec_ldl (int, unsigned long long *)".
17513 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17514 type tree bool_long_long_type_node and correct definition of
17515 bool_V2DI_type_node to make reference to this new type tree.
17516 (rs6000_mangle_type): Replace erroneous reference to
17517 bool_long_type_node with bool_long_long_type_node.
17518 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17519 comments to emphasize sign distinctions for char and int types and
17520 replace RS6000_BTI_bool_long constant with
17521 RS6000_BTI_bool_long_long constant. Also add comment to restrict
17522 use of RS6000_BTI_pixel.
17523 (bool_long_type_node): Remove this macro definition.
17524 (bool_long_long_type_node): New macro definition
17526 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17528 PR rtl-optimization/85300
17529 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17530 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17531 simplify_unary_operation fails.
17533 2018-04-10 Martin Liska <mliska@suse.cz>
17535 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17536 cgraph_edge and ipa_ref.
17538 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17542 * config/i386/sse.md
17543 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17544 computation of the VEC_MERGE selector from mask.
17545 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17546 Fix decoding of the VEC_MERGE selector into mask.
17548 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17550 PR tree-optimization/85286
17551 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17553 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17555 * final.c (final_1): Set insn_last_address as well as
17556 insn_current_address.
17558 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17561 * explow.c (emit_stack_probe): Call validize_mem on memory location
17562 before passing it to gen_probe_stack. Create address operand and
17563 legitimize it for the probe_stack_address case.
17565 2018-04-09 Jan Hubicka <jh@suse.cz>
17568 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17569 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17570 * tree.c (free_lang_data_in_type): Fix handling of binfos;
17572 (free_lang_data): Rebuild type inheritance graph.
17574 2018-04-09 Martin Sebor <msebor@redhat.com>
17576 * invoke.texi (-finline-small-functions): Mention other optimization
17578 (-findirect-inlining, -fpartial-inlining): Same.
17579 (-finline-functions-called-once): Same.
17580 (-freorder-blocks-and-partition): Same.
17582 2018-04-09 Jan Hubicka <jh@suse.cz>
17585 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17586 jumps; choose last target that matches the criteria (i.e.
17587 no partition changes for non-crossing jumps).
17588 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17589 support for redirecting crossing jumps to non-crossing.
17591 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
17593 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17594 also for naked functions.
17596 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
17598 * config/arc/arc.md (add_shift): New pattern.
17599 (add_shift2): Likewise.
17600 (sub_shift): Likewise.
17601 (sub_shift_cmp0_noout): Likewise.
17602 (compare_si_ashiftsi): Likewise.
17603 (xbfu_cmp0_noout): New combine pattern.
17604 (xbfu_cmp0"): Likewise.
17605 (movsi_set_cc_insn): Place the predicable variant first.
17606 (commutative_binary_cmp0_noout): Remove clobber.
17607 (commutative_binary_cmp0): New pattern.
17608 (noncommutative_binary_cmp0): Likewise.
17609 (noncommutative_binary_cmp0_noout): Likewise.
17610 (noncommutative_binary_comparison_result_used): Removed.
17611 (rsub_cmp0): New pattern.
17612 (rsub_cmp0_noout): Likewise.
17613 (extzvsi): Changed, keep only meaningful variants.
17614 (SQH, SEZ): New iterators.
17615 (SQH_postfix): New mode attribute.
17616 (SEZ_prefix): New code attribute.
17617 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17618 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17619 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17620 of numerical value.
17621 (noncommutative_operator): Check the availability of barrel
17624 2018-04-09 Richard Biener <rguenther@suse.de>
17626 PR tree-optimization/85284
17627 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17628 Only use the niter constraining form of simple_iv when the exit
17629 is always executed.
17631 2018-04-09 Tom de Vries <tom@codesourcery.com>
17634 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17635 (define_expand "*memory_barrier"): New define_expand.
17636 (define_insn "memory_barrier"): New insn.
17638 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17640 PR rtl-optimization/80463
17641 PR rtl-optimization/83972
17642 PR rtl-optimization/83480
17644 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17645 correct producer for the insn.
17646 (tidy_control_flow): Fixup seqnos in case of debug insns.
17648 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17650 PR rtl-optimization/83913
17652 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17653 different sched-times when merging exprs.
17655 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17657 PR rtl-optimization/83962
17659 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17660 tidy_fallthru_edge and tidy_control_flow.
17662 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17664 PR rtl-optimization/83530
17666 * sel-sched.c (force_next_insn): New global variable.
17667 (remove_insn_for_debug): When force_next_insn is true, also leave only
17668 next insn in the ready list.
17669 (sel_sched_region): When the region wasn't scheduled, make another pass
17670 over it with force_next_insn set to 1.
17672 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17674 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17676 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17677 for interrupt control.
17678 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17679 functions for interrupt control.
17680 * config/nds32/nds32-intrinsic.md: Likewise.
17681 * config/nds32/nds32_intrinsic.h: Likewise.
17682 * config/nds32/nds32.h (nds32_builtins): Likewise.
17684 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17686 * config/nds32/nds32.c (nds32_init_machine_status,
17687 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17688 strict_aligned_p field.
17689 (nds32_expand_to_rtl_hook): New function.
17690 (TARGET_EXPAND_TO_RTL_HOOK): Define.
17691 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17693 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17694 Chung-Ju Wu <jasonwucj@gmail.com>
17696 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17697 * config/nds32/nds32-n7.md: New file.
17698 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17699 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17701 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17702 * config/nds32/nds32.md (pipeline_model): Add n7.
17703 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17704 * config/nds32/pipelines.md: Include n7 settings.
17706 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17707 Chung-Ju Wu <jasonwucj@gmail.com>
17709 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17710 * config/nds32/nds32-e8.md: New file.
17711 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17712 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17714 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17715 * config/nds32/nds32.md (pipeline_model): Add e8.
17716 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17717 * config/nds32/pipelines.md: Include e8 settings.
17719 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17720 Chung-Ju Wu <jasonwucj@gmail.com>
17722 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17723 * config/nds32/nds32-n8.md: New file.
17724 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17725 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17727 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17728 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17729 * config/nds32/nds32.md (pipeline_model): Add n8.
17730 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17731 * config/nds32/pipelines.md: Include n8 settings.
17733 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17734 Chung-Ju Wu <jasonwucj@gmail.com>
17736 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17737 * config/nds32/nds32-n9-2r1w.md: New file.
17738 * config/nds32/nds32-n9-3r2w.md: New file.
17739 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17740 nds32_register_ports): New or modify for cpu n9.
17741 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17743 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17744 * config/nds32/nds32-utils.c: New file.
17745 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17746 TARGET_MUL_SLOW): Define.
17747 * config/nds32/nds32.md (pipeline_model): New attribute.
17748 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17749 New options that support cpu n9.
17750 * config/nds32/pipelines.md: Include n9 settings.
17751 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17753 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17755 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17756 information if necessary.
17757 (output_cond_branch_compare_zero): Likewise.
17758 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17759 (nds32_target_alignment): Refine for alignment.
17760 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17761 (FUNCTION_BOUNDARY): Modify.
17762 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17764 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17766 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17768 * config/nds32/constants.md (unspec_volatile_element): Add values for
17769 TLB operation and data prefetch.
17770 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17771 functions for TLB operation and data prefetch.
17772 * config/nds32/nds32-intrinsic.md: Likewise.
17773 * config/nds32/nds32_intrinsic.h: Likewise.
17774 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17775 (nds32_print_operand): Likewise.
17776 * config/nds32/nds32.h (nds32_builtins): Likewise.
17778 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17779 Andrew Pinski <pinsika@gcc.gnu.org>
17781 PR middle-end/82976
17782 * match.pd: Use constant_boolean_node of correct type instead of
17783 boolean_true_node or boolean_false_node for simplifying
17784 pointer comparisons to zero.
17786 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17788 PR tree-optimization/80021
17789 * tree.c (verify_type_variant): Make error call in verify_variant_match
17790 translatable and remove final full stop.
17792 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17794 * config/nds32/constants.md (unspec_volatile_element): Add
17795 UNSPEC_VOLATILE_EH_RETURN.
17796 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17797 nds32_output_stack_pop): Support dwarf exception handling process.
17798 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17799 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17800 exception handling process.
17801 (nds32_compute_stack_frame): Likewise.
17802 (nds32_return_addr_rtx): Likewise.
17803 (nds32_initial_elimination_offset): Likewise.
17804 (nds32_expand_prologue): Likewise.
17805 (nds32_expand_epilogue): Likewise.
17806 (nds32_dynamic_chain_address): New function.
17807 * config/nds32/nds32.h (machine_function): Add fields for dwarf
17808 exception handling.
17809 (DYNAMIC_CHAIN_ADDRESS): Define.
17810 (EH_RETURN_DATA_REGNO): Define.
17811 (EH_RETURN_STACKADJ_RTX): Define.
17812 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17813 patterns for dwarf exception handling.
17815 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17817 * config/nds32/nds32.h: Clean up obsolete macros.
17819 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17821 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17822 Add enum values for particular instructions.
17823 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17824 particular intrinsic functions.
17825 * config/nds32/nds32-intrinsic.md: Likewise.
17826 * config/nds32/nds32_intrinsic.h: Likewise.
17827 * config/nds32/nds32.h (nds32_builtins): Likewise.
17828 * config/nds32/nds32.md (type): Add pbsad and pbsada.
17829 (btst, ave): New patterns for particular instructions.
17831 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17833 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17834 Add enum values for atomic load/store and memory sync.
17835 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17837 * config/nds32/nds32-intrinsic.md: Likewise.
17838 * config/nds32/nds32_intrinsic.h: Likewise.
17839 * config/nds32/nds32.h (nds32_builtins): Likewise.
17841 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17843 PR tree-optimization/85257
17844 * fold-const.c (native_encode_vector): If not all elts could fit
17845 and off is -1, return 0 rather than offset.
17846 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17847 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17848 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
17849 adjust buffer in native_interpret_expr call.
17851 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17853 * config/nds32/constants.md (unspec_volatile_element): Add cache
17854 control enum values.
17855 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17856 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17857 * config/nds32/nds32.c (nds32_cctl_names): New.
17858 (nds32_print_operand): Handle cache control register names.
17859 * config/nds32/nds32.h (nds32_builtins): New enum values.
17860 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17862 * config/nds32/nds32.md (type): Add mmu.
17863 * config/nds32/pipelines.md (simple_insn): Add mmu.
17865 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17867 * config/nds32/nds32.md (type): Remove call.
17868 * config/nds32/pipelines.md (simple_insn): Likewise.
17870 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17872 * config/nds32/constants.md (unspec_volatile_element): Add
17873 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17874 UNSPEC_VOLATILE_FMFCFG.
17875 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17876 description for fmfcfg and fmfcsr.
17877 (bdesc_1arg): Add fmtcsr.
17878 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17879 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17880 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17881 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17882 unspec_fmfcfg): New patterns.
17883 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17884 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17885 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17886 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17887 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17888 __nds32__fmfcfg): Define.
17890 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17892 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17893 intrinsic register names.
17894 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17895 intrinsic register enum values and macros.
17897 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17899 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17900 for load/store addressing form.
17901 (nds32_print_operand_address): Likewise.
17903 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
17906 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17907 based on LABEL_REF. Remove useless assertion.
17908 (pic_address_needs_scratch): Fix formatting.
17909 (sparc_legitimize_pic_address): Minor tweaks.
17910 (sparc_delegitimize_address): Adjust assertion accordingly.
17911 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17912 into symbolic_operand.
17913 (movsi_high_pic_label_ref): Likewise.
17914 (movsi_lo_sum_pic_label_ref): Likewise.
17915 (movdi_pic_label_ref): Likewise.
17916 (movdi_high_pic_label_ref): Likewise.
17917 (movdi_lo_sum_pic_label_ref): Likewise.
17919 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
17921 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17922 custom LIB_SPEC setup.
17924 2018-04-06 Ruslan Bukin <br@bsdpad.com>
17925 Kito Cheng <kito.cheng@gmail.com>
17927 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17928 * config/riscv/freebsd.h: New.
17930 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17932 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17933 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17936 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17937 Kito Cheng <kito.cheng@gmail.com>
17939 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17940 nds32_output_call, nds32_symbol_binds_local_p): New functions.
17941 * config/nds32/nds32-protos.h (nds32_output_call,
17942 nds32_output_return): Declare.
17943 * config/nds32/nds32.md: Refine all the call and return patterns.
17945 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17948 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17949 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17951 PR rtl-optimization/84872
17952 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17953 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17954 EDGE_CROSSING edge.
17956 2018-04-06 Tamar Christina <tamar.christina@arm.com>
17958 * expr.c (copy_blkmode_to_reg): Revert 254862.
17959 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17961 2018-04-06 Richard Biener <rguenther@suse.de>
17963 PR middle-end/85244
17964 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17965 after seeing a component reference with an adjacent field. Treat
17966 refs to arrays at struct end of external decls similar to
17967 refs to unconstrained commons.
17969 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17972 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17973 look through SAVE_EXPRs with non-side-effects argument. Adjust
17975 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17978 2018-04-06 Richard Biener <rguenther@suse.de>
17980 PR middle-end/85180
17981 * alias.c (find_base_term): New wrapper around find_base_term
17982 unwinding CSELIB_VAL_PTR changes.
17983 (find_base_term): Do not restore CSELIB_VAL_PTR during the
17986 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17988 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17990 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17991 constant definitions.
17992 ("nop"): lr 0,0 -> nopr r0
17993 ("nop_lr0", "nop_lr1"): New insn definitions.
17995 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17997 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17998 NDS32_V3PUSH_AVAILABLE_P macro.
18000 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
18001 Chung-Ju Wu <jasonwucj@gmail.com>
18003 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18004 (nds32*-*-*): Add float and fpu_config into supported_defaults.
18005 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18006 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18007 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18008 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18009 * config/nds32/constraints.md: New constraints and checking for hard
18010 float configuration.
18011 * config/nds32/iterators.md: New mode iterator and attribute for hard
18012 float configuration.
18013 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18015 * config/nds32/nds32-fpu.md: New file.
18016 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18017 deal with hard float code generation.
18018 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18020 (abi_type, float_reg_number): New enum type.
18021 * config/nds32/nds32-predicates.c: New predicates for hard float.
18022 * config/nds32/nds32-protos.h: Declare functions for hard float.
18023 * config/nds32/nds32.c: Implementation for hard float configuration.
18024 * config/nds32/nds32.h: Definitions for hard float configuration.
18025 * config/nds32/nds32.md: Include hard float machine description and
18026 modify patterns for hard float configuration.
18027 * config/nds32/nds32.opt: New options for hard float configuration.
18028 * config/nds32/predicates.md: New predicates for hard float
18031 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18033 * common/config/nds32/nds32-common.c
18034 (nds32_option_optimization_table): Enable -mreleax-hint by default.
18036 2018-04-05 Jakub Jelinek <jakub@redhat.com>
18038 PR middle-end/85195
18039 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18040 CONSTRUCTOR_ELT (ctor, ...)->value.
18042 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
18045 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18047 2018-04-05 Tom de Vries <tom@codesourcery.com>
18050 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18053 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18054 Kito Cheng <kito.cheng@gmail.com>
18056 * config/nds32/constraints.md (U33): Fine-tune checking condition.
18057 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18058 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18059 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18061 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18062 Kito Cheng <kito.cheng@gmail.com>
18064 * config/nds32/constraints.md (Ufe): New memory constraint.
18065 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18066 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18067 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18069 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18070 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18072 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18074 * config/nds32/nds32.md: Use optimize_size in the condition for
18075 alu-shift instructions.
18077 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18079 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18081 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18083 * config/nds32/nds32.md (negsi2): Refine pattern.
18085 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18086 Chung-Ju Wu <jasonwucj@gmail.com>
18088 * config/nds32/iterators.md (shift_rotate): New code iterator.
18089 (shift): New code attribute.
18090 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18091 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18092 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18093 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18094 bit-wise operations.
18095 (andsi3, *andsi3): Ditto.
18096 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18097 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18098 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18099 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18100 nds32_ior_operand, nds32_xor_operand): New predicates.
18102 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18104 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18105 (addsi3, subsi3): ... this.
18107 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18109 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18111 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18113 * config/nds32/nds32.md: Adjust indention.
18115 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18117 * config/nds32/nds32.md (feature): New attribute.
18119 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18121 * config/nds32/nds32.md (subtype): New attribute.
18123 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
18126 * config/arm/arm-builtins.c (arm_expand_builtin): Change
18127 expansion to perform a bitwise AND of the argument followed by a
18128 boolean negation of the result.
18130 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
18132 PR rtl-optimization/84878
18133 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18134 the basic block. Assert the use reference is not artificial and that
18135 it has an associated insn.
18137 2018-04-04 Michael Matz <matz@suse.de>
18139 * builtins.c (compute_objsize): Pass correct operand
18140 to array_at_struct_end_p.
18142 2018-04-04 Richard Biener <rguenther@suse.de>
18145 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18146 from contexts for DINFO_LEVEL_TERSE and below.
18148 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18150 * config/nds32/nds32-doubleword.md (move_<mode>): Require
18151 resiter_operand condition.
18152 * config/nds32/nds32.md (*move<mode>): Ditto.
18154 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18155 Monk Chiang <sh.chiang04@gmail.com>
18157 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18159 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18161 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18163 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18164 Kito Cheng <kito.cheng@gmail.com>
18166 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18167 nds32_cond_code_str, output_cond_branch,
18168 output_cond_branch_compare_zero, nds32_expand_cbranch,
18169 nds32_expand_cstore, nds32_expand_movcc,
18170 nds32_output_cbranchsi4_equality_zero,
18171 nds32_output_cbranchsi4_equality_reg,
18172 nds32_output_cbranchsi4_equality_reg_or_const_int,
18173 nds32_output_cbranchsi4_greater_less_zero: New functions.
18174 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18175 nds32_expand_cstore, nds32_expand_movcc,
18176 nds32_output_cbranchsi4_equality_zero,
18177 nds32_output_cbranchsi4_equality_reg,
18178 nds32_output_cbranchsi4_equality_reg_or_const_int,
18179 nds32_output_cbranchsi4_greater_less_zero): Declare.
18180 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18181 nds32_rimm11s_operand): New predicates.
18182 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18183 * config/nds32/nds32.md: Rewrite all the branch and conditional move
18186 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18188 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18189 * config/nds32/nds32.md: Ditto.
18190 * config/nds32/pipelines.md: Ditto.
18192 2018-04-04 Richard Biener <rguenther@suse.de>
18194 PR tree-optimization/85168
18195 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18196 propagating abnormals.
18198 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18200 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18202 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18203 Kito Cheng <kito.cheng@gmail.com>
18205 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18206 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18207 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18208 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18209 * config/nds32/nds32.md (sibcall_internal): New.
18210 (sibcall_register): Remove.
18211 (sibcall_immediate): Remove.
18212 (sibcall_value_internal): New.
18213 (sibcall_value_register): Remove.
18214 (sibcall_value_immediate): Remove.
18215 * config/nds32/predicates.md (nds32_general_register_operand): New.
18216 (nds32_call_address_operand): New.
18218 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18220 PR rtl-optimization/85167
18221 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18222 bb_defs if *split_p, instead preinitialize it to NULL.
18224 PR tree-optimization/85156
18225 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18226 evaluating the argument multiple times.
18228 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
18230 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18232 (_mm_cvtpd_ps): Likewise.
18233 (_mm_cvttpd_epi32): Likewise.
18234 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18235 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18236 vector, pixel, and bool following altivec.h include.
18238 2018-04-03 Martin Sebor <msebor@redhat.com>
18240 * doc/extend.texi (Common Function Attributes): Clarify.
18241 (const attribute): Likewise.
18242 (pure attribute): Likewise.
18244 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18247 * config/i386/i386.c (ix86_expand_vector_set): Use
18248 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
18250 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
18252 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18253 instructions when changing rounding bits to preserve precision bits
18254 in the x87 control word.
18256 2018-04-03 Martin Liska <mliska@suse.cz>
18258 PR tree-optimization/82491
18259 * rtl.h (strip_offset_and_add): Replace += suboffset with
18260 poly_uint64 () + suboffset.
18262 2018-03-29 Martin Liska <mliska@suse.cz>
18263 Martin Jambor <mjambor@suse.cz>
18266 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18267 param_type is not an integral or pointer type.
18269 2018-04-03 Richard Biener <rguenther@suse.de>
18271 * sese.h (recompute_all_dominators): Remove.
18273 2018-04-02 Martin Sebor <msebor@redhat.com>
18275 * doc/invoke.texi (-Wrestrict): Fix typos.
18277 2018-04-02 Jim Wilson <jimw@sifive.com>
18279 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18280 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18281 (<optab>di3, <optab>si3_extend): Likewise.
18282 (<optab>si3_mask, <optab>si3_mask_1): New.
18283 (<optab>di3_mask, <optab>di3_mask_1): New.
18284 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18285 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18286 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18288 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
18290 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18293 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
18295 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18296 (nds32_canonicalize_comparison): New function.
18298 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18299 Kito Cheng <kito.cheng@gmail.com>
18300 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18302 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18303 * config/nds32/constants.md (unspec_volatile_element): Add
18304 UNSPEC_VOLATILE_RELAX_GROUP.
18305 * config/nds32/nds32-relax-opt.c: New file.
18306 * config/nds32/nds32-predicates.c
18307 (nds32_symbol_load_store_p): New function.
18308 * config/nds32/nds32-protos.h
18309 (nds32_symbol_load_store_p): Declare function.
18310 (make_pass_nds32_relax_opt): Declare new rtl pass function.
18311 * config/nds32/nds32.c
18312 (nds32_register_pass): New function to register pass.
18313 (nds32_register_passes): New function to register passes.
18314 * config/nds32/nds32.md (relax_group): New pattern.
18315 * config/nds32/nds32.opt (mrelax-hint): New option.
18316 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18318 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
18320 * config/nds32/t-nds32: Modify files dependency.
18322 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18324 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18325 (PROFILE_HOOK): Define its implementation.
18327 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18329 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18330 type and 32-bit size.
18332 2018-04-01 Jakub Jelinek <jakub@redhat.com>
18334 PR middle-end/85090
18335 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18336 (V_128_256): New mode iterator.
18337 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18338 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18339 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18341 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18342 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18344 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
18347 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18348 NaN inputs correctly.
18350 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
18353 * config/rs6000/vsx.md (??r): New mode attribute.
18354 (*vsx_mov<mode>_64bit): Use it.
18355 (*vsx_mov<mode>_32bit): Likewise.
18357 2018-03-30 Martin Sebor <msebor@redhat.com>
18359 PR tree-optimization/84818
18360 * builtins.c (check_access): Use warning_n.
18362 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18365 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18367 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18370 2018-03-30 Julia Koval <julia.koval@intel.com>
18373 * x86-tune.def (movx, partial_reg_dependency): Enable for
18376 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
18378 PR inline-asm/84985
18379 * lra-constraints.c (process_alt_operands): Move setting
18380 this_alternative_matches below.
18382 2018-03-29 Martin Liska <mliska@suse.cz>
18385 * doc/invoke.texi: Document how LTO works with debug info.
18386 Describe auto-load support of binutils. Mention 'x86-64'
18387 as valid option value of -march option.
18389 2018-03-29 Jakub Jelinek <jakub@redhat.com>
18391 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18394 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18395 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18396 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18399 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18402 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18403 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18404 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18405 for RS6000_BTM_POWERPC64.
18406 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18407 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18408 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18411 (DIVDEU): Likewise.
18413 2018-03-28 Carl Love <cel@us.ibm.com>
18416 2017-09-27 Carl Love <cel@us.ibm.com>
18418 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18419 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18420 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18423 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18425 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18426 instead of __vector bool.
18427 (_mm_max_pu8): Likewise.
18428 (_mm_min_pi16): Likewise.
18430 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18433 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18434 (DIVWEUO): Likewise.
18435 (DIVDEO): Likewise.
18436 (DIVDEUO): Likewise.
18437 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18438 DIVWEUO and DIVDEUO.
18439 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18440 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18441 (div_extend): Likewise.
18442 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18444 (__builtin_divweuo): Likewise.
18445 (__builtin_divdeo): Likewise.
18446 (__builtin_divdeuo): Likewise.
18448 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18451 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18452 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18454 PR tree-optimization/82004
18455 * gimple-match-head.c (optimize_pow_to_exp): New function.
18456 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18457 Don't fold to exp if optimize_pow_to_exp is false.
18459 2018-03-28 Martin Liska <mliska@suse.cz>
18462 * calls.c (initialize_argument_information): Fix trailing space.
18463 * common.opt: Fix typo and provide better explanation for
18464 -fsanitize-coverage option.
18465 * config/i386/i386.opt: Fix typo.
18467 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18468 Martin Liska <mliska@suse.cz>
18471 * gimplify.c (asan_poison_variable): Don't do the check for
18472 gimplify_omp_ctxp here.
18473 (gimplify_decl_expr): Do it here.
18474 (gimplify_target_expr): Likewise.
18476 2018-03-28 Martin Liska <mliska@suse.cz>
18479 * config/i386/i386.c (ix86_function_arg_advance): Do not call
18480 chkp_type_bounds_count if MPX is not enabled.
18482 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
18484 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18486 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
18489 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18490 function to create the function decl for complex long double
18491 multiply and divide for -mabi=ieeelongdouble.
18492 (init_float128_ieee): Call it.
18494 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
18497 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18498 -fcf-protection=branch -mibt.
18499 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18501 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18504 * config/arm/arm.c (arm_valid_symbolic_address): Handle
18505 arm_word_relocations.
18507 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
18510 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18511 extern array declarations.
18513 2018-03-27 Richard Biener <rguenther@suse.de>
18515 PR middle-end/84067
18516 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18517 explicit single_use checks.
18519 2018-03-27 Richard Biener <rguenther@suse.de>
18521 PR tree-optimization/85082
18522 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18525 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18527 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18528 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18529 Turn on fasynchronous-unwind-tables and funwind-tables.
18531 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
18534 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18535 (*bmi_blsr_<mode>_ccz): Ditto.
18537 2018-03-26 Tom de Vries <tom@codesourcery.com>
18539 PR tree-optimization/85063
18540 * omp-general.c (offloading_function_p): New function. Factor out
18542 * omp-offload.c (pass_omp_target_link::gate): ... here.
18543 * omp-general.h (offloading_function_p): Declare.
18544 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18545 with attribute omp declare target for offloading functions.
18547 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
18549 PR tree-optimization/84005
18550 * tree-data-ref.h (get_base_for_alignment): Declare.
18551 * tree-data-ref.c (get_base_for_alignment_1): New function.
18552 (get_base_for_alignment): Likewise.
18553 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18554 get_base_for_alignment to find a suitable base object, instead
18555 of always using drb->base_address.
18557 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18559 PR inline-asm/85022
18560 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18561 known size by default.
18563 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
18565 PR inline-asm/85030
18566 * lra-constraints.c (process_alt_operands): Don't match BLKmode
18567 and non BLKmode operands.
18569 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18572 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18573 Clean up attributes.
18575 2018-03-23 Richard Biener <rguenther@suse.de>
18578 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18579 we are going to emit early debug for LTO.
18581 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18583 PR inline-asm/85034
18584 * function.c (match_asm_constraints_1): Don't optimize if input
18585 doesn't satisfy general_operand predicate for output's mode.
18587 PR inline-asm/85022
18588 * alias.c (write_dependence_p): Don't require for x_canonicalized
18589 non-VOIDmode if x has VOIDmode.
18592 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18593 just don't try to optimize it rather than assert it never happens.
18595 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18597 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18598 macro expansions for definition of ST_INTERNAL_<mode> and
18599 LD_INTERNAL_<mode> builtins.
18600 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18602 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18604 (altivec_expand_st_builtin): Likewise.
18605 (altivec_expand_builtin): Remove calls to deleted functions.
18606 (rs6000_address_for_altivec): Delete this function.
18607 * config/rs6000/vector.md: Remove expands for
18608 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18610 2018-03-22 Sudakshina Das <sudi.das@arm.com>
18613 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18614 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18615 re-computing once computed.
18616 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18617 (arm_init_machine_status): Initialize
18618 machine->static_chain_stack_bytes.
18620 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18623 * doc/extend.texi: Add four new prototypes for vec_ld.
18624 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18625 definitions for more logical presentation.
18626 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18627 entries for V1TI variants of __builtin_altivec_ld builtin.
18628 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18629 handling of V1TI variant of LVX icode pattern.
18630 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18631 (rs6000_gimple_fold_builtin): Likewise.
18632 (altivec_init_builtins): Add code to define
18633 __builtin_altivec_lvx_v1ti function.
18635 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18637 PR inline-asm/84941
18638 * function.c (match_asm_constraints_1): Don't do the optimization
18639 if input isn't a REG, SUBREG, MEM or constant.
18641 2018-03-22 Tom de Vries <tom@codesourcery.com>
18643 PR tree-optimization/84956
18644 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18645 bb_has_abnormal_pred.
18647 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18650 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18651 DECL_INITIAL (decl) to decl at the end.
18652 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18653 adjust the comment.
18655 2018-03-21 Joseph Myers <joseph@codesourcery.com>
18657 * doc/extend.texi (__builtin_tgmath): Document when complex
18658 integer types are treated as _Complex _Float64.
18660 2018-03-21 Tom de Vries <tom@codesourcery.com>
18662 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18664 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18666 PR tree-optimization/84960
18667 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18668 if it is ENTRY block, move them into single succ of ENTRY in that case.
18670 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
18672 PR tree-optimization/84811
18673 * poly-int.h (poly_span_traits): Remove the T3 parameter and
18674 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18675 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18676 (known_subrange_p): Update accordingly. Cast each value involved
18677 in the size comparison, rather than casting the result of the
18680 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18682 PR tree-optimization/84982
18683 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18684 by flipping the least significant bit rather than all bits from
18685 bitpos to bitpos + bitsize - 1.
18687 2018-03-21 Nathan Sidwell <nathan@acm.org>
18689 * doc/extend.texi (Deprecated Features): Remove mention of
18690 long-deleted deprecations.
18692 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18695 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18696 * configure: Regenerate.
18698 2018-03-21 Tom de Vries <tom@codesourcery.com>
18700 PR tree-optimization/83126
18701 * tree-parloops.c (num_phis): New function.
18702 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18704 2018-03-21 Nathan Sidwell <nathan@acm.org>
18706 * doc/extend.texi (Deprecated Features): Update deprecated flags,
18707 mention anon-struct/union members and trailing attributes.
18709 2018-03-21 Bin Cheng <bin.cheng@arm.com>
18711 PR tree-optimization/84969
18712 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18713 builtin memset partitions if they set different rhs values.
18715 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18717 PR rtl-optimization/84989
18718 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18719 VEC_DUPLICATE with scalar result mode.
18721 2018-03-21 Martin Liska <mliska@suse.cz>
18724 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18725 not intended return statement.
18727 2018-03-21 Martin Liska <mliska@suse.cz>
18730 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18731 (chkp_find_bound_slots_1): Limit number of iterations.
18733 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
18736 * Minor grammar fixes for x86 options.
18738 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18741 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18742 holding REG_CFA_RESTORE notes, instead turn them into a USE.
18744 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
18747 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18748 (altivec_lvx_<mode>_1op): Likewise.
18749 (altivec_stvx_<mode>_2op): Likewise.
18750 (altivec_stvx_<mode>_1op): Likewise.
18751 (altivec_lvx_<VM2:mode>): New define_expand.
18752 (altivec_stvx_<VM2:mode>): Likewise.
18753 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18754 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18755 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18756 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18757 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18758 (rs6000_gen_lvx): Likewise.
18759 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18760 (altivec_expand_stv_builtin): Likewise.
18761 (altivec_expand_builtin): Likewise.
18762 * config/rs6000/vector.md: Likewise.
18764 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18767 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18770 2018-03-20 Richard Biener <rguenther@suse.de>
18773 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18774 sign-conversions as zero, fall back to standard scalar_stmt
18777 2018-03-20 Martin Liska <mliska@suse.cz>
18780 * predict.c (rebuild_frequencies): Handle case when we have
18781 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18783 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18786 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18787 flag_section_anchors.
18788 * varasm.c (use_blocks_for_decl_p): Remove hack for
18789 dw2_force_const_mem.
18792 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18794 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18795 be created, use lowpart_subreg of operands[0] rather than operands[0]
18797 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18798 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18799 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18800 and n constraint instead of aarch64_shift_imm_di and Usd.
18801 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18802 (*aarch64_<optab>_reg_minus<mode>3): ... this.
18804 2018-03-20 Sudakshina Das <sudi.das@arm.com>
18807 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18808 to favor GPR over NEON registers.
18809 (<shift>di3_neon): Likewise.
18811 2018-03-20 Tom de Vries <tom@codesourcery.com>
18814 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18815 (nvptx_process_pars): Emit bar.sync asap and alap.
18817 2018-03-20 Tom de Vries <tom@codesourcery.com>
18820 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18821 seen_label if seen_label is already set.
18823 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18826 * config/i386/i386.c (fold_builtin_cpu): For features above 31
18827 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18828 Use 1U instead of 1. Formatting fixes.
18831 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18832 instead of TREE_TYPE (s1) for the return value.
18834 2018-03-19 Jakub Jelinek <jakub@redhat.com>
18836 PR tree-optimization/84946
18837 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18838 bitsize + bitsize in poly_uint64 rather than poly_int64.
18841 * dwarf2asm.c: Include fold-const.c.
18842 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18843 of decl rather than decl itself.
18845 PR rtl-optimization/84643
18846 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18848 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18851 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18852 calling assemble_variable.
18854 2018-03-19 Sudakshina Das <sudi.das@arm.com>
18857 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18858 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18860 2018-03-19 Jim Wilson <jimw@sifive.com>
18863 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18864 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18865 (riscv_first_stack_step): Likewise.
18866 (riscv_option_override): Use STACK_BOUNDARY instead of
18867 MIN_STACK_BOUNDARY.
18868 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18869 MIN_STACK_BOUNDARY.
18870 (BIGGEST_ALIGNMENT): Set to 128.
18871 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18872 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18875 2018-03-19 Richard Biener <rguenther@suse.de>
18877 PR tree-optimization/84933
18878 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18879 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18881 2018-03-19 Richard Biener <rguenther@suse.de>
18883 PR tree-optimization/84859
18884 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18885 (cond_if_else_store_replacement): Perform sinking operation on
18886 single-store BBs regardless of MAX_STORES_TO_SINK setting.
18887 Generalize what a BB with a single eligible store is.
18889 2018-03-19 Richard Biener <rguenther@suse.de>
18891 PR tree-optimization/84929
18892 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18893 chrec_is_positive against non-chrec arg.
18895 2018-03-19 Tamar Christina <tamar.christina@arm.com>
18898 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18900 2018-03-18 Martin Liska <mliska@suse.cz>
18902 PR rtl-optimization/84635
18903 * regrename.c (build_def_use): Use matches_mode only when
18906 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
18908 PR tree-optimization/84913
18909 * tree-vect-loop.c (vectorizable_reduction): Don't try to
18910 vectorize chains of COND_EXPRs.
18912 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18914 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18916 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18918 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18920 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18922 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18924 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18925 Kito Cheng <kito.cheng@gmail.com>
18927 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18928 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18929 (nds32_adjust_reg_alloc_order): New function.
18930 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18932 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
18934 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18935 nds32_print_operand, nds32_print_operand_address): Use
18936 HOST_WIDE_INT_PRINT_DEC instead.
18938 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18940 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18942 2018-03-17 Jakub Jelinek <jakub@redhat.com>
18945 * config/i386/i386.c (initial_ix86_tune_features,
18946 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18947 unsigned long long.
18948 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18949 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18950 rather than 1u << ix86_tune. Formatting fix.
18951 (ix86_option_override_internal): Change ix86_arch_mask from
18952 unsigned int to unsigned HOST_WIDE_INT, initialize to
18953 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18954 (ix86_function_specific_restore): Likewise.
18956 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18959 * postreload.c (reload_combine_recognize_pattern): Perform
18960 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18961 truncate_int_for_mode the result for the destination's mode.
18964 * hsa-gen.c (mem_type_for_type): Fix comment typo.
18965 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18967 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18970 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
18973 * lra-assigns.c (lra_split_hard_reg_for): Don't use
18974 regno_allocno_class_array and sorted_pseudos.
18975 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18976 insns where regno is used.
18978 2018-03-16 Martin Liska <mliska@suse.cz>
18981 * multiple_target.c (create_dispatcher_calls): Redirect
18982 reference in the symbol table.
18984 2018-03-16 Martin Liska <mliska@suse.cz>
18987 * multiple_target.c (create_dispatcher_calls): Redirect also
18990 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18994 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18995 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18996 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18998 2018-03-16 Julia Koval <julia.koval@intel.com>
19000 * doc/invoke.texi (Skylake Server): Add CLWB.
19001 Cannonlake): Remove CLWB.
19003 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19005 PR tree-optimization/84841
19006 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19008 (FLOAT_ONE_CONST_TYPE): Define.
19009 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19010 (sort_by_operand_rank): Put entries with higher constant_type last
19011 rather than first to match comments.
19013 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
19015 * config/nios2/nios2.md (movsi_internal): Fix thinko in
19018 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19021 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19022 check and use address of target always.
19024 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
19027 * config/i386/i386.c (indirect_thunk_needed): Update comments.
19028 (indirect_thunk_bnd_needed): Likewise.
19029 (indirect_thunks_used): Likewise.
19030 (indirect_thunks_bnd_used): Likewise.
19031 (indirect_return_needed): New.
19032 (indirect_return_bnd_needed): Likewise.
19033 (output_indirect_thunk_function): Add a bool argument for
19035 (output_indirect_thunk_function): Don't generate alias for
19036 function return thunk.
19037 (ix86_code_end): Call output_indirect_thunk_function to generate
19038 function return thunks.
19039 (ix86_output_function_return): Set indirect_return_bnd_needed
19040 and indirect_return_needed instead of indirect_thunk_bnd_needed
19041 and indirect_thunk_needed.
19043 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
19045 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19046 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19047 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19049 2018-03-15 David Malcolm <dmalcolm@redhat.com>
19050 Paul Hua <paul.hua.gm@gmail.com>
19053 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19055 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
19057 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19058 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19059 resp. SFmode cases.
19061 2018-03-15 Tamar Christina <tamar.christina@arm.com>
19064 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19065 instead of GET_MODE_SIZE when comparing Units.
19067 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19070 * varasm.c (hash_section): Return an unchangeble hash value
19071 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19072 Return !aarch64_can_use_per_function_literal_pools_p ().
19074 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19077 * optabs.c (emit_conditional_move): Pass address of cmode's copy
19078 rather than address of cmode as last argument to prepare_cmp_insn.
19080 2018-03-15 Julia Koval <julia.koval@intel.com>
19082 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19083 F_AVX512VNNI, F_AVX512BITALG): New.
19085 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
19088 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
19089 insn for floating-point loads and stores.
19091 2018-03-14 Carl Love <cel@us.ibm.com>
19093 * config/rs6000/rs6000-c.c: Add macro definitions for
19094 ALTIVEC_BUILTIN_VEC_PERMXOR.
19095 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19096 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19097 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19098 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19100 * config/doc/extend.texi: Add prototypes for vec_permxor.
19102 2018-03-14 David Malcolm <dmalcolm@redhat.com>
19105 * diagnostic-show-locus.c (class layout_point): Convert m_line
19106 from int to linenum_type.
19107 (line_span::comparator): Use linenum "compare" function when
19108 comparing line numbers.
19109 (test_line_span): New function.
19110 (layout_range::contains_point): Convert param "row" from int to
19112 (layout_range::intersects_line_p): Likewise.
19113 (layout::will_show_line_p): Likewise.
19114 (layout::print_source_line): Likewise.
19115 (layout::should_print_annotation_line_p): Likewise.
19116 (layout::print_annotation_line): Likewise.
19117 (layout::print_leading_fixits): Likewise.
19118 (layout::annotation_line_showed_range_p): Likewise.
19119 (struct line_corrections): Likewise for field m_row.
19120 (line_corrections::line_corrections): Likewise for param "row".
19121 (layout::print_trailing_fixits): Likewise.
19122 (layout::get_state_at_point): Likewise.
19123 (layout::get_x_bound_for_row): Likewise.
19124 (layout::print_line): Likewise.
19125 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19126 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19127 * input.c (selftest::test_linenum_comparisons): New function.
19128 (selftest::input_c_tests): Call it.
19129 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19130 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19131 * selftest.h (ASSERT_GT): New macro.
19132 (ASSERT_GT_AT): New macro.
19133 (ASSERT_LT): New macro.
19134 (ASSERT_LT_AT): New macro.
19136 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
19138 PR rtl-optimization/84780
19139 * combine.c (distribute_links): Don't make a link based on pc_rtx.
19141 2018-03-14 Martin Liska <mliska@suse.cz>
19143 * tree.c (record_node_allocation_statistics): Use
19144 get_stats_node_kind.
19145 (get_stats_node_kind): New function extracted from
19146 record_node_allocation_statistics.
19147 (free_node): Use get_stats_node_kind.
19149 2018-03-14 Richard Biener <rguenther@suse.de>
19151 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19152 that the value-set of ANTIC_IN doesn't grow.
19155 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19157 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19158 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19160 2018-03-14 Julia Koval <julia.koval@intel.com>
19162 * config.gcc (icelake-client, icelake-server): New.
19164 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19165 (initial_ix86_arch_features): Ditto.
19166 (PTA_SKYLAKE): Add SGX.
19167 (PTA_ICELAKE): Remove.
19168 (PTA_ICELAKE_CLIENT): New.
19169 (PTA_ICELAKE_SERVER): New.
19170 (ix86_option_override_internal): Split up icelake on icelake client and
19172 (get_builtin_code_for_version): Ditto.
19173 (fold_builtin_cpu): Ditto.
19174 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19175 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19176 * config/i386/i386.h (processor_type): Ditto.
19177 * doc/invoke.texi: Ditto.
19179 2018-03-14 Jakub Jelinek <jakub@redhat.com>
19182 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19183 INTEGER_CST offset, add it together with bitpos / 8 and
19184 sign extend based on POINTER_SIZE.
19188 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
19191 * config/i386/constraints.md (Yc): New register constraint.
19192 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19193 Use Yc constraint for alternative 2 of operand 0. Remove
19194 preferred_for_speed attribute.
19196 2018-03-14 Richard Biener <rguenther@suse.de>
19198 PR tree-optimization/84830
19199 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19200 with the old one to avoid oscillations.
19202 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
19205 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19207 (assign_by_spills): Return a flag of reload assignment failure.
19208 Do not process the reload assignment failures. Do not spill other
19209 reload pseudos if they has the same reg class. Update n if
19211 (lra_assign): Add a return arg. Set up from the result of
19212 assign_by_spills call.
19213 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19214 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19215 usage_insns if it is not NULL.
19216 (spill_hard_reg_in_range): New function.
19217 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19218 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19219 function prototypes.
19220 (lra_assign): Change prototype.
19221 * lra.c (lra): Add code to deal with fails by splitting hard reg
19224 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
19226 * config/riscv/riscv.opt (mrelax): New option.
19227 * config/riscv/riscv.c (riscv_file_start): Emit ".option
19228 "norelax" when riscv_mrelax is disabled.
19229 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19231 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19234 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19235 reassociation for int modes.
19237 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19239 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19240 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19242 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19243 * config/aarch64/aarch64-sve.md
19244 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19245 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19246 (*extend<mode><Vwide>2): Rename to...
19247 (aarch64_sve_extend<mode><Vwide>2): ...this.
19248 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19249 renaming the old pattern to...
19250 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
19252 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19253 define_expand, renaming the old pattern to...
19254 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19255 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19256 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19257 account when deciding which SVE instruction the optab should use.
19258 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19260 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19262 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19263 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19264 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19265 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19266 (tlsdesc_small_<mode>): Turn a define_expand and use
19267 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
19268 (tlsdesc_small_advsimd_<mode>): ...this.
19269 (tlsdesc_small_sve_<mode>): New pattern.
19271 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19273 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19274 (UNSPEC_UMUL_HIGHPART): New constants.
19275 (MUL_HIGHPART): New int iteraor.
19276 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19277 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19279 (*<su>mul<mode>3_highpart): New define_insn.
19281 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19284 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19287 2018-03-13 Martin Liska <mliska@suse.cz>
19290 * (sem_item_optimizer::sem_item_optimizer): Initialize new
19292 (sem_item_optimizer::~sem_item_optimizer): Release it.
19293 (sem_item_optimizer::merge_classes): Register variable aliases.
19294 (sem_item_optimizer::fixup_pt_set): New function.
19295 (sem_item_optimizer::fixup_points_to_sets): Likewise.
19296 * ipa-icf.h: Declare new variables and functions.
19298 2018-03-13 Jakub Jelinek <jakub@redhat.com>
19300 PR middle-end/84834
19301 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19302 integer_pow2p@2 and test integer_pow2p in condition.
19303 (A < 0 ? C : 0): Similarly for @1.
19305 PR middle-end/84831
19306 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19307 characters starting at p contain '\0' character, don't look beyond
19311 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19312 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19315 * reg-stack.c (change_stack): Change update_end var from int to
19316 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19317 also call set_block_for_insn on the newly added insns and rescan.
19320 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19321 on the last operand.
19324 * tree.c (stabilize_reference_1): Return save_expr (e) for
19325 STATEMENT_LIST even if it doesn't have side-effects.
19327 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
19329 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19331 2018-03-12 Renlin Li <renlin.li@arm.com>
19333 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19334 aarch64_output_scalar_simd_mov_immediate.
19336 2018-03-12 Martin Sebor <msebor@redhat.com>
19338 PR tree-optimization/83456
19339 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19340 for perfectly overlapping calls to memcpy.
19341 (gimple_fold_builtin_memory_chk): Same.
19342 (gimple_fold_builtin_strcpy): Handle no-warning.
19343 (gimple_fold_builtin_stxcpy_chk): Same.
19344 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19346 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19348 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19349 parameter. Use it for SFmode.
19350 (rs6000_function_arg_advance_1): Adjust.
19351 (rs6000_function_arg): Adjust.
19352 (rs6000_gimplify_va_arg): Pass false for that new parameter.
19354 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19356 PR rtl-optimization/84169
19357 PR rtl-optimization/84780
19358 * combine.c (can_combine_p): Check for a 2-insn combination whether
19359 the destination register is used between the two insns, too.
19361 2018-03-12 Richard Biener <rguenther@suse.de>
19363 PR tree-optimization/84803
19364 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19365 for refs DR analysis didn't process.
19367 2018-03-12 Richard Biener <rguenther@suse.de>
19369 PR tree-optimization/84777
19370 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19371 force-vectorize loops ignore whether we are optimizing for size.
19373 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
19375 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19376 (TARGET_MD_ASM_ADJUST): Define.
19378 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
19379 Kito Cheng <kito.cheng@gmail.com>
19380 Chung-Ju Wu <jasonwucj@gmail.com>
19382 * config/nds32/nds32.c (nds32_compute_stack_frame,
19383 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19384 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19385 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19386 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19387 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19388 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19389 * config/nds32/nds32.md (prologue, epilogue): Use macro
19390 NDS32_V3PUSH_AVAILABLE_P to do checking.
19392 2018-03-11 Jakub Jelinek <jakub@redhat.com>
19395 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19396 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19397 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
19398 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
19399 addition of most attributes on !orig_type_die or the attribute not
19400 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19402 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19403 Chung-Ju Wu <jasonwucj@gmail.com>
19405 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19406 __NDS32_VH__ macro.
19407 * config/nds32/nds32.opt (mvh): New option.
19409 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19410 Chung-Ju Wu <jasonwucj@gmail.com>
19412 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19414 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19415 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19418 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19419 Chung-Ju Wu <jasonwucj@gmail.com>
19421 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19423 * config/nds32/nds32-multiple.md (strlensi): New pattern.
19424 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19426 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
19427 Kito Cheng <kito.cheng@gmail.com>
19428 Chung-Ju Wu <jasonwucj@gmail.com>
19430 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19431 UNSPEC_FFMISM and UNSPEC_FLMISM.
19432 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19433 for ffb, ffmism and flmism.
19434 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19435 (unspec_ffmism): Ditto.
19436 (unspec_flmism): Ditto.
19437 (nds32_expand_builtin_impl): Check if string extension is available.
19438 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19439 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19441 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
19444 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19447 * lra-assigns.c (assign_by_spills): Return a flag of reload
19448 assignment failure. Do not process the reload assignment
19449 failures. Do not spill other reload pseudos if they has the same
19451 (lra_assign): Add a return arg. Set up from the result of
19452 assign_by_spills call.
19453 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19454 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19455 usage_insns if it is not NULL.
19456 (spill_hard_reg_in_range): New function.
19457 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19458 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19459 function prototypes.
19460 (lra_assign): Change prototype.
19461 * lra.c (lra): Add code to deal with fails by splitting hard reg
19464 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
19467 * config/i386/i386.opt: Replace Enforcment with Enforcement.
19469 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
19472 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19473 (dw_val_node): Add val_symbolic_view.
19474 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19475 (symview_upper_bound): New.
19476 (new_line_info_table): Initialize symviews_since_reset.
19477 (dwarf2out_source_line): Count symviews_since_reset and set
19478 symview_upper_bound.
19479 (dw_val_equal_p): Handle symview.
19480 (add_AT_symview): New.
19481 (print_dw_val): Handle symview.
19482 (attr_checksum, attr_checksum_ordered): Likewise.
19483 (same_dw_val_p, size_of_die): Likewise.
19484 (value_format, output_die): Likewise.
19485 (add_high_low_attributes): Use add_AT_symview for entry_view.
19486 (dwarf2out_finish): Reset symview_upper_bound, clear
19489 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
19492 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19493 Add strict argument and use it.
19494 (rs6000_split_multireg_move): Update for new strict argument.
19495 (mem_operand_gpr): Disallow all non-offsettable addresses.
19496 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19498 2018-03-09 Jakub Jelinek <jakub@redhat.com>
19501 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19502 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19503 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19506 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19507 decl, use remap_type if we want to use the type.
19509 2018-03-09 Martin Sebor <msebor@redhat.com>
19511 PR tree-optimization/84526
19512 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19514 (builtin_access::generic_overlap): Be prepared to handle non-array
19517 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
19519 PR rtl-optimization/84682
19520 * lra-constraints.c (process_address_1): Check is_address flag
19521 for address constraints.
19522 (process_alt_operands): Likewise.
19523 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19524 preprocess_constraints.
19525 * recog.h (preprocess_constraints): Add oploc parameter.
19527 * recog.c (preprocess_constraints): Test address_operand for
19528 CT_ADDRESS constraints.
19530 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19533 * lra-assigns.c (assign_by_spills): Return a flag of reload
19534 assignment failure. Do not process the reload assignment
19535 failures. Do not spill other reload pseudos if they has the same
19537 (lra_assign): Add a return arg. Set up from the result of
19538 assign_by_spills call.
19539 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19540 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19541 usage_insns if it is not NULL.
19542 (spill_hard_reg_in_range): New function.
19543 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19544 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19545 function prototypes.
19546 (lra_assign): Change prototype.
19547 * lra.c (lra): Add code to deal with fails by splitting hard reg
19550 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19553 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19554 Accept complain bool parameter. Only emit errors if it is true.
19555 (arm_parse_cpu_option_name): Likewise.
19556 (arm_target_thumb_only): Adjust callers of the above.
19557 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19558 prototype to take a default true bool parameter.
19559 (arm_parse_arch_option_name): Likewise.
19561 2018-03-09 David Malcolm <dmalcolm@redhat.com>
19562 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
19566 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19567 * configure: Regenerate.
19568 * configure.ac ("linker --version-script option"): New.
19569 ("linker soname option"): New.
19571 2018-03-09 Richard Biener <rguenther@suse.de>
19573 PR tree-optimization/84775
19574 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19575 immediate uses of predicate stmts and mark them modified.
19578 PR tree-optimization/84178
19579 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19581 (version_loop_for_if_conversion): Delay update_ssa call.
19582 (tree_if_conversion): Delay update_ssa until after predicate
19585 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
19588 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19589 when the function accesses prior frames.
19591 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19594 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19595 gen_llsym, otherwise call maybe_gen_llsym.
19597 PR inline-asm/84742
19598 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19599 has ',' character inside of it.
19601 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19604 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19605 as clobbering CC_REGNUM.
19607 2018-03-08 Richard Biener <rguenther@suse.de>
19609 PR middle-end/84552
19610 * tree-scalar-evolution.c: Include tree-into-ssa.h.
19611 (follow_copies_to_constant): Do not follow SSA names registered
19614 2018-03-08 Richard Biener <rguenther@suse.de>
19616 PR tree-optimization/84178
19617 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19619 (version_loop_for_if_conversion): Delay update_ssa call.
19620 (tree_if_conversion): Delay update_ssa until after predicate
19623 2018-03-08 David Malcolm <dmalcolm@redhat.com>
19625 PR tree-optimization/84178
19626 * tree-if-conv.c (release_bb_predicate): Remove the
19627 the assertion that the stmts have NULL use_ops.
19628 Discard the statements, asserting that they haven't
19629 yet been added to a BB.
19631 2018-03-08 Richard Biener <rguenther@suse.de>
19633 PR tree-optimization/84746
19634 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19635 (phi_translate): Pass in destination ANTIC_OUT set.
19636 (phi_translate_1): Likewise. For a simplified result lookup
19637 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19638 (phi_translate_set): Adjust.
19639 (do_pre_regular_insertion): Likewise.
19640 (do_pre_partial_partial_insertion): Likewise.
19642 2018-03-08 Martin Liska <mliska@suse.cz>
19644 PR gcov-profile/84735
19645 * doc/gcov.texi: Document usage of profile files.
19646 * gcov-io.h: Document changes in the format.
19648 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
19652 * dwarf2out.c (struct dw_line_info_table): Update comments for
19654 (FORCE_RESET_NEXT_VIEW): New.
19655 (FORCE_RESETTING_VIEW_P): New.
19656 (RESETTING_VIEW_P): Check for -1 too.
19657 (ZERO_VIEW_P): Likewise.
19658 (new_line_info_table): Force-reset next view.
19659 (dwarf2out_begin_function): Likewise.
19660 (dwarf2out_source_line): Simplify zero_view_p initialization.
19661 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19662 view directly. Omit view when omitting .loc at line 0.
19664 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19666 PR tree-optimization/84740
19667 * tree-switch-conversion.c (process_switch): Call build_constructors
19668 only if info.phi_count is non-zero.
19670 PR tree-optimization/84739
19671 * tree-tailcall.c (find_tail_calls): Check call arguments against
19672 DECL_ARGUMENTS (current_function_decl) rather than
19673 DECL_ARGUMENTS (func) when checking for tail recursion.
19675 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19677 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19678 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19679 Volker Reichelt's entry and add entries for people that perform
19680 GCC fuzzy testing and report numerous bugs.
19682 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
19685 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19686 readonly data in sdata, if that is disabled.
19687 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19688 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19689 -mreadonly-in-sdata option.
19691 2018-03-07 Martin Sebor <msebor@redhat.com>
19693 PR tree-optimization/84468
19694 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19695 basic block when looking for nul assignment.
19697 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
19700 * except.h (output_function_exception_table): Adjust prototype.
19701 * except.c (output_function_exception_table): Remove FNNAME parameter
19702 and add SECTION parameter. Ouput one part of the table at a time.
19703 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19704 the first part of the exception table and emit unwind directives.
19705 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19706 (i386_pe_seh_cold_init): Likewise.
19707 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19708 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19709 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19710 (ix86_output_call_insn): Emit a nop in one more case for SEH.
19711 * config/i386/winnt.c: Include except.h.
19712 (struct seh_frame_state): Add reg_offset, after_prologue and
19713 in_cold_section fields.
19714 (i386_pe_seh_end_prologue): Set seh->after_prologue.
19715 (i386_pe_seh_cold_init): New function.
19716 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19717 to seh->in_cold_section.
19718 (seh_emit_push): Record the offset of the push.
19719 (seh_emit_save): Record the offet of the save.
19720 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19721 Test seh->after_prologue to disregard the epilogue.
19722 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19723 (i386_pe_end_cold_function): New function.
19725 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19728 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19729 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19732 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19734 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19735 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19737 PR middle-end/84723
19738 * multiple_target.c: Include tree-inline.h and intl.h.
19739 (expand_target_clones): Diagnose and fail if node->definition and
19740 !tree_versionable_function_p (node->decl).
19742 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
19744 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19746 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19747 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19748 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19750 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19753 * combine.c (try_combine): Use reg_or_subregno instead of handling
19754 just paradoxical SUBREGs and REGs.
19756 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
19758 * config/arc/arc.c (arc_finalize_pic): Remove function.
19759 (arc_must_save_register): We use single base PIC register, remove
19760 checks to save/restore the PIC register.
19761 (arc_expand_prologue): Likewise.
19762 * config/arc/arc-protos.h (arc_set_default_type_attributes):
19764 (arc_verify_short): Likewise.
19765 (arc_attr_type): Likewise.
19766 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19767 (walk_stores): Likewise.
19768 (arc_address_cost): Make it static.
19769 (arc_verify_short): Likewise.
19770 (branch_dest): Likewise.
19771 (arc_attr_type): Likewise.
19772 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19773 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19774 (arc_final_prescan_insn): Remove inserting the nops due to
19775 hardware hazards. It is done in reorg step.
19776 (insn_length_variant_t): Remove.
19777 (insn_length_parameters_t): Likewise.
19778 (arc_insn_length_parameters): Likewise.
19779 (arc_get_insn_variants): Likewise.
19780 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19782 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19784 PR inline-asm/84683
19785 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19788 PR tree-optimization/84687
19789 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19791 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19793 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19795 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19796 Rename to ppc_speculation_barrier.
19797 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19798 __builtin_ppc_speculation_barrier.
19800 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19803 * combine.c (combine_simplify_rtx): Don't try to simplify if
19804 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19807 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
19809 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19810 to 32 bytes when compiling for POWER9.
19812 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19815 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19816 regparm >= 3 with no arg reg available also for calls with
19817 flag_force_indirect_call. Pass decl to ix86_function_regparm.
19820 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19822 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
19824 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
19827 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19829 2018-03-05 Richard Biener <rguenther@suse.de>
19831 PR tree-optimization/84486
19832 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19833 When inserting a __builtin_assume_aligned call set the LHS
19834 SSA name alignment info accordingly.
19836 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
19838 PR tree-optimization/84114
19839 * config/aarch64/aarch64.c (aarch64_reassociation_width)
19840 Avoid reassociation of FLOAT_MODE addition.
19842 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
19844 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19845 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19846 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19847 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19848 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19849 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19850 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19852 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19853 __builtin_ia32_wbinvd): New builtins.
19854 (SPECIAL_ARGS2): New.
19855 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19856 (SPECIAL_ARGS2): New.
19857 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19858 (ix86_valid_target_attribute_inner_p): Ditto.
19859 (ix86_init_mmx_sse_builtins): Add special_args2.
19860 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19861 TARGET_WBNOINVD_P): New.
19862 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19863 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19864 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19865 * config/i386/immintrin.h (_wbinvd): New intrinsic.
19866 * config/i386/pconfigintrin.h: New file.
19867 * config/i386/wbnoinvdintrin.h: Ditto.
19868 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19870 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19872 2018-03-05 Richard Biener <rguenther@suse.de>
19874 PR tree-optimization/84670
19875 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19877 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19878 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19879 (compute_antic_aux): Only assert the number of values in ANTIC_IN
19880 doesn't grow if all successors (recursively) were visited at least
19883 2018-03-05 Richard Biener <rguenther@suse.de>
19885 PR tree-optimization/84650
19886 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19887 if executed in the loop pipeline.
19889 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
19891 * doc/configfiles.texi (Configuration Files): Move info about
19892 conditionalizing $target-protos.h to...
19893 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
19894 differs from $target-protos.h.
19896 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
19897 Chung-Ju Wu <jasonwucj@gmail.com>
19899 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19900 * config/nds32/nds32-multiple.md (setmemsi): Define.
19901 * config/nds32/nds32-memory-manipulation.c
19902 (nds32_gen_dup_4_byte_to_word_value): New.
19903 (emit_setmem_word_loop): New.
19904 (emit_setmem_byte_loop): New.
19905 (nds32_expand_setmem_loop): New.
19906 (nds32_expand_setmem_loop_v3m): New.
19907 (nds32_expand_setmem_unroll): New.
19908 (nds32_expand_setmem): New.
19910 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19911 Chung-Ju Wu <jasonwucj@gmail.com>
19913 * config/nds32/nds32-memory-manipulation.c
19914 (nds32_emit_load_store): New.
19915 (nds32_emit_post_inc_load_store): New.
19916 (nds32_emit_mem_move): New.
19917 (nds32_emit_mem_move_block): New.
19918 (nds32_expand_movmemsi_loop_unknown_size): New.
19919 (nds32_expand_movmemsi_loop_known_size): New.
19920 (nds32_expand_movmemsi_loop): New.
19921 (nds32_expand_movmemsi_unroll): New.
19922 (nds32_expand_movmemqi): Rename ...
19923 (nds32_expand_movmemsi): ... to this.
19924 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19925 (movmemsi): ... to this.
19926 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19927 (nds32_expand_movmemsi): ... to this.
19929 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19930 Monk Chiang <sh.chiang04@gmail.com>
19931 Chung-Ju Wu <jasonwucj@gmail.com>
19933 * config/nds32/nds32-protos.h
19934 (nds32_expand_load_multiple): New arguments.
19935 (nds32_expand_store_multiple): Ditto.
19936 (nds32_valid_multiple_load_store): Rename ...
19937 (nds32_valid_multiple_load_store_p): ... to this.
19938 * config/nds32/nds32-memory-manipulation.c
19939 (nds32_expand_load_multiple): Refine implementation.
19940 (nds32_expand_store_multiple): Ditto.
19941 * config/nds32/nds32-multiple.md
19942 (load_multiple): Update nds32_expand_load_multiple interface.
19943 (store_multiple): Update nds32_expand_store_multiple interface.
19944 * config/nds32/nds32-predicates.c
19945 (nds32_valid_multiple_load_store): Rename ...
19946 (nds32_valid_multiple_load_store_p): ... to this and refine
19948 * config/nds32/predicates.md
19949 (nds32_load_multiple_and_update_address_operation): New predicate.
19950 (nds32_store_multiple_and_update_address_operation): New predicate.
19952 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19953 Chung-Ju Wu <jasonwucj@gmail.com>
19955 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19956 (combo): New attribute.
19957 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19959 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
19961 * config/nds32/nds32.opt: Change -mcmodel= default value.
19963 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
19964 Monk Chiang <sh.chiang04@gmail.com>
19965 Chung-Ju Wu <jasonwucj@gmail.com>
19967 * config/nds32/constants.md (unspec_element): New enum.
19968 * config/nds32/constraints.md (Umw): New constraint.
19969 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19970 * config/nds32/nds32-intrinsic.md: Likewise.
19971 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19972 (nds32_valid_smw_lwm_base_p): New.
19973 (nds32_output_smw_single_word): New.
19974 (nds32_output_lmw_single_word): New.
19975 (nds32_expand_unaligned_load): New.
19976 (nds32_expand_unaligned_store): New.
19977 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19978 (nds32_output_smw_single_word): Declare.
19979 (nds32_output_lmw_single_word): Declare.
19980 (nds32_expand_unaligned_load): Declare.
19981 (nds32_expand_unaligned_store): Declare.
19982 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19983 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19984 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19985 NDS32_BUILTIN_UASTORE_DW.
19986 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19989 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
19990 Kito Cheng <kito.cheng@gmail.com>
19991 Chung-Ju Wu <jasonwucj@gmail.com>
19993 * config/nds32/nds32-intrinsic.c
19994 (nds32_expand_builtin_null_ftype_reg): Delete.
19995 (nds32_expand_builtin_reg_ftype_imm): Ditto.
19996 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19997 (nds32_read_argument): New.
19998 (nds32_legitimize_target): Ditto.
19999 (nds32_legitimize_argument): Ditto.
20000 (nds32_check_constant_argument): Ditto.
20001 (nds32_expand_unop_builtin): Ditto.
20002 (nds32_expand_unopimm_builtin): Ditto.
20003 (nds32_expand_binop_builtin): Ditto.
20004 (nds32_builtin_decl_impl): Ditto.
20005 (builtin_description): Ditto.
20006 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20007 (nds32_init_builtins_impl): Ditto.
20008 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20009 (nds32_builtin_decl): New.
20010 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20011 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20013 2018-03-02 Jeff Law <law@redhat.com>
20015 * reorg.c (stop_search_p): Handle DEBUG_INSN.
20016 (redundant_insn, fill_simple_delay_slots): Likewise.
20017 (fill_slots_from_thread): Likewise.
20018 * resource.c (mark_referenced_resources): Likewise.
20019 (mark_set_resources, find_dead_or_set_registers): Likewise.
20021 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20023 * substring-locations.h (format_warning_va): Formatting fix for
20024 ATTRIBUTE_GCC_DIAG.
20025 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20027 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20028 * substring-locations.c: Include intl.h.
20029 (format_warning_va): Turned into small wrapper around
20030 format_warning_n_va, renamed to ...
20031 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20032 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20034 (format_warning_at_substring_n): New function.
20035 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20036 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
20037 format_warning_at_substring with just a shorter name instead of
20038 const function pointer.
20039 (fmtwarn_n): New function.
20040 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20041 appropriate, get rid of all the fmtstr temporaries, move conditionals
20042 with G_() wrapped string literals directly into fmtwarn arguments,
20043 cast dir.len to (int), formatting fixes.
20045 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
20047 * doc/invoke.texi: Remove "Cilk Plus" references.
20049 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20050 Richard Biener <rguenther@suse.de>
20053 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20054 for error or warning attributes if CALL_FROM_THUNK_P is set.
20057 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20060 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20061 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20064 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20065 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20067 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20068 instead of -1U in last predictors element's probability member.
20070 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20073 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20074 arguments if they are comparable.
20076 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20078 PR tree-optimization/84634
20079 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20080 masks and masked_loop_p with a single loop_masks, making sure it's
20081 null for bb vectorization.
20083 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20085 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20086 (vect_analyze_data_ref_access): Use loop->safe_len rather than
20087 loop->force_vectorize to check whether there is no alias.
20089 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20092 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20094 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20096 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20097 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20098 instead of a loop around prev_real_insn.
20099 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20102 PR inline-asm/84625
20103 * config/i386/i386.c (ix86_print_operand): Use conditional
20104 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20107 2018-03-02 Richard Biener <rguenther@suse.de>
20109 PR tree-optimization/84427
20110 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20111 (bitmap_set_subtract_values): Rewrite to handle multiple
20114 (prune_clobbered_mems): Likewise.
20115 (phi_translate): Take edge instead of pred/phiblock.
20116 (phi_translate_1): Likewise.
20117 (phi_translate_set): Likewise. Insert all translated
20118 exprs for a value into the set, keeping possibly multiple
20119 expressions per value.
20120 (compute_antic_aux): Adjust for phi_translate changes.
20121 When intersecting union the expressions and prune those
20122 not in the final value set, keeping possibly multiple
20123 expressions per value. Do not use value-insertion
20124 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20125 all expressions. Add verification that the value-sets
20126 only shrink during iteration.
20127 (compute_partial_antic_aux): Adjust for the phi_translate changes.
20128 (do_pre_regular_insertion): Likewise.
20129 (do_pre_partial_partial_insertion): Likewise.
20131 2018-03-02 Richard Biener <rguenther@suse.de>
20134 * config/darwin.c (saved_debug_info_level): New static global.
20135 (darwin_asm_lto_start): Disable debug info generation for LTO out.
20136 (darwin_asm_lto_end): Restore debug info generation settings.
20138 2018-03-01 Martin Liska <mliska@suse.cz>
20141 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20142 volatile arguments.
20144 2018-03-01 Richard Biener <rguenther@suse.de>
20147 * dwarf2out.c (gen_variable_die): Properly handle late VLA
20148 type annotation with LTO when debug was disabled at compile-time.
20150 2018-03-01 Matthew Fortune <mfortune@gmail.com>
20152 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20154 (mips_final_postscan_insn): Likewise.
20156 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
20158 PR rtl-optimization/84528
20159 * alias.c (init_alias_target): Add commentary.
20160 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20161 a unique base value if the frame pointer is not eliminated
20162 to the stack pointer.
20164 2018-03-01 Tom de Vries <tom@codesourcery.com>
20166 PR rtl-optimization/83327
20167 * lra-int.h (hard_regs_spilled_into): Declare.
20168 * lra.c (hard_regs_spilled_into): Define.
20169 (init_reg_info): Init hard_regs_spilled_into.
20170 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20171 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20172 (process_bb_lives): Handle hard_regs_spilled_into.
20173 (lra_create_live_ranges_1): Before doing liveness propagation, clear
20174 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20176 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
20178 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20179 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20180 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20181 * config/rs6000/aix72.h: New file.
20183 2018-02-28 Jakub Jelinek <jakub@redhat.com>
20185 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20186 instead of warning_at with conditional singular and plural messages
20190 * stor-layout.c (update_alignment_for_field): For
20191 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20192 && !DECL_PACKED (field), do the alignment update, just use
20193 only desired_align instead of MAX (type_align, desired_align)
20195 (place_field): Don't do known_align < desired_align handling
20196 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20197 is non-NULL, instead do it after rli->prev_field handling and
20198 only if not within a bitfield word. For DECL_PACKED (field)
20199 use type_align of BITS_PER_UNIT.
20201 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
20203 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20204 superfluous parentheses and trailing spaces.
20206 2018-02-28 Richard Biener <rguenther@suse.de>
20208 PR tree-optimization/84584
20209 * graphite-scop-detection.c (scop_detection::add_scop): Discard
20210 SCoPs with fake exit edge.
20212 2018-02-28 Martin Liska <mliska@suse.cz>
20215 * timevar.c (timer::print): Fix format to properly print 100%
20218 2018-02-28 Richard Biener <rguenther@suse.de>
20220 PR middle-end/84607
20221 * genmatch.c (capture_info::walk_match): Do not mark
20222 captured expressions without operands as expr_p given
20223 they act more like predicates and should be subject to
20224 "lost tail" side-effect preserving.
20226 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
20228 PR rtl-optimization/81611
20229 * auto-inc-dec.c (attempt_change): Move dead note from
20230 mem_insn if it's the next use of regno
20231 (find_address): Take address use of reg holding
20232 non-incremented value. Add parm to limit search to the named
20234 (merge_in_block): Attempt to use a mem insn that is the next
20235 use of the original regno.
20237 2018-02-27 Martin Sebor <msebor@redhat.com>
20240 * doc/invoke.texi (-Wmissing-attributes): New option.
20241 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20243 2018-02-27 Martin Sebor <msebor@redhat.com>
20245 PR translation/84207
20246 * diagnostic-core.h (warning_n, error_n, inform_n): Change
20247 n argument to unsigned HOST_WIDE_INT.
20248 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20249 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
20250 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20251 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20253 2018-02-27 Richard Biener <rguenther@suse.de>
20255 PR tree-optimization/84512
20256 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20257 Do not use the estimate returned from record_stmt_cost for
20258 the scalar iteration cost but sum properly using add_stmt_cost.
20260 2018-02-27 Richard Biener <rguenther@suse.de>
20262 PR tree-optimization/84466
20263 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20264 Adjust last change to less strictly validate use operands.
20266 2018-02-27 Martin Liska <mliska@suse.cz>
20268 PR gcov-profile/84548
20269 * gcov.c (process_file): Allow partial overlap and consider it
20270 also as group functions.
20271 (output_lines): Properly calculate range of lines for a group.
20273 2018-02-27 Martin Liska <mliska@suse.cz>
20275 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20276 'ggc' suffixes. Change first column width.
20277 (timer::print): Fix formatting of the column.
20279 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
20281 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20282 preserve inline entry blocks for the sake of debug inline
20283 entry point markers alone.
20284 (remove_unused_locals): Suggest in comments a better place to
20285 force the preservation of inline entry blocks that are
20286 otherwise unused, but do not preserve them.
20288 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20290 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20292 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20295 * config/i386/constraints.md (Bs): Replace
20296 ix86_indirect_branch_register with
20297 TARGET_INDIRECT_BRANCH_REGISTER.
20299 * config/i386/i386.md (indirect_jump): Likewise.
20300 (tablejump): Likewise.
20301 (*sibcall_memory): Likewise.
20302 (*sibcall_value_memory): Likewise.
20303 Peepholes of indirect call and jump via memory: Likewise.
20304 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20305 (*sibcall_value_GOT_32): Likewise.
20306 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20307 (GOT_memory_operand): Likewise.
20308 (call_insn_operand): Likewise.
20309 (sibcall_insn_operand): Likewise.
20310 (GOT32_symbol_operand): Likewise.
20311 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20313 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20315 PR rtl-optimization/83496
20316 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20317 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
20318 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20319 redundant insn, if any.
20320 (relax_delay_slots): Likewise.
20321 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20323 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
20325 PR tree-optimization/83965
20326 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20327 that grouped statements are part of a reduction chain. Return
20328 true if the statement is not marked as a reduction itself but
20329 is part of a group.
20330 (vect_recog_dot_prod_pattern): Don't check whether the statement
20331 is part of a group here.
20332 (vect_recog_sad_pattern): Likewise.
20333 (vect_recog_widen_sum_pattern): Likewise.
20335 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20338 * final.c (rest_of_clean_state): Also look for calls inside sequences.
20340 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20343 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20345 (ix86_output_indirect_function_return): New prototype.
20346 (ix86_split_simple_return_pop_internal): Likewise.
20347 * config/i386/i386.c (indirect_return_via_cx): New.
20348 (indirect_return_via_cx_bnd): Likewise.
20349 (indirect_thunk_name): Handle return va CX_REG.
20350 (output_indirect_thunk_function): Create alias for
20351 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20352 (ix86_output_indirect_jmp): Remove the bool argument.
20353 (ix86_output_indirect_function_return): New function.
20354 (ix86_split_simple_return_pop_internal): Likewise.
20355 * config/i386/i386.md (*indirect_jump): Don't pass false
20356 to ix86_output_indirect_jmp.
20357 (*tablejump_1): Likewise.
20358 (simple_return_pop_internal): Change it to define_insn_and_split.
20359 Call ix86_split_simple_return_pop_internal to split it for
20360 -mfunction-return=.
20361 (simple_return_indirect_internal): Call
20362 ix86_output_indirect_function_return instead of
20363 ix86_output_indirect_jmp.
20365 2018-02-26 Jakub Jelinek <jakub@redhat.com>
20368 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20369 memset and value initialization afterwards.
20371 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
20373 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20375 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20378 * common/config/aarch64/aarch64-common.c
20379 (aarch_option_optimization_table[]): Switch
20380 off fomit-frame-pointer
20382 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20383 Chung-Ju Wu <jasonwucj@gmail.com>
20385 * config/nds32/nds32-multiple.md (load_multiple): Disallow
20387 (store_multiple): Ditto.
20389 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20391 * config.gcc: Add --with-cpu support for nds32 target.
20392 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20393 * config/nds32/nds32.opt: Add -mcpu= option.
20395 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
20397 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20398 isel=yes): Warn for these deprecated options.
20400 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
20402 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20403 ISA_2_5_MASKS_EMBEDDED.
20405 2018-02-23 Jakub Jelinek <jakub@redhat.com>
20407 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20408 p->max as pointers rather than using iterative_hash_expr.
20410 2018-02-23 Carl Love <cel@us.ibm.com>
20412 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20413 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20415 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20416 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20417 P8V_BUILTIN_VEC_VUNSIGNED2.
20419 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
20422 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20423 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20425 * lra-constraints.c (curr_insn_transform): Set up
20426 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
20428 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20429 * lra-eliminations.c (spill_pseudos): Ditto.
20430 (process_insn_for_elimination): Ditto.
20431 * lra-lives.c (reg_early_clobber_p): Use the new macros.
20432 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20433 LRA_NON_CLOBBERED_ALT.
20435 2018-02-22 Martin Sebor <msebor@redhat.com>
20437 PR tree-optimization/84480
20438 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20439 to maybe_diag_stxncpy_trunc. Call it.
20440 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20441 from gimple_fold_builtin_strcpy. Print inlining stack.
20442 (handle_builtin_stxncpy): Print inlining stack.
20443 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20445 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
20448 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20449 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20450 and -fcheck-pointer-bounds are used together.
20451 (indirect_thunk_prefix): New enum.
20452 (indirect_thunk_need_prefix): New function.
20453 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
20454 "_nt" instead of "_bnd" for NOTRACK prefix.
20455 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20456 (output_indirect_thunk_function): Likewise.
20458 (ix86_code_end): Update output_indirect_thunk_function calls.
20459 (ix86_output_indirect_branch_via_reg): Replace
20460 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20461 (ix86_output_indirect_branch_via_push): Likewise.
20462 (ix86_output_function_return): Likewise.
20463 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20464 incompatible with -fcf-protection=branch and
20465 -fcheck-pointer-bounds.
20467 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20470 * config/aarch64/aarch64.c (aarch64_print_address_internal):
20471 Change gcc_assert call to output_operand_lossage.
20473 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20475 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20477 2018-02-22 DJ Delorie <dj@redhat.com>
20478 Sebastian Perta <sebastian.perta@renesas.com>
20479 Oleg Endo <olegendo@gcc.gnu.org>
20481 * config/rx/rx.c (rx_rtx_costs): New function.
20482 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20484 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
20486 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20488 2018-02-22 Martin Liska <mliska@suse.cz>
20491 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20492 Add "native" as a possible value.
20494 2018-02-22 Martin Liska <mliska@suse.cz>
20497 * config/i386/i386.c (ix86_option_override_internal):
20498 Add "native" as a possible value for -march and -mtune.
20500 2018-02-22 Jakub Jelinek <jakub@redhat.com>
20503 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20504 to all type variants.
20506 PR tree-optimization/84503
20507 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20508 width as info->bitpos + info->bitsize - start.
20509 (merged_store_group::merge_overlapping): Simplify width computation.
20510 (check_no_overlap): New function.
20511 (imm_store_chain_info::try_coalesce_bswap): Compute expected
20512 start + width and last_order of the group, fail if check_no_overlap
20514 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20515 to group if check_no_overlap fails.
20517 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20519 * config/rs6000/altivec.md: Delete contraint arguments to
20520 define_expand, define_split, and define_peephole2, and in
20521 define_insn_and_split if always unused.
20522 * config/rs6000/darwin.md: Ditto.
20523 * config/rs6000/dfp.md: Ditto.
20524 * config/rs6000/rs6000.md: Ditto.
20525 * config/rs6000/sync.md: Ditto.
20526 * config/rs6000/vector.md: Ditto.
20527 * config/rs6000/vsx.md: Ditto.
20529 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20531 * config/rs6000/altivec.md: Write output control strings as braced
20532 blocks instead of double-quoted strings.
20533 * config/rs6000/darwin.md: Ditto.
20534 * config/rs6000/rs6000.md: Ditto.
20535 * config/rs6000/vector.md: Ditto.
20536 * config/rs6000/vsx.md: Ditto.
20538 2018-02-21 Jason Merrill <jason@redhat.com>
20540 PR c++/84314 - ICE with templates and fastcall attribute.
20541 * attribs.c (build_type_attribute_qual_variant): Remove assert.
20543 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20545 * ipa-cp.c (determine_versionability): Fix comment typos.
20547 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20550 * ipa-cp.c (determine_versionability): Do not version functions caling
20553 2018-02-21 Martin Liska <mliska@suse.cz>
20556 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20557 Add "native" as a possible value.
20558 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
20559 the macro when native cpu detection is available.
20561 2018-02-21 Martin Liska <mliska@suse.cz>
20564 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20565 Add "native" as a possible value.
20566 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20567 when native cpu detection is available.
20569 2018-02-21 Jakub Jelinek <jakub@redhat.com>
20570 Martin Sebor <msebor@redhat.com>
20572 PR tree-optimization/84478
20573 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20575 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20576 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20577 support which is conservatively correct, for 2 only stay conservative
20578 for maxlen. Formatting and comment capitalization fixes. Add STRICT
20579 argument to the 2 argument get_range_strlen, adjust 6 arg
20580 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20582 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20583 (gimple_fold_builtin_strlen): Pass true as last argument to
20586 2018-02-20 Martin Sebor <msebor@redhat.com>
20588 PR middle-end/84095
20589 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20590 (builtin_memref::set_base_and_offset): Same. Handle inner references.
20591 (builtin_memref::builtin_memref): Factor out parts into
20592 set_base_and_offset and call it.
20594 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20596 PR middle-end/84406
20597 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20598 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20599 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
20600 search at the associated MODE_INT.
20602 2018-02-20 Jeff Law <law@redhat.com>
20604 PR middle-end/82123
20605 PR tree-optimization/81592
20606 PR middle-end/79257
20607 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20608 for range data rather than using global data.
20609 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20610 range data rather than using global data.
20611 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20612 pass it to children as needed.
20613 (struct directive::fmtresult): Similarly.
20614 (struct directive::set_width): Similarly.
20615 (struct directive::set_precision): Similarly.
20616 (format_integer, format_directive, parse_directive): Similarly.
20617 (format_none): Accept unnamed vr_values parameter.
20618 (format_percent, format_floating, format_character): Similarly.
20619 (format_string, format_plain): Similarly.
20620 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20621 the EVRP range analyzer for range data rather than using global data.
20622 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20623 gimple-ssa-evrp-analyze.h
20624 (class sprintf_dom_walker): Add after_dom_children member function.
20625 Add evrp_range_analyzer member.
20626 (sprintf_dom_walker::before_dom_children): Call into the EVRP
20627 range analyzer as needed.
20628 (sprintf_dom_walker::after_dom_children): New member function.
20629 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20631 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20632 (evrp_range_analyzer::pop_to_marker): Likewise.
20634 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20636 PR tree-optimization/84419
20637 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20638 with the required type if its current type is compatible but
20641 2018-02-20 Jakub Jelinek <jakub@redhat.com>
20643 PR middle-end/82004
20644 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20645 after vectorization.
20647 2018-02-20 Martin Liska <mliska@suse.cz>
20650 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20651 possible values if we don't have a hint.
20653 2018-02-20 Martin Liska <mliska@suse.cz>
20657 * final.c (shorten_branches): Build align_tab array with one
20659 * opts.c (finish_options): Add alignment option limit check.
20660 (MAX_CODE_ALIGN): Likewise.
20661 (MAX_CODE_ALIGN_VALUE): Likewise.
20662 * doc/invoke.texi: Document maximum allowed option value for
20663 all -falign-* options.
20665 2018-02-19 Jakub Jelinek <jakub@redhat.com>
20668 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20669 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20670 * var-tracking.c (emit_note_insn_var_location): Remove all references
20671 to NOTE_INSN_CALL_ARG_LOCATION.
20672 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20673 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20674 Use copy_rtx_if_shared.
20675 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20676 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20677 (dwarf2out_var_location): Remove handling of
20678 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20680 * final.c (final_scan_insn): Remove all references to
20681 NOTE_INSN_CALL_ARG_LOCATION.
20682 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
20683 before dumping final insns.
20684 * except.c (emit_note_eh_region_end): Remove all references to
20685 NOTE_INSN_CALL_ARG_LOCATION.
20686 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20687 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20688 * config/arc/arc.c (hwloop_optimize): Likewise.
20689 * config/arm/arm.c (create_fix_barrier): Likewise.
20690 * config/s390/s390.c (s390_chunkify_start): Likewise.
20691 * config/sh/sh.c (find_barrier): Likewise.
20692 * config/i386/i386.c (rest_of_insert_endbranch,
20693 ix86_seh_fixup_eh_fallthru): Likewise.
20694 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20695 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20696 * config/frv/frv.c (frv_function_prologue): Likewise.
20697 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
20699 (note_outside_basic_block_p): Remove all references to
20700 NOTE_INSN_CALL_ARG_LOCATION.
20701 * gengtype.c (adjust_field_rtx_def): Likewise.
20702 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20704 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20705 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20708 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20711 PR tree-optimization/84452
20712 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20713 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20716 2018-02-19 Martin Liska <mliska@suse.cz>
20719 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20721 2018-02-19 Martin Liska <mliska@suse.cz>
20722 Richard Sandiford <richard.sandiford@linaro.org>
20724 PR tree-optimization/82491
20725 * gimple-fold.c (get_base_constructor): Make earlier bail out
20728 2018-02-19 Carl Love <cel@us.ibm.com>
20730 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20731 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20733 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20734 P8V_BUILTIN_VEC_NEG.
20736 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
20738 * config/rl78/rl78.md (movdf): New define expand.
20740 2018-02-19 Martin Liska <mliska@suse.cz>
20743 * doc/invoke.texi: Fix typo.
20744 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20746 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
20748 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20749 handle rs6000_single_float and rs6000_double_float specially for
20752 2018-02-16 Jeff Law <law@redhat.com>
20754 * config/rx/rx.c (add_pop_cfi_notes): New function.;
20755 (pop_regs): Use it.
20757 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20760 * ipa-inline.c (inline_small_functions): Fix a typo.
20762 2018-02-16 Nathan Sidwell <nathan@acm.org>
20764 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20766 2018-02-16 Carl Love <cel@us.ibm.com>
20768 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20769 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20770 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20771 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20772 expansion to P8V_BUILTIN_VEC_FLOAT2.
20774 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20776 PR rtl-optimization/70023
20777 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20778 src_regno into account.
20780 2018-02-16 Carl Love <cel@us.ibm.com>
20782 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20783 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20784 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20785 * config/rs6000/rs6000.c: Remove case statements for
20786 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20787 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20788 and P9V_BUILTIN_VEC_VINSERT4B.
20789 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20790 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20791 * config/rs6000/vsx.md:
20792 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20795 2018-02-16 Carl Love <cel@us.ibm.com>
20797 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20799 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20801 * config/rs6000/rs6000-c.c: Add the definitions for
20802 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20803 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20804 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20805 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
20806 definition for insert4b and define insn *insert3b_internal.
20807 * doc/extend.texi: Add documentation for vec_extract4b.
20809 2018-02-16 Nathan Sidwell <nathan@acm.org>
20811 * doc/extend.texi (Backwards Compatibility): Mention friend
20812 injection. Note for-scope is deprecated.
20813 * doc/invoke.texi (-ffriend-injection): Deprecate.
20815 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
20817 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20818 that moved to I2, also allow destinations that are a paradoxical
20819 subreg (instead of a normal reg).
20821 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
20824 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20827 2018-02-16 Richard Biener <rguenther@suse.de>
20829 PR tree-optimization/84037
20830 PR tree-optimization/84016
20832 * config/i386/i386.c (ix86_builtin_vectorization_cost):
20833 Adjust vec_construct for the fact we need additional higher latency
20834 128bit inserts for AVX256 and AVX512 vector builds.
20835 (ix86_add_stmt_cost): Scale vector construction cost for
20838 2018-02-16 Richard Biener <rguenther@suse.de>
20840 PR tree-optimization/84417
20841 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20842 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20843 (non_rewritable_lvalue_p): Likewise, use poly-ints.
20845 2018-02-16 Martin Liska <mliska@suse.cz>
20848 * internal-fn.def (ASAN_CHECK): Set proper flags.
20849 (ASAN_MARK): Likewise.
20851 2018-02-16 Julia Koval <julia.koval@intel.com>
20853 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20854 from PTA_CANNONLAKE.
20856 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20859 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20860 Use ++iter rather than iter++ for std::list iterators.
20861 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
20862 defer deleting them until all nodes in the forest are processed. Do
20863 free even leaf nodes. Change to_process into auto_vec.
20866 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20867 * vec.h (vec_default_construct): Use memset instead of placement new
20868 if BROKEN_VALUE_INITIALIZATION is defined.
20869 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20870 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20873 PR rtl-optimization/83723
20874 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20875 * lra.c (lra_substitute_pseudo): Likewise. If true, use
20876 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
20878 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20880 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20882 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
20884 PR rtl-optimization/81443
20885 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20886 from inner REGs to paradoxical SUBREGs.
20888 2018-02-16 Richard Biener <rguenther@suse.de>
20890 PR tree-optimization/84399
20891 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20892 For operands we can analyze at their definition make sure we can
20893 analyze them at each use as well.
20895 2018-02-16 Richard Biener <rguenther@suse.de>
20897 PR tree-optimization/84190
20898 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20899 volatile accesses if the decl isn't volatile.
20901 2018-02-15 Jason Merrill <jason@redhat.com>
20903 PR c++/84314 - ICE with templates and fastcall attribute.
20904 * attribs.c (build_type_attribute_qual_variant): Don't clobber
20905 TYPE_CANONICAL on an existing type.
20907 2018-02-15 Jakub Jelinek <jakub@redhat.com>
20909 PR tree-optimization/84383
20910 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20911 dstoff nor call operand_equal_p if dstbase is NULL.
20913 PR tree-optimization/84334
20914 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20915 also a CONSTANT_CLASS_P, punt.
20917 2018-02-14 Jim Wilson <jimw@sifive.com>
20919 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20920 first SMALL_OPERAND check. New local min_second_step. Move assert
20921 to where locals are set. Add TARGET_RVC support.
20922 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20924 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
20926 * doc/invoke.texi: Correct -Wformat-overflow code sample.
20928 2018-02-14 Martin Sebor <msebor@redhat.com>
20930 PR tree-optimization/83698
20931 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20932 arrays constrain the offset range to their bounds.
20933 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20934 (builtin_access::overlap): Avoid setting the size of overlap if it's
20936 (maybe_diag_overlap): Also consider arrays when deciding what values
20937 of offsets to include in diagnostics.
20939 2018-02-14 Martin Sebor <msebor@redhat.com>
20942 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20943 that correspond to the kind of a declaration.
20945 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
20948 * config/pa/pa.md: Load address of PIC label using the linkage table
20949 if the label is nonlocal.
20951 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
20953 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20954 warning message if user requests -maltivec=be.
20955 * doc/invoke.texi: Document deprecation of -maltivec=be.
20957 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
20960 * config/rs6000/rs6000-c.c: Update definitions for
20961 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20962 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20964 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20967 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20968 add _get_ssp intrinsics. Remove argument from
20969 __builtin_ia32_rdssp[d|q].
20970 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20971 * config/i386/i386-builtin.def: Remove argument from
20972 __builtin_ia32_rdssp[d|q].
20973 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20974 ix86_expand_special_args_builtin for _rdssp[d|q].
20975 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20976 Clear register before usage.
20977 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20978 Add documentation for new _get_ssp and _inc_ssp intrinsics.
20980 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
20982 PR tree-optimization/84357
20983 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20984 operand 1 of an ARRAY_REF too.
20986 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
20989 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20990 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20992 (set_of_reg): New struct.
20993 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20994 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20995 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20997 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20998 Split into bitclr, bitset, bitinvert patterns if appropriate.
20999 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21000 use rx_fuse_in_memory_bitop.
21001 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21002 to named insn, correct maximum insn length.
21004 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
21007 * machmode.def: Define a complex mode for PARTIAL_INT.
21008 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21010 * doc/rtl.texi: Document CSPImode.
21011 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21013 (msp430_hard_regno_nregs_with_padding): Likewise.
21015 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
21018 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21020 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
21022 PR rtl-optimization/84169
21023 * combine.c (try_combine): New variable split_i2i3. Set it to true if
21024 we generated a parallel as new i3 and we split that to new i2 and i3
21025 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
21026 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
21027 those to i2, not i1. Partially rewrite this scan code.
21029 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21032 * stor-layout.c (place_field): For variable length fields, adjust
21033 offset_align afterwards not just based on the field's alignment,
21034 but also on the size.
21036 PR middle-end/84309
21037 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21038 of exps and logs in the use_exp2 case.
21040 2018-02-13 Jeff Law <law@redhat.com>
21042 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21043 entry for "vector".
21045 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21048 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
21052 * common.opt (gas-loc-support, gas-locview-support): New.
21053 (ginline-points, ginternal-reset-location-views): New.
21054 * doc/invoke.texi: Document them. Use @itemx where intended.
21055 (gvariable-location-views): Adjust.
21056 * target.def (reset_location_view): New.
21057 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21058 (TARGET_RESET_LOCATION_VIEW): New.
21059 * doc/tm.texi: Rebuilt.
21060 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21061 (dwarf2out_default_as_locview_support): New.
21062 (output_asm_line_debug_info): Use option variables.
21063 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21064 (output_loc_list): Likewise.
21065 (add_high_low_attributes): Check option variables.
21066 Don't output entry view attribute in strict mode.
21067 (gen_inlined_subroutine_die): Check option variables.
21068 (dwarf2out_inline_entry): Likewise.
21069 (init_sections_and_labels): Likewise.
21070 (dwarf2out_early_finish): Likewise.
21071 (maybe_reset_location_view): New, from...
21072 (dwarf2out_var_location): ... here. Call it.
21073 * debug.h (dwarf2out_default_as_loc_support): Declare.
21074 (dwarf2out_default_as_locview_support): Declare.
21075 * hooks.c (hook_int_rtx_insn_0): New.
21076 * hooks.h (hook_int_rtx_insn_0): Declare.
21077 * toplev.c (process_options): Take -gas-loc-support and
21078 -gas-locview-support from dwarf2out. Enable
21079 -gvariable-location-views by default only with locview
21080 assembler support. Enable -ginternal-reset-location-views by
21081 default only if the target defines the corresponding hook.
21082 Enable -ginline-points by default if location views are
21083 enabled; force it disabled if statement frontiers are
21085 * tree-inline.c (expand_call_inline): Check option variables.
21086 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21088 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
21090 PR tree-optimization/84321
21091 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21092 handling. Also check whether the anti-range contains any values
21093 that satisfy the mask; switch to a VR_RANGE if not.
21095 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
21098 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21100 2018-02-13 Martin Jambor <mjambor@suse.cz>
21103 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21104 of call statements, also set location of a load to a temporary.
21106 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
21108 * config/rl78/rl78.c (add_vector_labels): New function.
21109 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21110 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21111 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21112 which checks that no arguments are passed.
21113 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21114 * doc/extend.texi: Documentation for the new attribute.
21116 2018-02-13 Andreas Schwab <schwab@suse.de>
21118 * config/riscv/linux.h (CPP_SPEC): Define.
21120 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21123 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21124 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21125 OPTION_MASK_ISA_AES as first argument to def_builtin_const
21126 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21127 instead of OPTION_MASK_ISA_PCLMUL as first argument to
21128 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21129 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21130 temporarily for AES and PCLMUL builtins.
21132 PR tree-optimization/84339
21133 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21134 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21137 PR middle-end/84309
21138 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21139 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21140 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21142 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21144 * omp-simd-clone.h: New file.
21145 * omp-simd-clone.c: Include omp-simd-clone.h.
21146 (expand_simd_clones): No longer static.
21147 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21148 cgraph.h and omp-simd-clone.h.
21149 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21150 (vect_recog_widen_shift_pattern): Formatting fix.
21151 (vect_pattern_recog_1): Don't check optab for calls.
21154 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21155 operands[2] into a REG before using gen_lowpart on it.
21157 2018-02-12 Jeff Law <law@redhat.com>
21160 * config/sh/sh.c (find_barrier): Consider a sibling call
21163 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21164 successfully back substituting a reg.
21166 2018-02-12 Richard Biener <rguenther@suse.de>
21168 PR tree-optimization/84037
21169 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21170 parameter, move visited init to caller.
21171 (vect_slp_analyze_operations): Separate cost from validity
21172 check, initialize visited once for all instances.
21173 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21175 * tree-vect-stmts.c (vect_model_simple_cost): Make early
21177 (vect_model_promotion_demotion_cost): Likewise.
21178 (vectorizable_bswap): Guard cost modeling with !slp_node
21179 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21181 (vectorizable_call): Likewise.
21182 (vectorizable_conversion): Likewise.
21183 (vectorizable_assignment): Likewise.
21184 (vectorizable_shift): Likewise.
21185 (vectorizable_operation): Likewise.
21186 (vectorizable_store): Likewise.
21187 (vectorizable_load): Likewise.
21188 (vectorizable_condition): Likewise.
21189 (vectorizable_comparison): Likewise.
21191 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
21194 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21195 (ASAN_MARK): Fix fnspec to account for return value, change pointer
21196 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21198 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21200 PR middle-end/83665
21201 * params.def (inline-min-speedup): Increase from 8 to 15.
21202 (max-inline-insns-auto): Decrease from 40 to 30.
21203 * ipa-split.c (consider_split): Add some buffer for function to
21204 be considered inlining candidate.
21205 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21208 2018-02-12 Richard Biener <rguenther@suse.de>
21210 PR tree-optimization/84037
21211 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21212 matched stmts if we cannot swap the non-matched ones.
21214 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21216 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21217 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21218 _mm_maskz_scalef_round_ss): New intrinsics.
21219 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21220 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21221 __builtin_ia32_scalefss_round): Remove.
21222 (__builtin_ia32_scalefsd_mask_round,
21223 __builtin_ia32_scalefss_mask_round): New intrinsics.
21224 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21225 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21226 ((match_operand:VF_128 2 "<round_nimm_predicate>"
21227 "<round_constraint>")): Changed to ...
21228 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21229 "<round_scalar_constraint>")): ... this.
21230 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21231 %0, %1, %2<round_op3>}"): Changed to ...
21232 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21233 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21234 %2<round_scalar_mask_op3>}"): ... this.
21235 * config/i386/subst.md (round_scalar_nimm_predicate): New.
21237 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21239 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21240 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21241 (_mm_maskz_sqrt_round_ss): New intrinsics.
21242 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21243 (__builtin_ia32_sqrtsd_mask_round)
21244 (__builtin_ia32_sqrtss_mask_round): New builtins.
21245 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21246 (__builtin_ia32_sqrtss_round): Remove.
21247 (__builtin_ia32_sqrtsd_mask_round)
21248 (__builtin_ia32_sqrtss_mask_round): New builtins.
21249 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21250 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21251 ((match_operand:VF_128 1 "vector_operand"
21252 "xBm,<round_constraint>")): Changed to ...
21253 ((match_operand:VF_128 1 "vector_operand"
21254 "xBm,<round_scalar_constraint>")): ... this.
21255 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21256 %0, %2, %<iptr>1<round_op3>}): Changed to ...
21257 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21258 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21259 %<iptr>1<round_scalar_mask_op3>}): ... this.
21260 ((set_attr "prefix" "<round_prefix>")): Changed to ...
21261 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21263 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
21266 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21267 Cast vec_cmpeq result to correct type.
21268 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21269 Cast vec_cmpgt result to correct type.
21271 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
21273 * final.c (final_scan_insn_1): Renamed from...
21274 (final_scan_insn): ... this. New wrapper, to recover
21275 seen from the outermost call in recursive ones.
21276 * config/sparc/sparc.c (output_return): Drop seen from call.
21277 (output_sibcall): Likewise.
21278 * config/visium/visium.c (output_branch): Likewise.
21280 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
21282 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21285 2018-02-10 Alan Modra <amodra@gmail.com>
21288 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21289 Specify LR as an input.
21291 2018-02-10 Jakub Jelinek <jakub@redhat.com>
21294 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21295 remove_member_access_dummy_vars): New functions.
21296 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21297 lower_omp_1, execute_lower_omp): Use them.
21299 PR rtl-optimization/84308
21300 * shrink-wrap.c (spread_components): Release todo vector.
21302 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
21304 PR rtl-optimization/57193
21305 * ira-color.c (struct allocno_color_data): Add member
21306 conflict_allocno_hard_prefs.
21307 (update_conflict_allocno_hard_prefs): New.
21308 (bucket_allocno_compare_func): Add a preference based on
21309 conflict_allocno_hard_prefs.
21310 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21311 (color_allocnos): Remove a dead code. Initiate
21312 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
21314 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21317 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21318 constraint from =wa to wa. Avoid a subreg on the output operand,
21319 instead use a pseudo and subreg it in a move.
21320 (p9_xxbrd_<mode>): Changed to ...
21321 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21322 (p9_xxbrd_v2df): New expander.
21323 (p9_xxbrw_<mode>): Changed to ...
21324 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21325 (p9_xxbrw_v4sf): New expander.
21327 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21329 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21331 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
21334 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21335 multiply in 32-bit mode.
21336 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21337 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21340 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21342 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21343 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21344 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21345 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21347 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
21350 * dwarf2out.c (is_trivial_indirect_ref): New function.
21351 (dwarf2out_late_global_decl): Do not generate a location
21352 attribute for variables that have a non-trivial DECL_VALUE_EXPR
21353 and that are not defined in the current unit.
21355 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21357 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21358 instead of a libcall for UNORDERED.
21360 2018-02-09 Tamar Christina <tamar.christina@arm.com>
21363 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21364 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21366 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21369 * config/s390/s390.c (s390_set_current_function): Invoke
21370 s390_indirect_branch_settings also if fndecl didn't change.
21372 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21374 * config/rs6000/rs6000.md (blockage): Set length to zero.
21376 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21378 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21380 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21383 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21384 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21388 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21389 PARALLEL incoming that failed vt_get_decl_and_offset check.
21391 PR middle-end/84237
21392 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21393 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21395 (get_variable_section): For decls in named .bss* sections pass true as
21396 second argument to bss_initializer_p.
21398 2018-02-09 Marek Polacek <polacek@redhat.com>
21399 Jakub Jelinek <jakub@redhat.com>
21402 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21403 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
21404 Sync some changes from cxx_fold_indirect_ref.
21406 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21408 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21410 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21411 (BLOCK_INLINE_ENTRY_LABEL): New.
21412 (dwarf2out_var_location): Disregard inline entry markers.
21413 (inline_entry_data): New struct.
21414 (inline_entry_data_hasher): New hashtable type.
21415 (inline_entry_data_hasher::hash): New.
21416 (inline_entry_data_hasher::equal): New.
21417 (inline_entry_data_table): New variable.
21418 (add_high_low_attributes): Add DW_AT_entry_pc and
21419 DW_AT_GNU_entry_view attributes if a pending entry is found
21420 in inline_entry_data_table. Add old entry_pc attribute only
21421 if debug nonbinding markers are disabled.
21422 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21423 markers are enabled.
21424 (block_within_block_p, dwarf2out_inline_entry): New.
21425 (dwarf2out_finish): Check that no entries remained in
21426 inline_entry_data_table.
21427 * final.c (reemit_insn_block_notes): Handle inline entry notes.
21428 (final_scan_insn, notice_source_line): Likewise.
21429 (rest_of_clean_state): Skip inline entry markers.
21430 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21432 * gimple.c (gimple_build_debug_inline_entry): New.
21433 * gimple.h (enum gimple_debug_subcode): Add
21434 GIMPLE_DEBUG_INLINE_ENTRY.
21435 (gimple_build_debug_inline_entry): Declare.
21436 (gimple_debug_inline_entry_p): New.
21437 (gimple_debug_nonbind_marker_p): Adjust.
21438 * insn-notes.def (INLINE_ENTRY): New.
21439 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21440 inline entry marker notes.
21441 (print_insn): Likewise.
21442 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21443 (INSN_DEBUG_MARKER_KIND): Likewise.
21444 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21445 * tree-inline.c (expand_call_inline): Build and insert
21446 debug_inline_entry stmt.
21447 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21448 inline entry blocks early, if nonbind markers are enabled.
21449 (dump_scope_block): Dump fragment info.
21450 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21451 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21452 (gimple_build_debug_inline_entry): New.
21453 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21454 Enable/disable inline entry points too.
21455 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21456 (DEBUG_INSN): Describe inline entry markers.
21458 * common.opt (gvariable-location-views): New.
21459 (gvariable-location-views=incompat5): New.
21460 * config.in: Rebuilt.
21461 * configure: Rebuilt.
21462 * configure.ac: Test assembler for view support.
21463 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21464 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21465 * dwarf2out.c (var_loc_view): New typedef.
21466 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21467 (dwarf2out_locviews_in_attribute): New.
21468 (dwarf2out_locviews_in_loclist): New.
21469 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21470 (enum dw_line_info_opcode): Add LI_adv_address.
21471 (struct dw_line_info_table): Add view.
21472 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21473 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21474 (zero_view_p): New variable.
21475 (ZERO_VIEW_P): New macro.
21476 (output_asm_line_debug_info): New.
21477 (struct var_loc_node): Add view.
21478 (add_AT_view_list, AT_loc_list): New.
21479 (add_var_loc_to_decl): Add view param. Test it against last.
21480 (new_loc_list): Add view params. Record them.
21481 (AT_loc_list_ptr): Handle loc and view lists.
21482 (view_list_to_loc_list_val_node): New.
21483 (print_dw_val): Handle dw_val_class_view_list.
21484 (size_of_die): Likewise.
21485 (value_format): Likewise.
21486 (loc_list_has_views): New.
21487 (gen_llsym): Set vl_symbol too.
21488 (maybe_gen_llsym, skip_loc_list_entry): New.
21489 (dwarf2out_maybe_output_loclist_view_pair): New.
21490 (output_loc_list): Output view list or entries too.
21491 (output_view_list_offset): New.
21492 (output_die): Handle dw_val_class_view_list.
21493 (output_dwarf_version): New.
21494 (output_compilation_unit_header): Use it.
21495 (output_skeleton_debug_sections): Likewise.
21496 (output_rnglists, output_line_info): Likewise.
21497 (output_pubnames, output_aranges): Update version comments.
21498 (output_one_line_info_table): Output view numbers in asm comments.
21499 (dw_loc_list): Determine current endview, pass it to new_loc_list.
21500 Call maybe_gen_llsym.
21501 (loc_list_from_tree_1): Adjust.
21502 (add_AT_location_description): Create view list attribute if
21503 needed, check it's absent otherwise.
21504 (convert_cfa_to_fb_loc_list): Adjust.
21505 (maybe_emit_file): Call output_asm_line_debug_info for test.
21506 (dwarf2out_var_location): Reset views as needed. Precompute
21507 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
21508 attribute. Set view.
21509 (new_line_info_table): Reset next view.
21510 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21511 (dwarf2out_source_line): Likewise. Output view resets and labels to
21512 the assembler, or select appropriate line info opcodes.
21513 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21514 (optimize_string_length): Catch it. Adjust.
21515 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
21516 dw_val_class_view_list, and remove it if no longer needed.
21517 (hash_loc_list): Hash view numbers.
21518 (loc_list_hasher::equal): Compare them.
21519 (optimize_location_lists): Check whether a view list symbol is
21520 needed, and whether the locview attribute is present, and
21521 whether they match. Remove the locview attribute if no longer
21523 (index_location_lists): Call skip_loc_list_entry for test.
21524 (dwarf2out_finish): Call output_asm_line_debug_info for test.
21525 Use output_dwarf_version.
21526 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21527 (struct dw_val_node): Add val_view_list.
21528 * final.c (SEEN_NEXT_VIEW): New.
21529 (set_next_view_needed): New.
21530 (clear_next_view_needed): New.
21531 (maybe_output_next_view): New.
21532 (final_start_function): Rename to...
21533 (final_start_function_1): ... this. Take pointer to FIRST,
21534 add SEEN parameter. Emit param bindings in the initial view.
21535 (final_start_function): Reintroduce SEEN-less interface.
21536 (final): Rename to...
21537 (final_1): ... this. Take SEEN parameter. Output final pending
21538 next view at the end.
21539 (final): Reintroduce seen-less interface.
21540 (final_scan_insn): Output pending next view before switching
21541 sections or ending a block. Mark the next view as needed when
21542 outputting variable locations. Notify debug backend of section
21543 changes, and of location view changes.
21544 (rest_of_handle_final): Adjust.
21545 * toplev.c (process_options): Autodetect value for debug variable
21546 location views option. Warn on incompat5 without -gdwarf-5.
21547 * doc/invoke.texi (gvariable-location-views): New.
21548 (gvariable-location-views=incompat5): New.
21549 (gno-variable-location-views): New.
21551 2018-02-08 David Malcolm <dmalcolm@redhat.com>
21553 PR tree-optimization/84136
21554 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21555 that the result of find_edge is non-NULL.
21557 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21560 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21561 storing integer register in SImode. Fix cost of 256 and 512
21562 byte aligned SSE register store.
21564 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21566 * config/i386/i386.c (ix86_multiplication_cost): Fix
21567 multiplication cost for TARGET_AVX512DQ.
21569 2018-02-08 Marek Polacek <polacek@redhat.com>
21571 PR tree-optimization/84238
21572 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21575 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21577 PR tree-optimization/84265
21578 * tree-vect-stmts.c (vectorizable_store): Don't treat
21579 VMAT_CONTIGUOUS accesses as grouped.
21580 (vectorizable_load): Likewise.
21582 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21584 PR tree-optimization/81635
21585 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21586 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21587 (test_round_for_mask): New functions.
21588 (wide_int_cc_tests): Call test_round_for_mask.
21589 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21590 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21591 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21592 range returned by get_range_info.
21594 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21597 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21598 * symtab.c: Include builtins.h
21599 (symtab_node::output_to_lto_symbol_table_p): Move here
21600 from lto-streamer-out.c:output_symbol_p.
21601 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21602 (output_symbol_p): Move all logic to symtab.c
21603 (produce_symtab): Update.
21605 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21607 * config/s390/s390-opts.h (enum indirect_branch): Define.
21608 * config/s390/s390-protos.h (s390_return_addr_from_memory)
21609 (s390_indirect_branch_via_thunk)
21610 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21611 (enum s390_indirect_branch_type): Define.
21612 * config/s390/s390.c (struct s390_frame_layout, struct
21613 machine_function): Remove.
21614 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21615 (indirect_branch_table_label_no, indirect_branch_table_name):
21617 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21618 (enum s390_indirect_branch_option): Define.
21619 (s390_return_addr_from_memory): New function.
21620 (s390_handle_string_attribute): New function.
21621 (s390_attribute_table): Add new attribute handler.
21622 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21623 (s390_indirect_branch_via_thunk): New function.
21624 (s390_indirect_branch_via_inline_thunk): New function.
21625 (s390_function_ok_for_sibcall): When jumping via thunk disallow
21626 sibling call optimization for non z10 compiles.
21627 (s390_emit_call): Force indirect branch target to be a single
21628 register. Add r1 clobber for non-z10 compiles.
21629 (s390_emit_epilogue): Emit return jump via return_use expander.
21630 (s390_reorg): Handle JUMP_INSNs as execute targets.
21631 (s390_option_override_internal): Perform validity checks for the
21632 new command line options.
21633 (s390_indirect_branch_attrvalue): New function.
21634 (s390_indirect_branch_settings): New function.
21635 (s390_set_current_function): Invoke s390_indirect_branch_settings.
21636 (s390_output_indirect_thunk_function): New function.
21637 (s390_code_end): Implement target hook.
21638 (s390_case_values_threshold): Implement target hook.
21639 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21641 * config/s390/s390.h (struct s390_frame_layout)
21642 (struct machine_function): Move here from s390.c.
21643 (TARGET_INDIRECT_BRANCH_NOBP_RET)
21644 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21645 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21646 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21647 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21648 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21649 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21650 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21651 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21652 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21653 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21654 (mnemonic attribute): Add values which aren't recognized
21656 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21657 pattern for branch conversion. Fix mnemonic attribute.
21658 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21659 indirect branch via thunk if requested.
21660 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21661 ("*indirect_jump"): Disable for branch conversion using out of
21663 ("indirect_jump_via_thunk<mode>_z10")
21664 ("indirect_jump_via_thunk<mode>")
21665 ("indirect_jump_via_inlinethunk<mode>_z10")
21666 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21667 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21668 ("casesi_jump_via_inlinethunk<mode>_z10")
21669 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21670 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21671 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21672 ("*indirect2_jump"): Disable for branch conversion.
21673 ("casesi_jump"): Turn into expander and expand patterns for branch
21675 ("return_use"): New expander.
21676 ("*return"): Emit return via thunk and rename it to ...
21677 ("*return<mode>"): ... this one.
21678 * config/s390/s390.opt: Add new options and and enum for the
21681 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21683 * lra-constraints.c (match_reload): Unconditionally use
21684 gen_lowpart_SUBREG, rather than selecting between that
21685 and equivalent gen_rtx_SUBREG code.
21687 2018-02-08 Richard Biener <rguenther@suse.de>
21689 PR tree-optimization/84233
21690 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21691 changed flag instead of boguously re-using phi_inserted.
21693 2018-02-08 Martin Jambor <mjambor@suse.cz>
21695 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21696 static local variables.
21698 2018-02-08 Richard Biener <rguenther@suse.de>
21700 PR tree-optimization/84278
21701 * tree-vect-stmts.c (vectorizable_store): When looking for
21702 smaller vector types to perform grouped strided loads/stores
21703 make sure the mode is supported by the target.
21704 (vectorizable_load): Likewise.
21706 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21708 * config/aarch64/aarch64.c (aarch64_components_for_bb):
21709 Increase LDP/STP opportunities by adding adjacent callee-saves.
21711 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21713 PR rtl-optimization/84068
21714 PR rtl-optimization/83459
21715 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21717 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
21719 PR tree-optimization/84224
21720 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21721 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21722 non-zero arguments.
21724 2018-02-07 Iain Sandoe <iain@codesourcery.com>
21727 * config/rs6000/altivec.md (*restore_world): Remove LR use.
21728 * config/rs6000/predicates.md (restore_world_operation): Adjust op
21729 count, remove one USE.
21731 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21733 * doc/install.texi (Configuration): Document the
21734 --with-long-double-format={ibm,ieee} PowerPC configuration
21738 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21739 Convert from define_expand to be define_insn_and_split. Rework
21740 float/double/_Float128 conversions to QI/HI/SImode to work with
21741 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
21742 conversions to QI/HImode types did a store and then a load to
21743 truncate the value. For conversions to VSX registers, don't split
21744 the insn, instead emit the code directly. Use the code iterator
21745 any_fix to combine signed and unsigned conversions.
21746 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21747 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21748 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21749 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21750 (fix_<mode>di2_hw): Likewise.
21751 (fixuns_<mode>di2_hw): Likewise.
21752 (fix_<mode>si2_hw): Likewise.
21753 (fixuns_<mode>si2_hw): Likewise.
21754 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21755 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21756 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21757 fix<uns>_trunc<SFDF:mode>si2_p8.
21758 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21759 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21760 (fix<uns>_<mode>_mem): Likewise.
21761 (fctiw<u>z_<mode>_mem): Likewise.
21762 (fix<uns>_<mode>_mem): Likewise.
21763 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21764 the register allocator from doing a direct move to the GPRs to do
21765 a store, and instead use the ISA 3.0 store byte/half-word from
21766 vector register instruction. For IEEE 128-bit floating point,
21767 also optimize stores of 32-bit ints.
21768 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21770 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21772 * genextract.c (push_pathstr_operand): New function to support
21774 (walk_rtx): Call push_pathstr_operand.
21775 (print_path): Support [a-zA-Z].
21777 2018-02-07 Richard Biener <rguenther@suse.de>
21779 PR tree-optimization/84037
21780 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21781 (cse_and_gimplify_to_preheader): Declare.
21782 (vect_get_place_in_interleaving_chain): Likewise.
21783 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21785 (_loop_vec_info::~_loop_vec_info): Delete it.
21786 (cse_and_gimplify_to_preheader): New function.
21787 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21788 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21789 (vectorizable_load): Likewise. For grouped stores always base
21790 the IV on the first element.
21791 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21792 condition before gimplifying.
21794 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21796 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21797 *DIV_EXPR and *MOD_EXPR.
21799 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
21802 * config/i386/i386.c (ix86_option_override_internal): Mask out
21803 the CF_SET bit when checking -fcf-protection.
21805 2018-02-07 Tom de Vries <tom@codesourcery.com>
21808 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21811 2018-02-07 Richard Biener <rguenther@suse.de>
21813 PR tree-optimization/84204
21814 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21817 PR tree-optimization/84205
21818 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21819 special-case isl_ast_op_zdiv_r.
21821 PR tree-optimization/84223
21822 * graphite-scop-detection.c (gather_bbs::before_dom_children):
21823 Only add conditions from within the region.
21824 (gather_bbs::after_dom_children): Adjust.
21826 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
21829 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21830 * config/avr/avr.md: Only post-reload split REG-REG moves if
21831 either register is GENERAL_REG_P.
21833 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21835 PR tree-optimization/84235
21836 * tree-ssa-scopedtables.c
21837 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21838 if the subtraction is performed in floating point type where NaNs are
21839 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21840 build 1. Formatting fix.
21842 2018-02-06 Jakub Jelinek <jakub@redhat.com>
21845 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21846 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21847 and skip it regardless of bb boundaries. Use CALL_P macro,
21848 don't test INSN_P (insn) together with CALL_P or JUMP_P check
21849 unnecessarily, formatting fix.
21851 2018-02-06 Michael Collison <michael.collison@arm.com>
21853 * config/arm/thumb2.md:
21854 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21855 (*thumb_mov_notscc): Ditto.
21857 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
21860 * config/rs6000/rs6000.md (su code attribute): Use "u" for
21861 unsigned_fix, not "s".
21863 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21865 * configure.ac (gcc_fn_eh_frame_ro): New function.
21866 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21867 correct .eh_frame permissions.
21868 * configure: Regenerate.
21870 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
21872 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21873 irrelevant options.
21875 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21877 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21878 Display warning message for -mno-speculate-indirect-jumps.
21880 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
21882 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21884 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21886 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
21888 PR tree-optimization/84225
21889 * tree-eh.c (find_trapping_overflow): Only call
21890 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21892 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21895 * config/i386/i386.c: Reimplement the check of possible options
21896 -mibt/-mshstk conbination. Change error messages.
21897 * doc/invoke.texi: Fix a typo: remove extra '='.
21899 2018-02-06 Marek Polacek <polacek@redhat.com>
21901 PR tree-optimization/84228
21902 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21904 2018-02-06 Tamar Christina <tamar.christina@arm.com>
21907 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21908 emitted arch directives.
21909 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21910 __ARM_FEATURE_COPROC before changing architectures.
21912 2018-02-06 Richard Biener <rguenther@suse.de>
21914 * config/i386/i386.c (print_reg): Fix typo.
21915 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21917 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
21919 * configure: Regenerate.
21921 2018-02-05 Martin Sebor <msebor@redhat.com>
21923 PR tree-optimization/83369
21924 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21927 2018-02-05 Martin Liska <mliska@suse.cz>
21929 * doc/invoke.texi: Cherry-pick upstream r323995.
21931 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
21933 * ira.c (ira_init_register_move_cost): Adjust comment.
21935 2018-02-05 Martin Liska <mliska@suse.cz>
21937 PR gcov-profile/84137
21938 * doc/gcov.texi: Fix typo in documentation.
21940 2018-02-05 Martin Liska <mliska@suse.cz>
21942 PR gcov-profile/83879
21943 * doc/gcov.texi: Document necessity of --dynamic-list-data when
21944 using dlopen functionality.
21946 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
21948 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21949 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21950 _mm_maskz_range_ss, _mm_mask_range_round_ss,
21951 _mm_maskz_range_round_ss): New intrinsics.
21952 (__builtin_ia32_rangesd128_round)
21953 (__builtin_ia32_rangess128_round): Remove.
21954 (__builtin_ia32_rangesd128_mask_round,
21955 __builtin_ia32_rangess128_mask_round): New builtins.
21956 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21957 __builtin_ia32_rangess128_round): Remove.
21958 (__builtin_ia32_rangesd128_mask_round,
21959 __builtin_ia32_rangess128_mask_round): New builtins.
21960 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21961 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21962 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21963 "<round_saeonly_constraint>")): Changed to ...
21964 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21965 "<round_saeonly_scalar_constraint>")): ... this.
21966 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21967 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21968 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21969 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21970 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21972 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
21974 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21976 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21977 Remove all values except native, 8540 and 8548.
21979 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
21981 * config/i386/i386.c (ix86_output_function_return): Pass
21982 INVALID_REGNUM, instead of -1, as invalid register number to
21983 indirect_thunk_name and output_indirect_thunk.
21985 2018-02-02 Julia Koval <julia.koval@intel.com>
21987 * config.gcc: Add -march=icelake.
21988 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21989 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21990 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21991 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21992 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21993 (processor_target_table): Add icelake.
21994 (ix86_option_override_internal): Handle new PTAs.
21995 (get_builtin_code_for_version): Handle icelake.
21996 (M_INTEL_COREI7_ICELAKE): New.
21997 (fold_builtin_cpu): Handle icelake.
21998 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21999 * doc/invoke.texi: Add -march=icelake.
22001 2018-02-02 Julia Koval <julia.koval@intel.com>
22003 * config/i386/i386.c (ix86_option_override_internal): Change flags type
22004 to wide_int_bitmask.
22005 * wide-int-bitmask.h: New.
22007 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22010 * config/i386/i386.md: Replace Pmode with word_mode in
22011 builtin_setjmp_setup and builtin_longjmp to support x32.
22013 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
22017 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22019 (rs6000_supported_cpu_names): New static variable.
22020 (linux_cpu_translation_table): Likewise.
22021 (elf_platform) <cpu>: Define new static variable and use it.
22022 Translate kernel AT_PLATFORM name to canonical name if needed.
22023 Error if platform name is unknown.
22025 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
22028 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22030 2018-02-01 Jeff Law <law@redhat.com>
22033 * config/i386/i386.c (release_scratch_register_on_entry): Add new
22034 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
22035 the scratch if RELEASE_VIA_POP is false.
22036 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22037 If we have to save a temporary register, decrement SIZE appropriately.
22038 Pass new arguments to release_scratch_register_on_entry.
22039 (ix86_adjust_stack_and_probe): Likewise.
22040 (ix86_emit_probe_stack_range): Pass new arguments to
22041 release_scratch_register_on_entry.
22043 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
22045 PR rtl-optimization/84157
22046 * combine.c (change_zero_ext): Use REG_P predicate in
22047 front of HARD_REGISTER_P predicate.
22049 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
22051 * config/avr/avr.c (avr_option_override): Move disabling of
22052 -fdelete-null-pointer-checks to...
22053 * common/config/avr/avr-common.c (avr_option_optimization_table):
22056 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22058 PR tree-optimization/81635
22059 * tree-data-ref.c (split_constant_offset_1): For types that
22060 wrap on overflow, try to use range info to prove that wrapping
22063 2018-02-01 Renlin Li <renlin.li@arm.com>
22066 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22067 TAILCALL_ADDR_REGS.
22068 (aarch64_register_move_cost): Likewise.
22069 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22070 TAILCALL_ADDR_REGS.
22071 (REG_CLASS_NAMES): Likewise.
22072 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22073 TAILCALL_ADDR_REGS. Remove IP registers.
22074 * config/aarch64/aarch64.md (Ucs): Update register constraint.
22076 2018-02-01 Richard Biener <rguenther@suse.de>
22078 * domwalk.h (dom_walker::dom_walker): Add additional constructor
22079 for specifying RPO order and allow NULL for that.
22080 * domwalk.c (dom_walker::dom_walker): Likewise.
22081 (dom_walker::walk): Handle NULL RPO order.
22082 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22084 (rewrite_update_dom_walker): Likewise.
22085 (mark_def_dom_walker): Likewise.
22087 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22089 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22090 (aarch64_maybe_expand_sve_subreg_move): Declare.
22091 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22092 * config/aarch64/predicates.md (aarch64_any_register_operand): New
22094 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22095 that are semantically a reverse operation.
22096 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22097 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22098 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22100 (aarch64_can_change_mode_class): For big-endian, forbid changes
22101 between two SVE modes if they have different element sizes.
22103 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22105 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22106 the TImode handling for big-endian targets.
22108 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22110 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22111 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
22113 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22114 Remove BSWAP handing for big-endian targets and use the form of
22115 LD1RQ appropariate for the mode.
22117 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22119 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22120 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22121 duplicated element.
22123 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22126 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22127 check for operands that need to go through aarch64_sve_reload_be.
22129 2018-02-01 Jakub Jelinek <jakub@redhat.com>
22131 PR tree-optimization/81661
22132 PR tree-optimization/84117
22133 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22134 * tree-eh.c: Include gimplify.h.
22135 (find_trapping_overflow, replace_trapping_overflow,
22136 rewrite_to_non_trapping_overflow): New functions.
22137 * tree-vect-loop.c: Include tree-eh.h.
22138 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22139 * tree-data-ref.c: Include tree-eh.h.
22140 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22142 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
22144 PR rtl-optimization/84123
22145 * combine.c (change_zero_ext): Check if hard register satisfies
22146 can_change_dest_mode before calling gen_lowpart_SUBREG.
22148 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
22151 * ira.c (ira_init_register_move_cost): Remove assert.
22153 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22155 PR rtl-optimization/84071
22156 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22157 * doc/tm.texi: Regenerate.
22159 2018-01-31 Richard Biener <rguenther@suse.de>
22161 PR tree-optimization/84132
22162 * tree-data-ref.c (analyze_miv_subscript): Properly
22163 check whether evolution_function_is_affine_multivariate_p
22164 before calling gcd_of_steps_may_divide_p.
22166 2018-01-31 Julia Koval <julia.koval@intel.com>
22169 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22170 * config/i386/i386.md (rdpid_rex64) New.
22171 (rdpid): Make 32bit only.
22173 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
22176 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22177 an IDENTIFIER_NODE for FUNCTION_TYPE's.
22179 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22182 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22184 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22186 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22188 PR rtl-optimization/84071
22189 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22190 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22192 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22194 * config/arc/arc.c (arc_handle_aux_attribute): New function.
22195 (arc_attribute_table): Add 'aux' attribute.
22196 (arc_in_small_data_p): Consider aux like variables.
22197 (arc_is_aux_reg_p): New function.
22198 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22199 (arc_get_aux_arg): New function.
22200 (prepare_move_operands): Handle aux-register access.
22201 (arc_handle_aux_attribute): New function.
22202 * doc/extend.texi (ARC Variable attributes): Add subsection.
22204 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22206 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22207 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22208 (arc_attribute_table): Add 'uncached' attribute.
22209 (arc_print_operand): Print '.di' flag for uncached memory
22211 (arc_in_small_data_p): Do not consider for small data the uncached
22213 (arc_is_uncached_mem_p): New function.
22214 * config/arc/predicates.md (compact_store_memory_operand): Check
22215 for uncached memory accesses.
22216 (nonvol_nonimm_operand): Likewise.
22217 * doc/extend.texi (ARC Type Attribute): New subsection.
22219 2018-01-31 Jakub Jelinek <jakub@redhat.com>
22222 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22223 falign-loops=): Add Optimization flag.
22225 2018-01-30 Jeff Law <law@redhat.com>
22228 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22229 INT_REGISTERS_SAVED. Check it prior to calling
22230 get_scratch_register_on_entry.
22231 (ix86_adjust_stack_and_probe): Similarly.
22232 (ix86_emit_probe_stack_range): Similarly.
22233 (ix86_expand_prologue): Corresponding changes.
22235 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22238 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22239 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22241 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
22244 * lra-constraints.c (curr_insn_transform): Process AND in the
22247 2018-01-30 Jakub Jelinek <jakub@redhat.com>
22249 PR rtl-optimization/83986
22250 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22251 dependence against last_pending_memory_flush in addition to
22252 pending_jump_insns.
22254 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
22256 PR tree-optimization/81611
22257 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22260 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22263 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22266 2018-01-30 Richard Biener <rguenther@suse.de>
22267 Jakub Jelinek <jakub@redhat.com>
22269 PR tree-optimization/84111
22270 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22271 inner loops added during recursion, as they don't have up-to-date
22274 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22277 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22278 (can_inline_edge_by_limits_p): ... here.
22279 (can_early_inline_edge_p, check_callers,
22280 update_caller_keys, update_callee_keys, recursive_inlining,
22281 add_new_edges_to_heap, speculation_useful_p,
22282 inline_small_functions,
22283 inline_small_functions, flatten_function,
22284 inline_to_all_callers_1): Update.
22286 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22288 * profile-count.c (profile_count::combine_with_ipa_count): Handle
22291 2018-01-30 Richard Biener <rguenther@suse.de>
22293 PR tree-optimization/83008
22294 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22295 invariant and constant vector uses in stmts when they need
22296 more than one stmt.
22298 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22301 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22302 * configure: Regenerate.
22304 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22306 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22308 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22309 Use gen_rtx_REG rather than gen_lowpart.
22311 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22313 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22314 rather than 0 when creating partial subregs.
22316 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22318 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22321 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
22324 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22325 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22326 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22327 flags. This restores the settings used before the 2017-07-24.
22328 Turning off pre increment/decrement/modify allows IVOPTS to
22329 optimize DF/SF loops where the index is an int.
22331 2018-01-29 Richard Biener <rguenther@suse.de>
22332 Kelvin Nilsen <kelvin@gcc.gnu.org>
22335 * tree-vect-stmts.c (vectorizable_call): Don't call
22336 targetm.vectorize_builtin_md_vectorized_function if callee is
22339 2018-01-22 Carl Love <cel@us.ibm.com>
22341 * doc/extend.tex: Fix typo in second arg in
22342 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22344 2018-01-29 Richard Biener <rguenther@suse.de>
22346 PR tree-optimization/84086
22347 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22348 (flush_ssaname_freelist): When SSA names were released reset
22349 the SCEV hash table.
22351 2018-01-29 Richard Biener <rguenther@suse.de>
22353 PR tree-optimization/84057
22354 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22355 removed paths when removing edges.
22357 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
22359 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22360 -mfunction-return=@var{choice}.
22362 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22364 PR diagnostic/84034
22365 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22366 Handle CR like TAB.
22367 (layout::print_source_line): Likewise.
22368 (test_get_line_width_without_trailing_whitespace): Add test cases.
22370 2018-01-27 Jakub Jelinek <jakub@redhat.com>
22372 PR middle-end/84040
22373 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22376 2018-01-26 Jim Wilson <jimw@sifive.com>
22378 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22380 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22383 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22385 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22386 and CMP + SUB-immediate -> SUBS.
22388 2018-01-26 Martin Sebor <msebor@redhat.com>
22390 PR tree-optimization/83896
22391 * tree-ssa-strlen.c (get_string_len): Rename...
22392 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
22393 Avoid assuming length is constant.
22394 (handle_char_store): Use HOST_WIDE_INT for string length.
22396 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
22399 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22400 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22402 2018-01-26 Richard Biener <rguenther@suse.de>
22404 PR rtl-optimization/84003
22405 * dse.c (record_store): Only record redundant stores when
22406 the earlier store aliases at least all accesses the later one does.
22408 2018-01-26 Jakub Jelinek <jakub@redhat.com>
22410 PR rtl-optimization/83985
22411 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22412 REG_CFA_RESTORE insns.
22413 (delete_unmarked_insns): Don't ignore separate shrink wrapping
22414 REG_CFA_RESTORE insns here.
22417 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22418 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22420 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22422 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22423 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22424 (arc_init): Likewise.
22425 (arc_override_options): Likewise.
22426 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22428 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22430 * config/arc/arc.h (TARGET_DBNZ): Define.
22431 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22432 properly set the tune attribute.
22433 (dbnz): Use TARGET_DBNZ guard.
22434 * config/arc/arc.opt (mtune): Add core3 option.
22436 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22438 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22439 recognize new pic like addresses.
22440 (arc_delegitimize_address): Clean up.
22442 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22444 * config/arc/arc-arches.def: Option mrf16 valid for all
22446 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22447 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22448 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22449 * config/arc/arc-tables.opt: Regenerate.
22450 * config/arc/arc.c (arc_conditional_register_usage): Handle
22451 reduced register file case.
22452 (arc_file_start): Set must have build attributes.
22453 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22454 mrf16 option value.
22455 * config/arc/arc.opt (mrf16): Add new option.
22456 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22457 * config/arc/genmultilib.awk: Handle new mrf16 option.
22458 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22459 * config/arc/t-multilib: Regenerate.
22460 * doc/invoke.texi (ARC Options): Document mrf16 option.
22462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22464 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22465 * config/arc/arc.c (arc_handle_secure_attribute): New function.
22466 (arc_attribute_table): Add 'secure_call' attribute.
22467 (arc_print_operand): Print secure call operand.
22468 (arc_function_ok_for_sibcall): Don't optimize tail calls when
22470 (arc_is_secure_call_p): New function. * config/arc/arc.md
22471 (call_i): Add support for sjli instruction.
22472 (call_value_i): Likewise.
22473 * config/arc/constraints.md (Csc): New constraint.
22475 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22476 John Eric Martin <John.Martin@emmicro-us.com>
22478 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22479 * config/arc/arc.c (_arc_jli_section): New struct.
22480 (arc_jli_section): New type.
22481 (rc_jli_sections): New static variable.
22482 (arc_handle_jli_attribute): New function.
22483 (arc_attribute_table): Add jli_always and jli_fixed attribute.
22484 (arc_file_end): New function.
22485 (TARGET_ASM_FILE_END): Define.
22486 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22487 (arc_add_jli_section): New function.
22488 (jli_call_scan): Likewise.
22489 (arc_reorg): Call jli_call_scan.
22490 (arc_output_addsi): Remove 'S' from printing asm operand.
22491 (arc_is_jli_call_p): New function.
22492 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22494 (movhi_insn): Likewise.
22495 (movsi_insn): Likewise.
22496 (movsi_set_cc_insn): Likewise.
22497 (loadqi_update): Likewise.
22498 (load_zeroextendqisi_update): Likewise.
22499 (load_signextendqisi_update): Likewise.
22500 (loadhi_update): Likewise.
22501 (load_zeroextendhisi_update): Likewise.
22502 (load_signextendhisi_update): Likewise.
22503 (loadsi_update): Likewise.
22504 (loadsf_update): Likewise.
22505 (movsicc_insn): Likewise.
22506 (bset_insn): Likewise.
22507 (bxor_insn): Likewise.
22508 (bclr_insn): Likewise.
22509 (bmsk_insn): Likewise.
22510 (bicsi3_insn): Likewise.
22511 (cmpsi_cc_c_insn): Likewise.
22512 (movsi_ne): Likewise.
22513 (movsi_cond_exec): Likewise.
22514 (clrsbsi2): Likewise.
22515 (norm_f): Likewise.
22525 (call_i): Remove 'S' asm letter, add jli instruction.
22526 (call_value_i): Likewise.
22527 * config/arc/arc.op (mjli-always): New option.
22528 * config/arc/constraints.md (Cji): New constraint.
22529 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22531 (subsf3_fpx): Likewise.
22532 (mulsf3_fpx): Likewise.
22533 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22535 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22536 function attrbutes.
22537 * doc/invoke.texi (ARC): Document mjli-always option.
22539 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
22541 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22542 avoid addition with 0 and use incw and decw where possible.
22544 2018-01-26 Richard Biener <rguenther@suse.de>
22546 PR tree-optimization/81082
22547 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22548 association if it requires casting to unsigned.
22549 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22550 from fold_plusminus_mult_expr to catch important cases late when
22551 range info is available.
22553 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22555 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22556 * configure.ac (hidden_linkonce): New test.
22557 * configure: Regenerate.
22558 * config.in: Regenerate.
22560 2018-01-26 Julia Koval <julia.koval@intel.com>
22562 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22563 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22564 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22565 _mm_mask_bitshuffle_epi64_mask): Fix type.
22566 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22567 USI_FTYPE_V4DI_V4DI_USI): Remove.
22568 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22569 __builtin_ia32_vpshufbitqmb256_mask,
22570 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22571 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22572 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22574 2018-01-26 Alan Modra <amodra@gmail.com>
22577 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22578 UNSPEC_VBPERMQ. Sort other unspecs.
22580 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
22582 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22584 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
22586 PR middle-end/83055
22587 * predict.c (drop_profile): Do not push/pop cfun; update also
22589 (handle_missing_profiles): Fix logic looking for zero profiles.
22591 2018-01-25 Jakub Jelinek <jakub@redhat.com>
22593 PR middle-end/83977
22594 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22595 on functions with #pragma omp declare simd or functions with simd
22597 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22598 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22599 Remove trailing \n from warning_at calls.
22601 2018-01-25 Tom de Vries <tom@codesourcery.com>
22604 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22605 for neutered workers.
22607 2018-01-24 Joseph Myers <joseph@codesourcery.com>
22610 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22611 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22613 2018-01-24 Jeff Law <law@redhat.com>
22616 * i386.c (get_probe_interval): Move to earlier point.
22617 (ix86_compute_frame_layout): If -fstack-clash-protection and
22618 the frame is larger than the probe interval, then use pushes
22619 to save registers rather than reg->mem moves.
22620 (ix86_expand_prologue): Remove conditional for int_registers_saved
22623 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
22626 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22627 min/max for never referenced object.
22629 2018-01-24 Jakub Jelinek <jakub@redhat.com>
22631 PR middle-end/83977
22632 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22634 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22635 attributes from DECL_ATTRIBUTES (decl) without affecting
22636 DECL_ATTRIBUTES (current_function_decl).
22637 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22638 functions with non-NULL DECL_ABSTRACT_ORIGIN.
22640 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
22642 PR tree-optimization/83979
22643 * fold-const.c (fold_comparison): Use constant_boolean_node
22644 instead of boolean_{true,false}_node.
22646 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
22648 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22651 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22653 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22654 Simplify the clause that sets the length attribute.
22655 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22656 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22657 clause that sets the length attribute.
22658 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22660 2018-01-24 Tom de Vries <tom@codesourcery.com>
22663 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22664 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22665 Add strict parameter.
22666 (prevent_branch_around_nothing): Insert dummy insn between branch to
22667 label and label with no ptx insn inbetween.
22668 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22670 2018-01-24 Tom de Vries <tom@codesourcery.com>
22673 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22674 for neutered threads in warp.
22675 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22677 2018-01-24 Richard Biener <rguenther@suse.de>
22679 PR tree-optimization/83176
22680 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22683 2018-01-24 Richard Biener <rguenther@suse.de>
22685 PR tree-optimization/82819
22686 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22687 code generating pluses that are no-ops in the target precision.
22689 2018-01-24 Richard Biener <rguenther@suse.de>
22691 PR middle-end/84000
22692 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22694 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22696 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22697 to merge probabilities.
22698 * predict.c (probably_never_executed): Also mark as cold functions
22699 with global 0 profile and guessed local profile.
22700 * profile-count.c (profile_probability::combine_with_count): New
22702 * profile-count.h (profile_probability::operator*,
22703 profile_probability::operator*=, profile_probability::operator/,
22704 profile_probability::operator/=): Reduce precision to adjusted
22705 and set value to guessed on contradictory divisions.
22706 (profile_probability::combine_with_freq): Remove.
22707 (profile_probability::combine_wiht_count): Declare.
22708 (profile_count::force_nonzero):: Set to adjusted.
22709 (profile_count::probability_in):: Set quality to adjusted.
22710 * tree-ssa-tail-merge.c (replace_block_by): Use
22711 combine_with_count.
22713 2018-01-23 Andrew Waterman <andrew@sifive.com>
22714 Jim Wilson <jimw@sifive.com>
22716 * config/riscv/riscv.c (riscv_stack_boundary): New.
22717 (riscv_option_override): Set riscv_stack_boundary. Handle
22718 riscv_preferred_stack_boundary_arg.
22719 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22720 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22721 (STACK_BOUNDARY): Set to riscv_stack_boundary.
22722 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22723 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22724 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22726 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
22729 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22730 of struct ix86_frame.
22731 (ix86_expand_epilogue): Likewise. Add a local variable for
22732 the reg_save_offset field in struct ix86_frame.
22734 2018-01-23 Bin Cheng <bin.cheng@arm.com>
22736 PR tree-optimization/82604
22737 * tree-loop-distribution.c (enum partition_kind): New enum item
22738 PKIND_PARTIAL_MEMSET.
22739 (partition_builtin_p): Support above new enum item.
22740 (generate_code_for_partition): Ditto.
22741 (compute_access_range): Differentiate cases that equality can be
22742 proven at all loops, the innermost loops or no loops.
22743 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22744 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
22745 (finalize_partitions, distribute_loop): Don't fuse partition of
22746 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22747 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22748 parloop is enabled.
22750 2018-01-23 Martin Liska <mliska@suse.cz>
22752 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22753 order to ignore the predictor.
22754 (PRED_POLYMORPHIC_CALL): Likewise.
22755 (PRED_RECURSIVE_CALL): Likewise.
22757 2018-01-23 Martin Liska <mliska@suse.cz>
22759 * tree-profile.c (tree_profiling): Print function header to
22760 aware reader which function we are working on.
22761 * value-prof.c (gimple_find_values_to_profile): Do not print
22762 not interesting value histograms.
22764 2018-01-23 Martin Liska <mliska@suse.cz>
22766 * profile-count.h (enum profile_quality): Add
22767 profile_uninitialized as the first value. Do not number values
22768 as they are zero based.
22769 (profile_count::verify): Update sanity check.
22770 (profile_probability::verify): Likewise.
22772 2018-01-23 Nathan Sidwell <nathan@acm.org>
22774 * doc/invoke.texi (ffor-scope): Deprecate.
22776 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22778 PR tree-optimization/83510
22779 * domwalk.c (set_all_edges_as_executable): New function.
22780 (dom_walker::dom_walker): Convert bool param
22781 "skip_unreachable_blocks" to enum reachability. Move setup of
22782 edge flags to set_all_edges_as_executable and only do it when
22783 reachability is REACHABLE_BLOCKS.
22784 * domwalk.h (enum dom_walker::reachability): New enum.
22785 (dom_walker::dom_walker): Convert bool param
22786 "skip_unreachable_blocks" to enum reachability.
22787 (set_all_edges_as_executable): New decl.
22788 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22789 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22791 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22792 but converting true to REACHABLE_BLOCKS.
22793 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22795 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22796 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22797 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22799 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22800 if check_all_array_refs will be called.
22802 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22804 * tree.c (selftest::test_location_wrappers): Add more test
22807 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22809 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22810 (selftest::test_bit_in_range): Likewise.
22812 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22815 * doc/sourcebuild.texi (vect_float): Say that the selector
22816 only describes the situation when -funsafe-math-optimizations is on.
22817 (vect_float_strict): Document.
22819 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22821 PR tree-optimization/83965
22822 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22823 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22824 instead of checking only for a reduction.
22825 (vect_recog_widen_sum_pattern): Likewise.
22827 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22829 * predict.c (probably_never_executed): Only use precise profile info.
22830 (compute_function_frequency): Skip after inlining hack since we now
22831 have quality checking.
22833 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22835 * profile-count.h (profile_probability::very_unlikely,
22836 profile_probability::unlikely, profile_probability::even): Set
22837 precision to guessed.
22839 2018-01-23 Richard Biener <rguenther@suse.de>
22841 PR tree-optimization/83963
22842 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22843 Properly terminate dominator walk when crossing the exit edge not
22844 when visiting its source block.
22846 2018-01-23 Jakub Jelinek <jakub@redhat.com>
22849 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22850 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22852 2018-01-22 Jakub Jelinek <jakub@redhat.com>
22854 PR tree-optimization/83957
22855 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
22856 semicolon after for body surrounded by braces.
22858 PR tree-optimization/83081
22859 * profile-count.h (profile_probability::split): New method.
22860 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22861 Use profile_probability::split.
22862 (do_compare_rtx_and_jump): Fix adjustment of probabilities
22863 when splitting a single conditional jump into 2.
22865 2018-01-22 David Malcolm <dmalcolm@redhat.com>
22867 PR tree-optimization/69452
22868 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22871 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22873 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22874 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22875 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22877 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22879 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22881 * config/rl78/rl78.md (movdi): New define_expand.
22882 * config/rl78/rl78.c (rl78_split_movdi): New function.
22884 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
22887 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22889 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22890 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22891 128-bit to produce an UNSPEC move to get the double word with the
22892 signbit and then a shift directly to do signbit.
22893 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22894 implementation with a new version that just does either a direct
22895 move or a regular move. Move memory interface to separate insns.
22896 Move insns so they are next to the expander.
22897 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22898 with signbit move. Split big and little endian case.
22899 (signbit<mode>2_dm_mem_le): Likewise.
22900 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22901 (signbit<mode>2_dm2): Likewise.
22903 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22905 * config/rl78/rl78.md (anddi3): New define_expand.
22907 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22909 * config/rl78/rl78.md (umindi3): New define_expand.
22911 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22913 * config/rl78/rl78.md (smindi3): New define_expand.
22915 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22917 * config/rl78/rl78.md (smaxdi3): New define_expand.
22919 2018-01-22 Carl Love <cel@us.ibm.com>
22921 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22922 LVX_V1TI): Add macro expansion.
22923 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22924 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22925 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22926 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22927 Change check to determine if the instruction is a byte reversing
22928 entry. Fix typo in comment.
22929 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22930 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22931 Add def_builtin calls for new builtins.
22932 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22933 Add define_insn expansion.
22935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22937 * config/rl78/rl78.md (umaxdi3): New define_expand.
22939 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22941 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22942 for non-QImode registers.
22944 2018-01-22 Richard Biener <rguenther@suse.de>
22946 PR tree-optimization/83963
22947 * graphite-scop-detection.c (scop_detection::get_sese): Delay
22948 including the loop exit block.
22949 (scop_detection::merge_sese): Likewise.
22950 (scop_detection::add_scop): Do it here instead.
22952 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22954 * doc/sourcebuild.texi (arm_softfloat): Document.
22956 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
22959 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22960 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22961 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22963 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22964 David Edelsohn <dje.gcc@gmail.com>
22967 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22968 Change "crset eq" to "crset 2".
22969 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22970 (*call_indirect_aix<mode>_nospec): Likewise.
22971 (*call_value_indirect_aix<mode>_nospec): Likewise.
22972 (*call_indirect_elfv2<mode>_nospec): Likewise.
22973 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22974 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22975 change assembly output from . to $.
22976 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22977 (indirect_jump<mode>_nospec): Change assembly output from . to $.
22978 (*tablejump<mode>_internal1_nospec): Likewise.
22980 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
22983 * config/sh/sh_optimize_sett_clrt.cc:
22984 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22986 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
22988 PR tree-optimization/83940
22989 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22990 offset_dt to vect_constant_def rather than vect_unknown_def_type.
22991 (vect_check_load_store_mask): Add a mask_dt_out parameter and
22992 use it to pass back the definition type.
22993 (vect_check_store_rhs): Likewise rhs_dt_out.
22994 (vect_build_gather_load_calls): Add a mask_dt argument and use
22995 it instead of a call to vect_is_simple_use.
22996 (vectorizable_store): Update calls to vect_check_load_store_mask
22997 and vect_check_store_rhs. Use the dt returned by the latter instead
22998 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
22999 instead of calls to vect_is_simple_use. Pass the scalar rather
23000 than the vector operand to vect_is_simple_use when handling
23001 second and subsequent copies of an rhs value.
23002 (vectorizable_load): Update calls to vect_check_load_store_mask
23003 and vect_build_gather_load_calls. Use the cached mask_dt and
23004 gs_info.offset_dt instead of calls to vect_is_simple_use.
23006 2018-01-20 Jakub Jelinek <jakub@redhat.com>
23008 PR middle-end/83945
23009 * tree-emutls.c: Include gimplify.h.
23010 (lower_emutls_2): New function.
23011 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23012 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23013 it before further processing.
23016 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23017 UINTVAL (trueop1) instead of INTVAL (op1).
23019 2018-01-19 Jakub Jelinek <jakub@redhat.com>
23023 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23024 INCOMING_FRAME_SP_OFFSET if not defined.
23025 (scan_trace): Add ENTRY argument. If true and
23026 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23027 emit a note to adjust the CFA offset.
23028 (create_cfi_notes): Adjust scan_trace callers.
23029 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23030 INCOMING_FRAME_SP_OFFSET in the CIE.
23031 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23032 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23034 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23035 * doc/tm.texi: Regenerated.
23037 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23039 PR rtl-optimization/83147
23040 * lra-constraints.c (remove_inheritance_pseudos): Use
23041 lra_substitute_pseudo_within_insn.
23043 2018-01-19 Tom de Vries <tom@codesourcery.com>
23044 Cesar Philippidis <cesar@codesourcery.com>
23047 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23049 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
23052 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23053 spaces for function labels.
23055 2018-01-19 Martin Liska <mliska@suse.cz>
23057 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23058 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23059 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23060 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23061 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23062 (PRED_CONST_RETURN): Change from 69 to 65.
23063 (PRED_NULL_RETURN): Change from 91 to 71.
23064 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23065 (PRED_LOOP_GUARD): Change from 66 to 73.
23067 2018-01-19 Martin Liska <mliska@suse.cz>
23069 * predict.c (predict_insn_def): Add new assert.
23070 (struct branch_predictor): Change type to signed integer.
23071 (test_prediction_value_range): Amend test to cover
23072 PROB_UNINITIALIZED.
23073 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23074 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23075 (PRED_LOOP_ITERATIONS_MAX): Likewise.
23076 (PRED_LOOP_IV_COMPARE): Likewise.
23077 * predict.h (PROB_UNINITIALIZED): Define new constant.
23079 2018-01-19 Martin Liska <mliska@suse.cz>
23081 * predict.c (dump_prediction): Add new format for
23082 analyze_brprob.py script which is enabled with -details
23084 * profile-count.h (precise_p): New function.
23086 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23088 PR tree-optimization/83922
23089 * tree-vect-loop.c (vect_verify_full_masking): Return false if
23090 there are no statements that need masking.
23091 (vect_active_double_reduction_p): New function.
23092 (vect_analyze_loop_operations): Use it when handling phis that
23093 are not in the loop header.
23095 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23097 PR tree-optimization/83914
23098 * tree-vect-loop.c (vectorizable_induction): Don't convert
23099 init_expr or apply the peeling adjustment for inductions
23100 that are nested within the vectorized loop.
23102 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23104 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23107 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23111 * function.h (gimplify_parameters): Add gimple_seq * argument.
23112 * function.c: Include gimple.h and options.h.
23113 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23114 for the added local temporaries if needed.
23115 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23116 if there are any parameter cleanups, wrap whole body into a
23117 try/finally with the cleanups.
23119 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
23122 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23123 Use GET_MODE_CLASS for scalar floating point.
23125 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23129 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23130 Fix call of call_cgraph_insertion_hooks.
23132 2018-01-18 Martin Sebor <msebor@redhat.com>
23134 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23136 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23139 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23142 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
23145 * common.opt: (-ffile-prefix-map): New option.
23146 * opts.c (common_handle_option): Defer it.
23147 * opts-global.c (handle_common_deferred_options): Handle it.
23148 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23149 * file-prefix-map.h: New file.
23150 (remap_debug_filename, add_debug_prefix_map): ...here.
23151 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23152 * final.c (debug_prefix_map, add_debug_prefix_map
23153 remap_debug_filename): Move to...
23154 * file-prefix-map.c: New file.
23155 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23156 generalize, get rid of alloca(), use strrchr() instead of strchr().
23157 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23158 Implement in terms of add_prefix_map().
23159 (remap_macro_filename, remap_debug_filename): Implement in term of
23161 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23162 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23163 * dbxout.c: Include file-prefix-map.h.
23164 * varasm.c: Likewise.
23165 * vmsdbgout.c: Likewise.
23166 * xcoffout.c: Likewise.
23167 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23168 * doc/cppopts.texi (-fmacro-prefix-map): Document.
23169 * doc/invoke.texi (-ffile-prefix-map): Document.
23170 (-fdebug-prefix-map): Update description.
23172 2018-01-18 Martin Liska <mliska@suse.cz>
23174 * config/i386/i386.c (indirect_thunk_name): Document that also
23176 (output_indirect_thunk): Document why both instructions
23177 (pause and lfence) are generated.
23179 2018-01-18 Richard Biener <rguenther@suse.de>
23181 PR tree-optimization/83887
23182 * graphite-scop-detection.c
23183 (scop_detection::get_nearest_dom_with_single_entry): Remove.
23184 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23185 (scop_detection::merge_sese): Re-implement with a flood-fill
23186 algorithm that properly finds a SESE region if it exists.
23188 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23191 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23192 pointer_diff optimizations use view_convert instead of convert.
23194 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23196 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23197 Generate different code for -mno-speculate-indirect-jumps.
23198 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23199 (*call_indirect_aix<mode>): Disable for
23200 -mno-speculate-indirect-jumps.
23201 (*call_indirect_aix<mode>_nospec): New define_insn.
23202 (*call_value_indirect_aix<mode>): Disable for
23203 -mno-speculate-indirect-jumps.
23204 (*call_value_indirect_aix<mode>_nospec): New define_insn.
23205 (*sibcall_nonlocal_sysv<mode>): Generate different code for
23206 -mno-speculate-indirect-jumps.
23207 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23209 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
23211 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23212 long double type, set the flags for noting the default long double
23213 type, even if we don't pass or return a long double type.
23215 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
23218 * ipa-inline.c (flatten_function): Do not overwrite final inlining
23221 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
23223 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23224 support for merge[hl].
23225 (fold_mergehl_helper): New helper function.
23226 (tree-vector-builder.h): New #include for tree_vector_builder usage.
23227 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23228 (altivec_vmrglw_direct): Add xxmrglw insn.
23230 2018-01-17 Andrew Waterman <andrew@sifive.com>
23232 * config/riscv/riscv.c (riscv_conditional_register_usage): If
23233 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23235 2018-01-17 David Malcolm <dmalcolm@redhat.com>
23238 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23239 call the lto_location_cache before reading the
23240 DECL_SOURCE_LOCATION of the types.
23242 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
23243 Richard Sandiford <richard.sandiford@linaro.org>
23245 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23246 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23247 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23248 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23249 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23251 * config/aarch64/constraints.md (aarch64_movti_operand):
23253 * config/aarch64/predicates.md (Uti): Add new constraint.
23255 2018-01-17 Carl Love <cel@us.ibm.com>
23257 * config/rs6000/vsx.md (define_expand xl_len_r,
23258 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23259 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23261 (define_expand, define_insn): Move the shift left from the
23262 define_insn to the define_expand for lxvl and stxvl instructions.
23263 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23264 and XL_LEN_R definitions to PURE.
23266 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
23268 * config/i386/i386.c (indirect_thunk_name): Declare regno
23269 as unsigned int. Compare regno with INVALID_REGNUM.
23270 (output_indirect_thunk): Ditto.
23271 (output_indirect_thunk_function): Ditto.
23272 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
23273 in the call to output_indirect_thunk_function.
23275 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
23277 PR middle-end/83884
23278 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23279 rather than the size of inner_type to determine the stack slot size
23280 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23282 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
23285 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23288 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
23290 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23291 endian Linux systems to optionally enable multilibs for selecting
23292 the long double type if the user configured an explicit type.
23293 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23294 have no long double multilibs if not defined.
23295 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23296 warn if the user used -mabi={ieee,ibm}longdouble and we built
23297 multilibs for long double.
23298 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23299 appropriate multilib option.
23300 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23302 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23303 for building long double multilibs.
23304 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23306 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
23308 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23311 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23313 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23316 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23319 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23322 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23324 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23325 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23327 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
23329 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23330 different rtl trees depending on TARGET_64BIT.
23331 (rs6000_gen_lvx): Likewise.
23333 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23335 * config/visium/visium.md (nop): Tweak comment.
23336 (hazard_nop): Likewise.
23338 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23340 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23341 -mspeculate-indirect-jumps.
23342 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23343 for -mno-speculate-indirect-jumps.
23344 (*call_indirect_elfv2<mode>_nospec): New define_insn.
23345 (*call_value_indirect_elfv2<mode>): Disable for
23346 -mno-speculate-indirect-jumps.
23347 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23348 (indirect_jump): Emit different RTL for
23349 -mno-speculate-indirect-jumps.
23350 (*indirect_jump<mode>): Disable for
23351 -mno-speculate-indirect-jumps.
23352 (*indirect_jump<mode>_nospec): New define_insn.
23353 (tablejump): Emit different RTL for
23354 -mno-speculate-indirect-jumps.
23355 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23356 (tablejumpsi_nospec): New define_expand.
23357 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23358 (tablejumpdi_nospec): New define_expand.
23359 (*tablejump<mode>_internal1): Disable for
23360 -mno-speculate-indirect-jumps.
23361 (*tablejump<mode>_internal1_nospec): New define_insn.
23362 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23365 2018-01-16 Artyom Skrobov tyomitch@gmail.com
23367 * caller-save.c (insert_save): Drop unnecessary parameter. All
23370 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23371 Richard Biener <rguenth@suse.de>
23374 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23375 return early, inline manually is_gimple_sizepos. Make sure if we
23376 call gimplify_expr we don't end up with a gimple constant.
23377 * tree.c (variably_modified_type_p): Don't return true for
23378 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
23379 * gimplify.h (is_gimple_sizepos): Remove.
23381 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23383 PR tree-optimization/83857
23384 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23385 vectorizable_live_operation for pure SLP statements.
23386 (vectorizable_live_operation): Handle PHIs.
23388 2018-01-16 Richard Biener <rguenther@suse.de>
23390 PR tree-optimization/83867
23391 * tree-vect-stmts.c (vect_transform_stmt): Precompute
23392 nested_in_vect_loop_p since the scalar stmt may get invalidated.
23394 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23397 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23398 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23399 If off is not INTEGER_CST, issue a may not be aligned warning
23400 rather than isn't aligned. Use isn%'t rather than isn't.
23401 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23403 <case MULT_EXPR>: Improve the case when bottom and one of the
23404 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23405 operand, in that case check if the other operand is multiple of
23406 bottom divided by the INTEGER_CST operand.
23408 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23411 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23412 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23413 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23414 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23415 * config/pa/pa.c (pa_function_arg_advance): Likewise.
23416 (pa_function_arg, pa_arg_partial_bytes): Likewise.
23417 (pa_function_arg_size): New function.
23419 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23421 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23422 in a separate statement.
23424 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23426 PR tree-optimization/83847
23427 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23428 group gathers and scatters.
23430 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23432 PR rtl-optimization/86620
23433 * params.def (max-sched-ready-insns): Bump minimum value to 1.
23435 PR rtl-optimization/83213
23436 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23437 to last if both are JUMP_INSNs.
23439 PR tree-optimization/83843
23440 * gimple-ssa-store-merging.c
23441 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23442 store_immediate_info for bswap/nop orig_stores.
23444 2018-01-15 Andrew Waterman <andrew@sifive.com>
23446 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23448 <UDIV>: Increase cost if !TARGET_DIV.
23450 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
23452 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23453 (define_attr "cr_logical_3op"): New.
23454 (cceq_ior_compare): Adjust.
23455 (cceq_ior_compare_complement): Adjust.
23456 (*cceq_rev_compare): Adjust.
23457 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23458 (is_cracked_insn): Adjust.
23459 (insn_must_be_first_in_group): Adjust.
23460 * config/rs6000/40x.md: Adjust.
23461 * config/rs6000/440.md: Adjust.
23462 * config/rs6000/476.md: Adjust.
23463 * config/rs6000/601.md: Adjust.
23464 * config/rs6000/603.md: Adjust.
23465 * config/rs6000/6xx.md: Adjust.
23466 * config/rs6000/7450.md: Adjust.
23467 * config/rs6000/7xx.md: Adjust.
23468 * config/rs6000/8540.md: Adjust.
23469 * config/rs6000/cell.md: Adjust.
23470 * config/rs6000/e300c2c3.md: Adjust.
23471 * config/rs6000/e500mc.md: Adjust.
23472 * config/rs6000/e500mc64.md: Adjust.
23473 * config/rs6000/e5500.md: Adjust.
23474 * config/rs6000/e6500.md: Adjust.
23475 * config/rs6000/mpc.md: Adjust.
23476 * config/rs6000/power4.md: Adjust.
23477 * config/rs6000/power5.md: Adjust.
23478 * config/rs6000/power6.md: Adjust.
23479 * config/rs6000/power7.md: Adjust.
23480 * config/rs6000/power8.md: Adjust.
23481 * config/rs6000/power9.md: Adjust.
23482 * config/rs6000/rs64.md: Adjust.
23483 * config/rs6000/titan.md: Adjust.
23485 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23487 * config/i386/predicates.md (indirect_branch_operand): Rewrite
23488 ix86_indirect_branch_register logic.
23490 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23492 * config/i386/constraints.md (Bs): Update
23493 ix86_indirect_branch_register check. Don't check
23494 ix86_indirect_branch_register with GOT_memory_operand.
23496 * config/i386/predicates.md (GOT_memory_operand): Don't check
23497 ix86_indirect_branch_register here.
23498 (GOT32_symbol_operand): Likewise.
23500 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23502 * config/i386/predicates.md (constant_call_address_operand):
23503 Rewrite ix86_indirect_branch_register logic.
23504 (sibcall_insn_operand): Likewise.
23506 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23508 * config/i386/constraints.md (Bs): Replace
23509 ix86_indirect_branch_thunk_register with
23510 ix86_indirect_branch_register.
23512 * config/i386/i386.md (indirect_jump): Likewise.
23513 (tablejump): Likewise.
23514 (*sibcall_memory): Likewise.
23515 (*sibcall_value_memory): Likewise.
23516 Peepholes of indirect call and jump via memory: Likewise.
23517 * config/i386/i386.opt: Likewise.
23518 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23519 (GOT_memory_operand): Likewise.
23520 (call_insn_operand): Likewise.
23521 (sibcall_insn_operand): Likewise.
23522 (GOT32_symbol_operand): Likewise.
23524 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23526 PR middle-end/83837
23527 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23528 type rather than type addr's type points to.
23529 (expand_omp_atomic_mutex): Likewise.
23530 (expand_omp_atomic): Likewise.
23532 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23535 * config/i386/i386.c (output_indirect_thunk_function): Use
23536 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23537 for __x86_return_thunk.
23539 2018-01-15 Richard Biener <rguenther@suse.de>
23541 PR middle-end/83850
23542 * expmed.c (extract_bit_field_1): Fix typo.
23544 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23547 * config/arm/iterators.md (VF): New mode iterator.
23548 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23549 Remove integer-related logic from pattern.
23550 (neon_vabd<mode>_3): Likewise.
23552 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23554 PR middle-end/82694
23555 * common.opt (fstrict-overflow): No longer an alias.
23556 (fwrapv-pointer): New option.
23557 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23558 also for pointer types based on flag_wrapv_pointer.
23559 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23560 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23561 opts->x_flag_wrapv got set.
23562 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23563 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23564 POINTER_TYPE_OVERFLOW_UNDEFINED.
23565 * match.pd: Likewise in address comparison pattern.
23566 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23568 2018-01-15 Richard Biener <rguenther@suse.de>
23571 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23572 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
23573 Reset type names to their identifier if their TYPE_DECL doesn't
23574 have linkage (and thus is used for ODR and devirt).
23575 (save_debug_info_for_decl): Remove.
23576 (save_debug_info_for_type): Likewise.
23577 (add_tree_to_fld_list): Adjust.
23578 * tree-pretty-print.c (dump_generic_node): Make dumping of
23579 type names more robust.
23581 2018-01-15 Richard Biener <rguenther@suse.de>
23583 * BASE-VER: Bump to 8.0.1.
23585 2018-01-14 Martin Sebor <msebor@redhat.com>
23588 * builtins.c (check_access): Avoid warning when the no-warning bit
23591 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
23593 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23594 * ira-color (allocno_hard_regs_compare): Likewise.
23596 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
23599 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23600 Use .pushsection/.popsection.
23602 2018-01-14 Martin Sebor <msebor@redhat.com>
23605 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23607 2018-01-14 Jakub Jelinek <jakub@redhat.com>
23609 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23610 entry from extra_headers.
23611 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23612 extra_headers, make the list bitwise identical to the i?86-*-* one.
23614 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23616 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23617 -mcmodel=large with -mindirect-branch=thunk,
23618 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23619 -mfunction-return=thunk-extern.
23620 * doc/invoke.texi: Document -mcmodel=large is incompatible with
23621 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23622 -mfunction-return=thunk and -mfunction-return=thunk-extern.
23624 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23626 * config/i386/i386.c (print_reg): Print the name of the full
23627 integer register without '%'.
23628 (ix86_print_operand): Handle 'V'.
23629 * doc/extend.texi: Document 'V' modifier.
23631 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23633 * config/i386/constraints.md (Bs): Disallow memory operand for
23634 -mindirect-branch-register.
23636 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23637 (GOT_memory_operand): Likewise.
23638 (call_insn_operand): Likewise.
23639 (sibcall_insn_operand): Likewise.
23640 (GOT32_symbol_operand): Likewise.
23641 * config/i386/i386.md (indirect_jump): Call convert_memory_address
23642 for -mindirect-branch-register.
23643 (tablejump): Likewise.
23644 (*sibcall_memory): Likewise.
23645 (*sibcall_value_memory): Likewise.
23646 Disallow peepholes of indirect call and jump via memory for
23647 -mindirect-branch-register.
23648 (*call_pop): Replace m with Bw.
23649 (*call_value_pop): Likewise.
23650 (*sibcall_pop_memory): Replace m with Bs.
23651 * config/i386/i386.opt (mindirect-branch-register): New option.
23652 * doc/invoke.texi: Document -mindirect-branch-register option.
23654 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23656 * config/i386/i386-protos.h (ix86_output_function_return): New.
23657 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23658 set function_return_type.
23659 (indirect_thunk_name): Add ret_p to indicate thunk for function
23661 (output_indirect_thunk_function): Pass false to
23662 indirect_thunk_name.
23663 (ix86_output_indirect_branch_via_reg): Likewise.
23664 (ix86_output_indirect_branch_via_push): Likewise.
23665 (output_indirect_thunk_function): Create alias for function
23666 return thunk if regno < 0.
23667 (ix86_output_function_return): New function.
23668 (ix86_handle_fndecl_attribute): Handle function_return.
23669 (ix86_attribute_table): Add function_return.
23670 * config/i386/i386.h (machine_function): Add
23671 function_return_type.
23672 * config/i386/i386.md (simple_return_internal): Use
23673 ix86_output_function_return.
23674 (simple_return_internal_long): Likewise.
23675 * config/i386/i386.opt (mfunction-return=): New option.
23676 (indirect_branch): Mention -mfunction-return=.
23677 * doc/extend.texi: Document function_return function attribute.
23678 * doc/invoke.texi: Document -mfunction-return= option.
23680 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23682 * config/i386/i386-opts.h (indirect_branch): New.
23683 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23684 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23685 with local indirect jump when converting indirect call and jump.
23686 (ix86_set_indirect_branch_type): New.
23687 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23688 (indirectlabelno): New.
23689 (indirect_thunk_needed): Likewise.
23690 (indirect_thunk_bnd_needed): Likewise.
23691 (indirect_thunks_used): Likewise.
23692 (indirect_thunks_bnd_used): Likewise.
23693 (INDIRECT_LABEL): Likewise.
23694 (indirect_thunk_name): Likewise.
23695 (output_indirect_thunk): Likewise.
23696 (output_indirect_thunk_function): Likewise.
23697 (ix86_output_indirect_branch_via_reg): Likewise.
23698 (ix86_output_indirect_branch_via_push): Likewise.
23699 (ix86_output_indirect_branch): Likewise.
23700 (ix86_output_indirect_jmp): Likewise.
23701 (ix86_code_end): Call output_indirect_thunk_function if needed.
23702 (ix86_output_call_insn): Call ix86_output_indirect_branch if
23704 (ix86_handle_fndecl_attribute): Handle indirect_branch.
23705 (ix86_attribute_table): Add indirect_branch.
23706 * config/i386/i386.h (machine_function): Add indirect_branch_type
23707 and has_local_indirect_jump.
23708 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23710 (tablejump): Likewise.
23711 (*indirect_jump): Use ix86_output_indirect_jmp.
23712 (*tablejump_1): Likewise.
23713 (simple_return_indirect_internal): Likewise.
23714 * config/i386/i386.opt (mindirect-branch=): New option.
23715 (indirect_branch): New.
23718 (thunk-inline): Likewise.
23719 (thunk-extern): Likewise.
23720 * doc/extend.texi: Document indirect_branch function attribute.
23721 * doc/invoke.texi: Document -mindirect-branch= option.
23723 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
23726 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23728 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
23730 * ipa-inline.c (want_inline_small_function_p): Return false if
23731 inlining has already failed with CIF_FINAL_ERROR.
23732 (update_caller_keys): Call want_inline_small_function_p before
23734 (update_callee_keys): Likewise.
23736 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23738 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23740 (rs6000_quadword_masked_address_p): Likewise.
23741 (quad_aligned_load_p): Likewise.
23742 (quad_aligned_store_p): Likewise.
23743 (const_load_sequence_p): Add comment to describe the outer-most loop.
23744 (mimic_memory_attributes_and_flags): New function.
23745 (rs6000_gen_stvx): Likewise.
23746 (replace_swapped_aligned_store): Likewise.
23747 (rs6000_gen_lvx): Likewise.
23748 (replace_swapped_aligned_load): Likewise.
23749 (replace_swapped_load_constant): Capitalize argument name in
23750 comment describing this function.
23751 (rs6000_analyze_swaps): Add a third pass to search for vector loads
23752 and stores that access quad-word aligned addresses and replace
23753 with stvx or lvx instructions when appropriate.
23754 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23755 New function prototype.
23756 (rs6000_quadword_masked_address_p): Likewise.
23757 (rs6000_gen_lvx): Likewise.
23758 (rs6000_gen_stvx): Likewise.
23759 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23760 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23761 when memory address is aligned.
23762 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23763 this split to select lvx instruction when memory address is aligned.
23764 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23765 instruction when memory address is aligned.
23766 (*vsx_le_perm_load_v16qi): Likewise.
23767 (four unnamed splitters): Modify to select the stvx instruction
23768 when memory is aligned.
23770 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23772 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23773 which appears in the queue multiple times.
23775 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23776 Alan Hayward <alan.hayward@arm.com>
23777 David Sherwood <david.sherwood@arm.com>
23779 * tree-vectorizer.h (vec_lower_bound): New structure.
23780 (_loop_vec_info): Add check_nonzero and lower_bounds.
23781 (LOOP_VINFO_CHECK_NONZERO): New macro.
23782 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23783 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23784 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23785 fields. Make seg_len the distance travelled, not including the
23787 (dr_direction_indicator): Declare.
23788 (dr_zero_step_indicator): Likewise.
23789 (dr_known_forward_stride_p): Likewise.
23790 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23792 (runtime_alias_check_p): Allow runtime alias checks with
23794 (operator ==): Compare access_size and align.
23795 (prune_runtime_alias_test_list): Rework for new distinction between
23796 the access_size and seg_len.
23797 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23799 (get_segment_min_max): New function.
23800 (create_intersect_range_checks): Use it.
23801 (dr_step_indicator): New function.
23802 (dr_direction_indicator): Likewise.
23803 (dr_zero_step_indicator): Likewise.
23804 (dr_known_forward_stride_p): Likewise.
23805 * tree-loop-distribution.c (data_ref_segment_size): Return
23806 DR_STEP * (niters - 1).
23807 (compute_alias_check_pairs): Update call to the dr_with_seg_len
23809 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23810 (vect_preserves_scalar_order_p): New function, split out from...
23811 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
23812 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23813 (vect_vfa_access_size): New function.
23814 (vect_vfa_align): Likewise.
23815 (vect_compile_time_alias): Take access_size_a and access_b arguments.
23816 (dump_lower_bound): New function.
23817 (vect_check_lower_bound): Likewise.
23818 (vect_small_gap_p): Likewise.
23819 (vectorizable_with_step_bound_p): Likewise.
23820 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23821 depencies if the vectorization factor is 1. Convert the checks
23822 for nonzero steps into checks on the bounds of DR_STEP. Try using
23823 a bunds check for variable steps if the minimum required step is
23824 relatively small. Update calls to the dr_with_seg_len
23825 constructor and to vect_compile_time_alias.
23826 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23828 (vect_loop_versioning): Call it.
23829 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23831 (vect_estimate_min_profitable_iters): Account for any bounds checks.
23833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23834 Alan Hayward <alan.hayward@arm.com>
23835 David Sherwood <david.sherwood@arm.com>
23837 * doc/sourcebuild.texi (vect_scatter_store): Document.
23838 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23840 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23842 * genopinit.c (main): Add supports_vec_scatter_store and
23843 supports_vec_scatter_store_cached to target_optabs.
23844 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23845 IFN_MASK_SCATTER_STORE.
23846 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23848 * internal-fn.h (internal_store_fn_p): Declare.
23849 (internal_fn_stored_value_index): Likewise.
23850 * internal-fn.c (scatter_store_direct): New macro.
23851 (expand_scatter_store_optab_fn): New function.
23852 (direct_scatter_store_optab_supported_p): New macro.
23853 (internal_store_fn_p): New function.
23854 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23855 IFN_MASK_SCATTER_STORE.
23856 (internal_fn_mask_index): Likewise.
23857 (internal_fn_stored_value_index): New function.
23858 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23859 for scatter stores.
23860 * optabs-query.h (supports_vec_scatter_store_p): Declare.
23861 * optabs-query.c (supports_vec_scatter_store_p): New function.
23862 * tree-vectorizer.h (vect_get_store_rhs): Declare.
23863 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23864 true for scatter stores.
23865 (vect_gather_scatter_fn_p): Handle scatter stores too.
23866 (vect_check_gather_scatter): Consider using scatter stores if
23867 supports_vec_scatter_store_p.
23868 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23869 scatter stores too.
23870 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23871 internal_fn_stored_value_index.
23872 (check_load_store_masking): Handle scatter stores too.
23873 (vect_get_store_rhs): Make public.
23874 (vectorizable_call): Use internal_store_fn_p.
23875 (vectorizable_store): Handle scatter store internal functions.
23876 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23877 when deciding whether the end of the group has been reached.
23878 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23879 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23880 (mask_scatter_store<mode>): New insns.
23882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23883 Alan Hayward <alan.hayward@arm.com>
23884 David Sherwood <david.sherwood@arm.com>
23886 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23887 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23888 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23890 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23891 Use vect_truncate_gather_scatter_offset if we can't treat the
23892 operation as a normal gather load or scatter store.
23893 (get_group_load_store_type): Take the gather_scatter_info
23894 as argument. Try using a gather load or scatter store for
23895 single-element groups.
23896 (get_load_store_type): Update calls to get_group_load_store_type
23897 and vect_use_strided_gather_scatters_p.
23899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23900 Alan Hayward <alan.hayward@arm.com>
23901 David Sherwood <david.sherwood@arm.com>
23903 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23904 optional tree argument.
23905 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23907 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23908 but continue to use the current value as a fallback.
23909 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23910 to compare the updates.
23911 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23912 (get_load_store_type): Use it when handling a strided access.
23913 (vect_get_strided_load_store_ops): New function.
23914 (vect_get_data_ptr_increment): Likewise.
23915 (vectorizable_load): Handle strided gather loads. Always pass
23916 a step to vect_create_data_ref_ptr and bump_vector_ptr.
23918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23919 Alan Hayward <alan.hayward@arm.com>
23920 David Sherwood <david.sherwood@arm.com>
23922 * doc/md.texi (gather_load@var{m}): Document.
23923 (mask_gather_load@var{m}): Likewise.
23924 * genopinit.c (main): Add supports_vec_gather_load and
23925 supports_vec_gather_load_cached to target_optabs.
23926 * optabs-tree.c (init_tree_optimization_optabs): Use
23927 ggc_cleared_alloc to allocate target_optabs.
23928 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23929 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23931 * internal-fn.h (internal_load_fn_p): Declare.
23932 (internal_gather_scatter_fn_p): Likewise.
23933 (internal_fn_mask_index): Likewise.
23934 (internal_gather_scatter_fn_supported_p): Likewise.
23935 * internal-fn.c (gather_load_direct): New macro.
23936 (expand_gather_load_optab_fn): New function.
23937 (direct_gather_load_optab_supported_p): New macro.
23938 (direct_internal_fn_optab): New function.
23939 (internal_load_fn_p): Likewise.
23940 (internal_gather_scatter_fn_p): Likewise.
23941 (internal_fn_mask_index): Likewise.
23942 (internal_gather_scatter_fn_supported_p): Likewise.
23943 * optabs-query.c (supports_at_least_one_mode_p): New function.
23944 (supports_vec_gather_load_p): Likewise.
23945 * optabs-query.h (supports_vec_gather_load_p): Declare.
23946 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23947 and memory_type field.
23948 (NUM_PATTERNS): Bump to 15.
23949 * tree-vect-data-refs.c: Include internal-fn.h.
23950 (vect_gather_scatter_fn_p): New function.
23951 (vect_describe_gather_scatter_call): Likewise.
23952 (vect_check_gather_scatter): Try using internal functions for
23953 gather loads. Recognize existing calls to a gather load function.
23954 (vect_analyze_data_refs): Consider using gather loads if
23955 supports_vec_gather_load_p.
23956 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23957 (vect_get_gather_scatter_offset_type): Likewise.
23958 (vect_convert_mask_for_vectype): Likewise.
23959 (vect_add_conversion_to_patterm): Likewise.
23960 (vect_try_gather_scatter_pattern): Likewise.
23961 (vect_recog_gather_scatter_pattern): New pattern recognizer.
23962 (vect_vect_recog_func_ptrs): Add it.
23963 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23964 internal_fn_mask_index and internal_gather_scatter_fn_p.
23965 (check_load_store_masking): Take the gather_scatter_info as an
23966 argument and handle gather loads.
23967 (vect_get_gather_scatter_ops): New function.
23968 (vectorizable_call): Check internal_load_fn_p.
23969 (vectorizable_load): Likewise. Handle gather load internal
23971 (vectorizable_store): Update call to check_load_store_masking.
23972 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23973 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23974 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23975 (aarch64_gather_scale_operand_d): New predicates.
23976 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23977 (mask_gather_load<mode>): New insns.
23979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23980 Alan Hayward <alan.hayward@arm.com>
23981 David Sherwood <david.sherwood@arm.com>
23983 * optabs.def (fold_left_plus_optab): New optab.
23984 * doc/md.texi (fold_left_plus_@var{m}): Document.
23985 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23986 * internal-fn.c (fold_left_direct): Define.
23987 (expand_fold_left_optab_fn): Likewise.
23988 (direct_fold_left_optab_supported_p): Likewise.
23989 * fold-const-call.c (fold_const_fold_left): New function.
23990 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23991 * tree-parloops.c (valid_reduction_p): New function.
23992 (gather_scalar_reductions): Use it.
23993 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23994 (vect_finish_replace_stmt): Declare.
23995 * tree-vect-loop.c (fold_left_reduction_fn): New function.
23996 (needs_fold_left_reduction_p): New function, split out from...
23997 (vect_is_simple_reduction): ...here. Accept reductions that
23998 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23999 (vect_force_simple_reduction): Also store the reduction type in
24000 the assignment's STMT_VINFO_REDUC_TYPE.
24001 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24002 (merge_with_identity): New function.
24003 (vect_expand_fold_left): Likewise.
24004 (vectorize_fold_left_reduction): Likewise.
24005 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
24006 scalar phi in place for it. Check for target support and reject
24007 cases that would reassociate the operation. Defer the transform
24008 phase to vectorize_fold_left_reduction.
24009 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24010 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24011 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24015 * tree-if-conv.c (predicate_mem_writes): Remove redundant
24016 call to ifc_temp_var.
24018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24019 Alan Hayward <alan.hayward@arm.com>
24020 David Sherwood <david.sherwood@arm.com>
24022 * target.def (legitimize_address_displacement): Take the original
24023 offset as a poly_int.
24024 * targhooks.h (default_legitimize_address_displacement): Update
24026 * targhooks.c (default_legitimize_address_displacement): Likewise.
24027 * doc/tm.texi: Regenerate.
24028 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24029 as an argument, moving assert of ad->disp == ad->disp_term to...
24030 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
24031 Try calling targetm.legitimize_address_displacement before expanding
24032 the address rather than afterwards, and adjust for the new interface.
24033 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24034 Match the new hook interface. Handle SVE addresses.
24035 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24036 new hook interface.
24038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24040 * Makefile.in (OBJS): Add early-remat.o.
24041 * target.def (select_early_remat_modes): New hook.
24042 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24043 * doc/tm.texi: Regenerate.
24044 * targhooks.h (default_select_early_remat_modes): Declare.
24045 * targhooks.c (default_select_early_remat_modes): New function.
24046 * timevar.def (TV_EARLY_REMAT): New timevar.
24047 * passes.def (pass_early_remat): New pass.
24048 * tree-pass.h (make_pass_early_remat): Declare.
24049 * early-remat.c: New file.
24050 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24052 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24055 Alan Hayward <alan.hayward@arm.com>
24056 David Sherwood <david.sherwood@arm.com>
24058 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24059 vfm1 with a bound_epilog parameter.
24060 (vect_do_peeling): Update calls accordingly, and move the prologue
24061 call earlier in the function. Treat the base bound_epilog as 0 for
24062 fully-masked loops and retain vf - 1 for other loops. Add 1 to
24063 this base when peeling for gaps.
24064 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24065 with fully-masked loops.
24066 (vect_estimate_min_profitable_iters): Handle the single peeled
24067 iteration in that case.
24069 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24070 Alan Hayward <alan.hayward@arm.com>
24071 David Sherwood <david.sherwood@arm.com>
24073 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24074 single-element interleaving even if the size is not a power of 2.
24075 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24076 accesses for single-element interleaving if the group size is
24079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24080 Alan Hayward <alan.hayward@arm.com>
24081 David Sherwood <david.sherwood@arm.com>
24083 * doc/md.texi (fold_extract_last_@var{m}): Document.
24084 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24085 * optabs.def (fold_extract_last_optab): New optab.
24086 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24087 * internal-fn.c (fold_extract_direct): New macro.
24088 (expand_fold_extract_optab_fn): Likewise.
24089 (direct_fold_extract_optab_supported_p): Likewise.
24090 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24091 * tree-vect-loop.c (vect_model_reduction_cost): Handle
24092 EXTRACT_LAST_REDUCTION.
24093 (get_initial_def_for_reduction): Do not create an initial vector
24094 for EXTRACT_LAST_REDUCTION reductions.
24095 (vectorizable_reduction): Leave the scalar phi in place for
24096 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
24097 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
24098 epilogue code for EXTRACT_LAST_REDUCTION and defer the
24099 transform phase to vectorizable_condition.
24100 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24102 (vect_finish_stmt_generation): ...here.
24103 (vect_finish_replace_stmt): New function.
24104 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24105 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24107 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24110 Alan Hayward <alan.hayward@arm.com>
24111 David Sherwood <david.sherwood@arm.com>
24113 * doc/md.texi (extract_last_@var{m}): Document.
24114 * optabs.def (extract_last_optab): New optab.
24115 * internal-fn.def (EXTRACT_LAST): New internal function.
24116 * internal-fn.c (cond_unary_direct): New macro.
24117 (expand_cond_unary_optab_fn): Likewise.
24118 (direct_cond_unary_optab_supported_p): Likewise.
24119 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24120 loops using EXTRACT_LAST.
24121 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24122 (extract_last_<mode>): ...this optab.
24123 (vec_extract<mode><Vel>): Update accordingly.
24125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24126 Alan Hayward <alan.hayward@arm.com>
24127 David Sherwood <david.sherwood@arm.com>
24129 * target.def (empty_mask_is_expensive): New hook.
24130 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24131 * doc/tm.texi: Regenerate.
24132 * targhooks.h (default_empty_mask_is_expensive): Declare.
24133 * targhooks.c (default_empty_mask_is_expensive): New function.
24134 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24135 if the target says that empty masks are expensive.
24136 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24138 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24141 Alan Hayward <alan.hayward@arm.com>
24142 David Sherwood <david.sherwood@arm.com>
24144 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24145 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24146 (vect_use_loop_mask_for_alignment_p): New function.
24147 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24148 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24149 niters_skip argument. Make sure that the first niters_skip elements
24150 of the first iteration are inactive.
24151 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24152 Update call to vect_set_loop_masks_directly.
24153 (get_misalign_in_elems): New function, split out from...
24154 (vect_gen_prolog_loop_niters): ...here.
24155 (vect_update_init_of_dr): Take a code argument that specifies whether
24156 the adjustment should be added or subtracted.
24157 (vect_update_init_of_drs): Likewise.
24158 (vect_prepare_for_masked_peels): New function.
24159 (vect_do_peeling): Skip prologue peeling if we're using a mask
24160 instead. Update call to vect_update_inits_of_drs.
24161 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24163 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24164 alignment. Do not include the number of peeled iterations in
24165 the minimum threshold in that case.
24166 (vectorizable_induction): Adjust the start value down by
24167 LOOP_VINFO_MASK_SKIP_NITERS iterations.
24168 (vect_transform_loop): Call vect_prepare_for_masked_peels.
24169 Take the number of skipped iterations into account when calculating
24171 * tree-vect-stmts.c (vect_gen_while_not): New function.
24173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24174 Alan Hayward <alan.hayward@arm.com>
24175 David Sherwood <david.sherwood@arm.com>
24177 * doc/sourcebuild.texi (vect_fully_masked): Document.
24178 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24179 default value to 0.
24180 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24182 (vect_analyze_loop_2): ...here. Don't check the vectorization
24183 factor against the number of loop iterations if the loop is
24186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24187 Alan Hayward <alan.hayward@arm.com>
24188 David Sherwood <david.sherwood@arm.com>
24190 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24191 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24192 (dump_groups): Update accordingly.
24193 (iv_use::mem_type): New member variable.
24194 (address_p): New function.
24195 (record_use): Add a mem_type argument and initialize the new
24197 (record_group_use): Add a mem_type argument. Use address_p.
24198 Remove obsolete null checks of base_object. Update call to record_use.
24199 (find_interesting_uses_op): Update call to record_group_use.
24200 (find_interesting_uses_cond): Likewise.
24201 (find_interesting_uses_address): Likewise.
24202 (get_mem_type_for_internal_fn): New function.
24203 (find_address_like_use): Likewise.
24204 (find_interesting_uses_stmt): Try find_address_like_use before
24205 calling find_interesting_uses_op.
24206 (addr_offset_valid_p): Use the iv mem_type field as the type
24207 of the addressed memory.
24208 (add_autoinc_candidates): Likewise.
24209 (get_address_cost): Likewise.
24210 (split_small_address_groups_p): Use address_p.
24211 (split_address_groups): Likewise.
24212 (add_iv_candidate_for_use): Likewise.
24213 (autoinc_possible_for_pair): Likewise.
24214 (rewrite_groups): Likewise.
24215 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24216 (determine_group_iv_cost): Update after split of USE_ADDRESS.
24217 (get_alias_ptr_type_for_ptr_address): New function.
24218 (rewrite_use_address): Rewrite address uses in calls that were
24219 identified by find_address_like_use.
24221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24222 Alan Hayward <alan.hayward@arm.com>
24223 David Sherwood <david.sherwood@arm.com>
24225 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24227 * gimple-expr.h (is_gimple_addressable: Likewise.
24228 * gimple-expr.c (is_gimple_address): Likewise.
24229 * internal-fn.c (expand_call_mem_ref): New function.
24230 (expand_mask_load_optab_fn): Use it.
24231 (expand_mask_store_optab_fn): Likewise.
24233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24234 Alan Hayward <alan.hayward@arm.com>
24235 David Sherwood <david.sherwood@arm.com>
24237 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24238 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24239 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24240 (cond_umax@var{mode}): Document.
24241 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24242 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24243 (cond_umin_optab, cond_umax_optab): New optabs.
24244 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24245 (COND_IOR, COND_XOR): New internal functions.
24246 * internal-fn.h (get_conditional_internal_fn): Declare.
24247 * internal-fn.c (cond_binary_direct): New macro.
24248 (expand_cond_binary_optab_fn): Likewise.
24249 (direct_cond_binary_optab_supported_p): Likewise.
24250 (get_conditional_internal_fn): New function.
24251 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24252 Cope with reduction statements that are vectorized as calls rather
24254 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24255 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24256 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24257 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24258 (UNSPEC_COND_EOR): New unspecs.
24259 (optab): Add mappings for them.
24260 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24261 (sve_int_op, sve_fp_op): New int attributes.
24263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24264 Alan Hayward <alan.hayward@arm.com>
24265 David Sherwood <david.sherwood@arm.com>
24267 * optabs.def (while_ult_optab): New optab.
24268 * doc/md.texi (while_ult@var{m}@var{n}): Document.
24269 * internal-fn.def (WHILE_ULT): New internal function.
24270 * internal-fn.h (direct_internal_fn_supported_p): New override
24271 that takes two types as argument.
24272 * internal-fn.c (while_direct): New macro.
24273 (expand_while_optab_fn): New function.
24274 (convert_optab_supported_p): Likewise.
24275 (direct_while_optab_supported_p): New macro.
24276 * wide-int.h (wi::udiv_ceil): New function.
24277 * tree-vectorizer.h (rgroup_masks): New structure.
24278 (vec_loop_masks): New typedef.
24279 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24280 and fully_masked_p.
24281 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24282 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24283 (vect_max_vf): New function.
24284 (slpeel_make_loop_iterate_ntimes): Delete.
24285 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24286 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24287 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24288 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24289 internal-fn.h, stor-layout.h and optabs-query.h.
24290 (vect_set_loop_mask): New function.
24291 (add_preheader_seq): Likewise.
24292 (add_header_seq): Likewise.
24293 (interleave_supported_p): Likewise.
24294 (vect_maybe_permute_loop_masks): Likewise.
24295 (vect_set_loop_masks_directly): Likewise.
24296 (vect_set_loop_condition_masked): Likewise.
24297 (vect_set_loop_condition_unmasked): New function, split out from
24298 slpeel_make_loop_iterate_ntimes.
24299 (slpeel_make_loop_iterate_ntimes): Rename to..
24300 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
24301 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24302 (vect_do_peeling): Update call accordingly.
24303 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24305 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24306 mask_compare_type, can_fully_mask_p and fully_masked_p.
24307 (release_vec_loop_masks): New function.
24308 (_loop_vec_info): Use it to free the loop masks.
24309 (can_produce_all_loop_masks_p): New function.
24310 (vect_get_max_nscalars_per_iter): Likewise.
24311 (vect_verify_full_masking): Likewise.
24312 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24313 retries, and free the mask rgroups before retrying. Check loop-wide
24314 reasons for disallowing fully-masked loops. Make the final decision
24315 about whether use a fully-masked loop or not.
24316 (vect_estimate_min_profitable_iters): Do not assume that peeling
24317 for the number of iterations will be needed for fully-masked loops.
24318 (vectorizable_reduction): Disable fully-masked loops.
24319 (vectorizable_live_operation): Likewise.
24320 (vect_halve_mask_nunits): New function.
24321 (vect_double_mask_nunits): Likewise.
24322 (vect_record_loop_mask): Likewise.
24323 (vect_get_loop_mask): Likewise.
24324 (vect_transform_loop): Handle the case in which the final loop
24325 iteration might handle a partial vector. Call vect_set_loop_condition
24326 instead of slpeel_make_loop_iterate_ntimes.
24327 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24328 (check_load_store_masking): New function.
24329 (prepare_load_store_mask): Likewise.
24330 (vectorizable_store): Handle fully-masked loops.
24331 (vectorizable_load): Likewise.
24332 (supportable_widening_operation): Use vect_halve_mask_nunits for
24334 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24335 (vect_gen_while): New function.
24336 * config/aarch64/aarch64.md (umax<mode>3): New expander.
24337 (aarch64_uqdec<mode>): New insn.
24339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24340 Alan Hayward <alan.hayward@arm.com>
24341 David Sherwood <david.sherwood@arm.com>
24343 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24344 (reduc_xor_scal_optab): New optabs.
24345 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24346 (reduc_xor_scal_@var{m}): Document.
24347 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24348 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24349 internal functions.
24350 * fold-const-call.c (fold_const_call): Handle them.
24351 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24352 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24353 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24354 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24355 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24356 (UNSPEC_XORV): New unspecs.
24357 (optab): Add entries for them.
24358 (BITWISEV): New int iterator.
24359 (bit_reduc_op): New int attributes.
24361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24362 Alan Hayward <alan.hayward@arm.com>
24363 David Sherwood <david.sherwood@arm.com>
24365 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24366 * internal-fn.def (VEC_SHL_INSERT): New internal function.
24367 * optabs.def (vec_shl_insert_optab): New optab.
24368 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24369 (duplicate_and_interleave): Likewise.
24370 * tree-vect-loop.c: Include internal-fn.h.
24371 (neutral_op_for_slp_reduction): New function, split out from
24372 get_initial_defs_for_reduction.
24373 (get_initial_def_for_reduction): Handle option 2 for variable-length
24374 vectors by loading the neutral value into a vector and then shifting
24375 the initial value into element 0.
24376 (get_initial_defs_for_reduction): Replace the code argument with
24377 the neutral value calculated by neutral_op_for_slp_reduction.
24378 Use gimple_build_vector for constant-length vectors.
24379 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24380 but the first group_size elements have a neutral value.
24381 Use duplicate_and_interleave otherwise.
24382 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24383 Update call to get_initial_defs_for_reduction. Handle SLP
24384 reductions for variable-length vectors by creating one vector
24385 result for each scalar result, with the elements associated
24386 with other scalar results stubbed out with the neutral value.
24387 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24388 Require IFN_VEC_SHL_INSERT for double reductions on
24389 variable-length vectors, or SLP reductions that have
24390 a neutral value. Require can_duplicate_and_interleave_p
24391 support for variable-length unchained SLP reductions if there
24392 is no neutral value, such as for MIN/MAX reductions. Also require
24393 the number of vector elements to be a multiple of the number of
24394 SLP statements when doing variable-length unchained SLP reductions.
24395 Update call to vect_create_epilog_for_reduction.
24396 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24397 and remove initial values.
24398 (duplicate_and_interleave): Make public.
24399 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24400 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24403 Alan Hayward <alan.hayward@arm.com>
24404 David Sherwood <david.sherwood@arm.com>
24406 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24407 (can_duplicate_and_interleave_p): New function.
24408 (vect_get_and_check_slp_defs): Take the vector of statements
24409 rather than just the current one. Remove excess parentheses.
24410 Restriction rejectinon of vect_constant_def and vect_external_def
24411 for variable-length vectors to boolean types, or types for which
24412 can_duplicate_and_interleave_p is false.
24413 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24414 (duplicate_and_interleave): New function.
24415 (vect_get_constant_vectors): Use gimple_build_vector for
24416 constant-length vectors and suitable variable-length constant
24417 vectors. Use duplicate_and_interleave for other variable-length
24418 vectors. Don't defer the update when inserting new statements.
24420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24421 Alan Hayward <alan.hayward@arm.com>
24422 David Sherwood <david.sherwood@arm.com>
24424 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24425 min_profitable_iters doesn't go negative.
24427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24428 Alan Hayward <alan.hayward@arm.com>
24429 David Sherwood <david.sherwood@arm.com>
24431 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24432 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24433 * optabs.def (vec_mask_load_lanes_optab): New optab.
24434 (vec_mask_store_lanes_optab): Likewise.
24435 * internal-fn.def (MASK_LOAD_LANES): New internal function.
24436 (MASK_STORE_LANES): Likewise.
24437 * internal-fn.c (mask_load_lanes_direct): New macro.
24438 (mask_store_lanes_direct): Likewise.
24439 (expand_mask_load_optab_fn): Handle masked operations.
24440 (expand_mask_load_lanes_optab_fn): New macro.
24441 (expand_mask_store_optab_fn): Handle masked operations.
24442 (expand_mask_store_lanes_optab_fn): New macro.
24443 (direct_mask_load_lanes_optab_supported_p): Likewise.
24444 (direct_mask_store_lanes_optab_supported_p): Likewise.
24445 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24447 (vect_load_lanes_supported): Likewise.
24448 * tree-vect-data-refs.c (strip_conversion): New function.
24449 (can_group_stmts_p): Likewise.
24450 (vect_analyze_data_ref_accesses): Use it instead of checking
24451 for a pair of assignments.
24452 (vect_store_lanes_supported): Take a masked_p parameter.
24453 (vect_load_lanes_supported): Likewise.
24454 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24455 vect_store_lanes_supported and vect_load_lanes_supported.
24456 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24457 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24458 parameter. Don't allow gaps for masked accesses.
24459 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
24460 and vect_load_lanes_supported.
24461 (get_load_store_type): Take a masked_p parameter and update
24462 call to get_group_load_store_type.
24463 (vectorizable_store): Update call to get_load_store_type.
24464 Handle IFN_MASK_STORE_LANES.
24465 (vectorizable_load): Update call to get_load_store_type.
24466 Handle IFN_MASK_LOAD_LANES.
24468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24469 Alan Hayward <alan.hayward@arm.com>
24470 David Sherwood <david.sherwood@arm.com>
24472 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24474 * config/aarch64/aarch64-protos.h
24475 (aarch64_sve_struct_memory_operand_p): Declare.
24476 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24477 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24478 (VPRED, vpred): Handle SVE structure modes.
24479 * config/aarch64/constraints.md (Utx): New constraint.
24480 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24481 (aarch64_sve_struct_nonimmediate_operand): New predicates.
24482 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24483 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24484 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24485 structure modes. Split into pieces after RA.
24486 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24487 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24489 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24490 SVE structure modes.
24491 (aarch64_classify_address): Likewise.
24492 (sizetochar): Move earlier in file.
24493 (aarch64_print_operand): Handle SVE register lists.
24494 (aarch64_array_mode): New function.
24495 (aarch64_sve_struct_memory_operand_p): Likewise.
24496 (TARGET_ARRAY_MODE): Redefine.
24498 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24499 Alan Hayward <alan.hayward@arm.com>
24500 David Sherwood <david.sherwood@arm.com>
24502 * target.def (array_mode): New target hook.
24503 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24504 * doc/tm.texi: Regenerate.
24505 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24506 * hooks.c (hook_optmode_mode_uhwi_none): New function.
24507 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24508 targetm.array_mode.
24509 * stor-layout.c (mode_for_array): Likewise. Support polynomial
24512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24513 Alan Hayward <alan.hayward@arm.com>
24514 David Sherwood <david.sherwood@arm.com>
24516 * fold-const.c (fold_binary_loc): Check the argument types
24517 rather than the result type when testing for a vector operation.
24519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24521 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24522 * doc/tm.texi: Regenerate.
24524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24525 Alan Hayward <alan.hayward@arm.com>
24526 David Sherwood <david.sherwood@arm.com>
24528 * doc/invoke.texi (-msve-vector-bits=): Document new option.
24529 (sve): Document new AArch64 extension.
24530 * doc/md.texi (w): Extend the description of the AArch64
24531 constraint to include SVE vectors.
24532 (Upl, Upa): Document new AArch64 predicate constraints.
24533 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24535 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24536 (msve-vector-bits=): New option.
24537 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24538 SVE when these are disabled.
24539 (sve): New extension.
24540 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24541 modes. Adjust their number of units based on aarch64_sve_vg.
24542 (MAX_BITSIZE_MODE_ANY_MODE): Define.
24543 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24544 aarch64_addr_query_type.
24545 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24546 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24547 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24548 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24549 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24550 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24551 (aarch64_simd_imm_zero_p): Delete.
24552 (aarch64_check_zero_based_sve_index_immediate): Declare.
24553 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24554 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24555 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24556 (aarch64_sve_float_mul_immediate_p): Likewise.
24557 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24558 rather than an rtx.
24559 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24560 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24561 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24562 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24563 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24564 (aarch64_regmode_natural_size): Likewise.
24565 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24566 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24568 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24569 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24570 for VG and the SVE predicate registers.
24571 (V_ALIASES): Add a "z"-prefixed alias.
24572 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24573 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24574 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24575 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24576 (REG_CLASS_NAMES): Add entries for them.
24577 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
24578 and the predicate registers.
24579 (aarch64_sve_vg): Declare.
24580 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24581 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24582 (REGMODE_NATURAL_SIZE): Define.
24583 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24585 * config/aarch64/aarch64.c: Include cfgrtl.h.
24586 (simd_immediate_info): Add a constructor for series vectors,
24587 and an associated step field.
24588 (aarch64_sve_vg): New variable.
24589 (aarch64_dbx_register_number): Handle VG and the predicate registers.
24590 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24591 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24592 (VEC_ANY_DATA, VEC_STRUCT): New constants.
24593 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24594 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24595 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24596 (aarch64_get_mask_mode): New functions.
24597 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24598 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24599 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
24600 predicate modes and predicate registers. Explicitly restrict
24601 GPRs to modes of 16 bytes or smaller. Only allow FP registers
24602 to store a vector mode if it is recognized by
24603 aarch64_classify_vector_mode.
24604 (aarch64_regmode_natural_size): New function.
24605 (aarch64_hard_regno_caller_save_mode): Return the original mode
24607 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24608 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24609 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24610 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24612 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
24613 does not overlap dest if the function is frame-related. Handle
24615 (aarch64_split_add_offset): New function.
24616 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24617 them aarch64_add_offset.
24618 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24619 and update call to aarch64_sub_sp.
24620 (aarch64_add_cfa_expression): New function.
24621 (aarch64_expand_prologue): Pass extra temporary registers to the
24622 functions above. Handle the case in which we need to emit new
24623 DW_CFA_expressions for registers that were originally saved
24624 relative to the stack pointer, but now have to be expressed
24625 relative to the frame pointer.
24626 (aarch64_output_mi_thunk): Pass extra temporary registers to the
24628 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
24629 IP0 and IP1 values for SVE frames.
24630 (aarch64_expand_vec_series): New function.
24631 (aarch64_expand_sve_widened_duplicate): Likewise.
24632 (aarch64_expand_sve_const_vector): Likewise.
24633 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24634 Handle SVE constants. Use emit_move_insn to move a force_const_mem
24635 into the register, rather than emitting a SET directly.
24636 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24637 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24638 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24639 (offset_9bit_signed_scaled_p): New functions.
24640 (aarch64_replicate_bitmask_imm): New function.
24641 (aarch64_bitmask_imm): Use it.
24642 (aarch64_cannot_force_const_mem): Reject expressions involving
24643 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
24644 (aarch64_classify_index): Handle SVE indices, by requiring
24645 a plain register index with a scale that matches the element size.
24646 (aarch64_classify_address): Handle SVE addresses. Assert that
24647 the mode of the address is VOIDmode or an integer mode.
24648 Update call to aarch64_classify_symbol.
24649 (aarch64_classify_symbolic_expression): Update call to
24650 aarch64_classify_symbol.
24651 (aarch64_const_vec_all_in_range_p): New function.
24652 (aarch64_print_vector_float_operand): Likewise.
24653 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
24654 "vN" for FP registers with SVE modes. Handle (const ...) vectors
24655 and the FP immediates 1.0 and 0.5.
24656 (aarch64_print_address_internal): Handle SVE addresses.
24657 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24658 (aarch64_regno_regclass): Handle predicate registers.
24659 (aarch64_secondary_reload): Handle big-endian reloads of SVE
24661 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24662 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24663 (aarch64_convert_sve_vector_bits): New function.
24664 (aarch64_override_options): Use it to handle -msve-vector-bits=.
24665 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24666 rather than an rtx.
24667 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24668 Handle SVE vector and predicate modes. Accept VL-based constants
24669 that need only one temporary register, and VL offsets that require
24670 no temporary registers.
24671 (aarch64_conditional_register_usage): Mark the predicate registers
24672 as fixed if SVE isn't available.
24673 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24674 Return true for SVE vector and predicate modes.
24675 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24676 rather than an unsigned int. Handle SVE modes.
24677 (aarch64_preferred_simd_mode): Update call accordingly. Handle
24679 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24681 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24682 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24683 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24684 (aarch64_sve_float_mul_immediate_p): New functions.
24685 (aarch64_sve_valid_immediate): New function.
24686 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24687 Explicitly reject structure modes. Check for INDEX constants.
24688 Handle PTRUE and PFALSE constants.
24689 (aarch64_check_zero_based_sve_index_immediate): New function.
24690 (aarch64_simd_imm_zero_p): Delete.
24691 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24692 vector modes. Accept constants in the range of CNT[BHWD].
24693 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24694 ask for an Advanced SIMD mode.
24695 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24696 (aarch64_simd_vector_alignment): Handle SVE predicates.
24697 (aarch64_vectorize_preferred_vector_alignment): New function.
24698 (aarch64_simd_vector_alignment_reachable): Use it instead of
24700 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24701 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24703 (MAX_VECT_LEN): Delete.
24704 (expand_vec_perm_d): Add a vec_flags field.
24705 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24706 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24707 (aarch64_evpc_ext): Don't apply a big-endian lane correction
24709 (aarch64_evpc_rev): Rename to...
24710 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
24711 (aarch64_evpc_rev_global): New function.
24712 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24713 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24715 (aarch64_evpc_sve_tbl): New function.
24716 (aarch64_expand_vec_perm_const_1): Update after rename of
24717 aarch64_evpc_rev. Handle SVE permutes too, trying
24718 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24719 than aarch64_evpc_tbl.
24720 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24721 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24722 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24723 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24724 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24725 (aarch64_expand_sve_vcond): New functions.
24726 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24727 of aarch64_vector_mode_p.
24728 (aarch64_dwarf_poly_indeterminate_value): New function.
24729 (aarch64_compute_pressure_classes): Likewise.
24730 (aarch64_can_change_mode_class): Likewise.
24731 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24732 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24733 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24734 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24735 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24736 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24737 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24738 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24740 (Dn, Dl, Dr): Accept const as well as const_vector.
24741 (Dz): Likewise. Compare against CONST0_RTX.
24742 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24743 of "vector" where appropriate.
24744 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24745 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24746 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24747 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24748 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24749 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24750 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24751 (v_int_equiv): Extend to SVE modes.
24752 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24754 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24755 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24756 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24757 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24758 (SVE_COND_FP_CMP): New int iterators.
24759 (perm_hilo): Handle the new unpack unspecs.
24760 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24762 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24763 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24764 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24765 (aarch64_equality_operator, aarch64_constant_vector_operand)
24766 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24767 (aarch64_sve_nonimmediate_operand): Likewise.
24768 (aarch64_sve_general_operand): Likewise.
24769 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24770 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24771 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24772 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24773 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24774 (aarch64_sve_float_arith_immediate): Likewise.
24775 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24776 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24777 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24778 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24779 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24780 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24781 (aarch64_sve_float_arith_operand): Likewise.
24782 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24783 (aarch64_sve_float_mul_operand): Likewise.
24784 (aarch64_sve_vec_perm_operand): Likewise.
24785 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24786 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24787 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24788 as well as const_vector.
24789 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24790 in file. Use CONST0_RTX and CONSTM1_RTX.
24791 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24792 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24793 Use aarch64_simd_imm_zero.
24794 * config/aarch64/aarch64-sve.md: New file.
24795 * config/aarch64/aarch64.md: Include it.
24796 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24797 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24798 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24799 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24800 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24801 (sve): New attribute.
24802 (enabled): Disable instructions with the sve attribute unless
24804 (movqi, movhi): Pass CONST_POLY_INT operaneds through
24805 aarch64_expand_mov_immediate.
24806 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24807 CNT[BHSD] immediates.
24808 (movti): Split CONST_POLY_INT moves into two halves.
24809 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24810 Split additions that need a temporary here if the destination
24811 is the stack pointer.
24812 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24813 (*add<mode>3_poly_1): New instruction.
24814 (set_clobber_cc): New expander.
24816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24818 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24819 parameter and use it instead of GET_MODE_SIZE (innermode). Use
24820 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24821 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24822 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
24823 Change innermode from fixed_mode_size to machine_mode.
24824 (simplify_subreg): Update call accordingly. Handle a constant-sized
24825 subreg of a variable-length CONST_VECTOR.
24827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24828 Alan Hayward <alan.hayward@arm.com>
24829 David Sherwood <david.sherwood@arm.com>
24831 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24832 (add_offset_to_base): New function, split out from...
24833 (create_mem_ref): ...here. When handling a scale other than 1,
24834 check first whether the address is valid without the offset.
24835 Add it into the base if so, leaving the index and scale as-is.
24837 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24840 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24841 fold_for_warn before checking if arg2 is INTEGER_CST.
24843 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
24845 * config/rs6000/predicates.md (load_multiple_operation): Delete.
24846 (store_multiple_operation): Delete.
24847 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24848 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24849 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24850 guarded by TARGET_STRING.
24851 (rs6000_output_load_multiple): Delete.
24852 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24853 OPTION_MASK_STRING / TARGET_STRING handling.
24854 (print_operand) <'N', 'O'>: Add comment that these are unused now.
24855 (const rs6000_opt_masks) <"string">: Change mask to 0.
24856 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24857 (MASK_STRING): Delete.
24858 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24860 (load_multiple): Delete.
24867 (store_multiple): Delete.
24874 (movmemsi_8reg): Delete.
24875 (corresponding unnamed define_insn): Delete.
24876 (movmemsi_6reg): Delete.
24877 (corresponding unnamed define_insn): Delete.
24878 (movmemsi_4reg): Delete.
24879 (corresponding unnamed define_insn): Delete.
24880 (movmemsi_2reg): Delete.
24881 (corresponding unnamed define_insn): Delete.
24882 (movmemsi_1reg): Delete.
24883 (corresponding unnamed define_insn): Delete.
24884 * config/rs6000/rs6000.opt (mno-string): New.
24885 (mstring): Replace by deprecation warning stub.
24886 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24888 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24890 * regrename.c (regrename_do_replace): If replacing the same
24891 reg multiple times, try to reuse last created gen_raw_REG.
24894 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24895 main to workaround a bug in GDB.
24897 2018-01-12 Tom de Vries <tom@codesourcery.com>
24900 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24902 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
24904 PR rtl-optimization/80481
24905 * ira-color.c (get_cap_member): New function.
24906 (allocnos_conflict_by_live_ranges_p): Use it.
24907 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24908 (setup_slot_coalesced_allocno_live_ranges): Ditto.
24910 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
24913 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24914 (*saddl_se_1): Ditto.
24915 (*ssubsi_1): Ditto.
24916 (*ssubl_se_1): Ditto.
24918 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24920 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24921 rather than wi::to_widest for DR_INITs.
24922 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24923 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24924 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24926 (vect_analyze_group_access_1): Note that here.
24928 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24930 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24931 polynomial type sizes.
24933 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24935 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24936 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24937 (gimple_add_tmp_var): Likewise.
24939 2018-01-12 Martin Liska <mliska@suse.cz>
24941 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24942 (gimple_alloc_sizes): Likewise.
24943 (dump_gimple_statistics): Use PRIu64 in printf format.
24944 * gimple.h: Change uint64_t to int.
24946 2018-01-12 Martin Liska <mliska@suse.cz>
24948 * tree-core.h: Use uint64_t instead of int.
24949 * tree.c (tree_node_counts): Likewise.
24950 (tree_node_sizes): Likewise.
24951 (dump_tree_statistics): Use PRIu64 in printf format.
24953 2018-01-12 Martin Liska <mliska@suse.cz>
24955 * Makefile.in: As qsort_chk is implemented in vec.c, add
24956 vec.o to linkage of gencfn-macros.
24957 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24958 passing the info to record_node_allocation_statistics.
24959 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24961 * ggc-common.c (struct ggc_usage): Add operator== and use
24962 it in operator< and compare function.
24963 * mem-stats.h (struct mem_usage): Likewise.
24964 * vec.c (struct vec_usage): Remove operator< and compare
24965 function. Can be simply inherited.
24967 2018-01-12 Martin Jambor <mjambor@suse.cz>
24970 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24971 * tree-ssa-math-opts.c: Include domwalk.h.
24972 (convert_mult_to_fma_1): New function.
24973 (fma_transformation_info): New type.
24974 (fma_deferring_state): Likewise.
24975 (cancel_fma_deferring): New function.
24976 (result_of_phi): Likewise.
24977 (last_fma_candidate_feeds_initial_phi): Likewise.
24978 (convert_mult_to_fma): Added deferring logic, split actual
24979 transformation to convert_mult_to_fma_1.
24980 (math_opts_dom_walker): New type.
24981 (math_opts_dom_walker::after_dom_children): New method, body moved
24982 here from pass_optimize_widening_mul::execute, added deferring logic
24984 (pass_optimize_widening_mul::execute): Moved most of code to
24985 math_opts_dom_walker::after_dom_children.
24986 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24987 * config/i386/i386.c (ix86_option_override_internal): Added
24988 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24990 2018-01-12 Richard Biener <rguenther@suse.de>
24993 * dwarf2out.c (gen_variable_die): Do not reset old_die for
24994 inline instance vars.
24996 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
24999 * config/rx/rx.c (rx_is_restricted_memory_address):
25000 Handle SUBREG case.
25002 2018-01-12 Richard Biener <rguenther@suse.de>
25004 PR tree-optimization/80846
25005 * target.def (split_reduction): New target hook.
25006 * targhooks.c (default_split_reduction): New function.
25007 * targhooks.h (default_split_reduction): Declare.
25008 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25009 target requests first reduce vectors by combining low and high
25011 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25012 (get_vectype_for_scalar_type_and_size): Export.
25013 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25014 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25015 * doc/tm.texi: Regenerate.
25016 * config/i386/i386.c (ix86_split_reduction): Implement
25017 TARGET_VECTORIZE_SPLIT_REDUCTION.
25019 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25022 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25023 in PIC mode except for TARGET_VXWORKS_RTP.
25024 * config/sparc/sparc.c: Include cfgrtl.h.
25025 (TARGET_INIT_PIC_REG): Define.
25026 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25027 (sparc_pic_register_p): New predicate.
25028 (sparc_legitimate_address_p): Use it.
25029 (sparc_legitimize_pic_address): Likewise.
25030 (sparc_delegitimize_address): Likewise.
25031 (sparc_mode_dependent_address_p): Likewise.
25032 (gen_load_pcrel_sym): Remove 4th parameter.
25033 (load_got_register): Adjust call to above. Remove obsolete stuff.
25034 (sparc_expand_prologue): Do not call load_got_register here.
25035 (sparc_flat_expand_prologue): Likewise.
25036 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25037 (sparc_use_pseudo_pic_reg): New function.
25038 (sparc_init_pic_reg): Likewise.
25039 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25040 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25042 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25044 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25045 Add item for branch_cost.
25047 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25049 PR rtl-optimization/83565
25050 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25051 not extend the result to a larger mode for rotate operations.
25052 (num_sign_bit_copies1): Likewise.
25054 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25057 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25059 Use values-Xc.o for -pedantic.
25060 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25062 2018-01-12 Martin Liska <mliska@suse.cz>
25065 * ipa-devirt.c (final_warning_record::grow_type_warnings):
25067 (possible_polymorphic_call_targets): Use it.
25068 (ipa_devirt): Likewise.
25070 2018-01-12 Martin Liska <mliska@suse.cz>
25072 * profile-count.h (enum profile_quality): Use 0 as invalid
25073 enum value of profile_quality.
25075 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
25077 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25078 -mext-string options.
25080 2018-01-12 Richard Biener <rguenther@suse.de>
25082 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25083 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25084 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25086 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25088 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
25090 * configure.ac (--with-long-double-format): Add support for the
25091 configuration option to change the default long double format on
25093 * config.gcc (powerpc*-linux*-*): Likewise.
25094 * configure: Regenerate.
25095 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25096 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25097 used without modification.
25099 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25101 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25102 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25103 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25104 MISC_BUILTIN_SPEC_BARRIER.
25105 (rs6000_init_builtins): Likewise.
25106 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25108 (speculation_barrier): New define_insn.
25109 * doc/extend.texi: Document __builtin_speculation_barrier.
25111 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25114 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25115 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25116 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25118 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
25119 integral modes instead of "ss" and "sd".
25120 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25121 vectors with 32-bit and 64-bit elements.
25122 (vecdupssescalarmodesuffix): New mode attribute.
25123 (vec_dup<mode>): Use it.
25125 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
25128 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25129 frame if argument is passed on stack.
25131 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25134 * ree.c (combine_reaching_defs): Optimize also
25135 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25136 reg2=any_extend(exp); reg1=reg2;, formatting fix.
25138 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25140 PR middle-end/83189
25141 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25143 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25145 PR middle-end/83718
25146 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25147 after they are computed.
25149 2018-01-11 Bin Cheng <bin.cheng@arm.com>
25151 PR tree-optimization/83695
25152 * gimple-loop-linterchange.cc
25153 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25154 reset cached scev information after interchange.
25155 (pass_linterchange::execute): Remove call to scev_reset_htab.
25157 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25159 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25160 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25161 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25162 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25163 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25164 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25165 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25166 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25167 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25168 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25169 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25170 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25171 (V_lane_reg): Likewise.
25172 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25174 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25175 (vfmal_lane_low<mode>_intrinsic,
25176 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25177 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25178 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25179 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25180 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25181 vfmsl_lane_high<mode>_intrinsic): New define_insns.
25183 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25185 * config/arm/arm-cpus.in (fp16fml): New feature.
25186 (ALL_SIMD): Add fp16fml.
25187 (armv8.2-a): Add fp16fml as an option.
25188 (armv8.3-a): Likewise.
25189 (armv8.4-a): Add fp16fml as part of fp16.
25190 * config/arm/arm.h (TARGET_FP16FML): Define.
25191 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25193 * config/arm/arm-modes.def (V2HF): Define.
25194 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25195 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25196 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25197 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25198 vfmsl_low, vfmsl_high): New set of builtins.
25199 * config/arm/iterators.md (PLUSMINUS): New code iterator.
25200 (vfml_op): New code attribute.
25201 (VFMLHALVES): New int iterator.
25202 (VFML, VFMLSEL): New mode attributes.
25203 (V_reg): Define mapping for V2HF.
25204 (V_hi, V_lo): New mode attributes.
25205 (VF_constraint): Likewise.
25206 (vfml_half, vfml_half_selector): New int attributes.
25207 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25209 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25210 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25212 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25213 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25214 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25215 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
25217 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25218 Document new effective target and option set.
25220 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25222 * config/arm/arm-cpus.in (armv8_4): New feature.
25223 (ARMv8_4a): New fgroup.
25224 (armv8.4-a): New arch.
25225 * config/arm/arm-tables.opt: Regenerate.
25226 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25227 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25228 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25229 Add matching rules for -march=armv8.4-a and extensions.
25230 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25232 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
25235 * config/rx/rx.md (BW): New mode attribute.
25236 (sync_lock_test_and_setsi): Add mode suffix to insn output.
25238 2018-01-11 Richard Biener <rguenther@suse.de>
25240 PR tree-optimization/83435
25241 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25242 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25243 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25245 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25246 Alan Hayward <alan.hayward@arm.com>
25247 David Sherwood <david.sherwood@arm.com>
25249 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25251 (aarch64_classify_address): Initialize it. Track polynomial offsets.
25252 (aarch64_print_address_internal): Use it to check for a zero offset.
25254 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25255 Alan Hayward <alan.hayward@arm.com>
25256 David Sherwood <david.sherwood@arm.com>
25258 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25259 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25260 Return a poly_int64 rather than a HOST_WIDE_INT.
25261 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25262 rather than a HOST_WIDE_INT.
25263 * config/aarch64/aarch64.h (aarch64_frame): Protect with
25264 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
25265 hard_fp_offset, frame_size, initial_adjust, callee_offset and
25266 final_offset from HOST_WIDE_INT to poly_int64.
25267 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25268 to_constant when getting the number of units in an Advanced SIMD
25270 (aarch64_builtin_vectorized_function): Check for a constant number
25272 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25274 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25275 attribute instead of GET_MODE_NUNITS.
25276 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25277 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25278 GET_MODE_SIZE for fixed-size registers.
25279 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25280 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25281 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25282 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25283 (aarch64_print_operand, aarch64_print_address_internal)
25284 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25285 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25286 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25287 Handle polynomial GET_MODE_SIZE.
25288 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
25289 wider than SImode without modification.
25290 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25291 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25292 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25293 passing and returning SVE modes.
25294 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25295 rather than GEN_INT.
25296 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25297 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25298 (aarch64_allocate_and_probe_stack_space): Likewise.
25299 (aarch64_layout_frame): Cope with polynomial offsets.
25300 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25301 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
25302 polynomial offsets.
25303 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25304 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25305 poly_int64 rather than a HOST_WIDE_INT.
25306 (aarch64_get_separate_components, aarch64_process_components)
25307 (aarch64_expand_prologue, aarch64_expand_epilogue)
25308 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25309 (aarch64_anchor_offset): New function, split out from...
25310 (aarch64_legitimize_address): ...here.
25311 (aarch64_builtin_vectorization_cost): Handle polynomial
25312 TYPE_VECTOR_SUBPARTS.
25313 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25315 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25316 number of elements from the PARALLEL rather than the mode.
25317 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25318 rather than GET_MODE_BITSIZE.
25319 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25320 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25321 (aarch64_expand_vec_perm_const_1): Handle polynomial
25322 d->perm.length () and d->perm elements.
25323 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
25324 Apply to_constant to d->perm elements.
25325 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25326 polynomial CONST_VECTOR_NUNITS.
25327 (aarch64_move_pointer): Take amount as a poly_int64 rather
25329 (aarch64_progress_pointer): Avoid temporary variable.
25330 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25331 the mode attribute instead of GET_MODE.
25333 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25334 Alan Hayward <alan.hayward@arm.com>
25335 David Sherwood <david.sherwood@arm.com>
25337 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25338 x exists before using it.
25339 (aarch64_add_constant_internal): Rename to...
25340 (aarch64_add_offset_1): ...this. Replace regnum with separate
25341 src and dest rtxes. Handle the case in which they're different,
25342 including when the offset is zero. Replace scratchreg with an rtx.
25343 Use 2 additions if there is no spare register into which we can
25344 move a 16-bit constant.
25345 (aarch64_add_constant): Delete.
25346 (aarch64_add_offset): Replace reg with separate src and dest
25347 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
25348 Use aarch64_add_offset_1.
25349 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25350 an rtx rather than an int. Take the delta as a poly_int64
25351 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
25352 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25353 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25354 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25355 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25356 and aarch64_add_sp.
25357 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25358 aarch64_add_constant.
25360 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25362 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25363 Use scalar_float_mode.
25365 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25367 * config/aarch64/aarch64-simd.md
25368 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25369 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25370 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25371 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25372 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25373 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25374 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25375 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25376 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25377 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25379 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25382 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25383 targ_options->x_arm_arch_string is non NULL.
25385 2018-01-11 Tamar Christina <tamar.christina@arm.com>
25387 * config/aarch64/aarch64.h
25388 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
25390 2018-01-11 Sudakshina Das <sudi.das@arm.com>
25393 * expmed.c (emit_store_flag_force): Swap if const op0
25394 and change VOIDmode to mode of op0.
25396 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25398 PR rtl-optimization/83761
25399 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25400 than bytes to mode_for_size.
25402 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25404 PR middle-end/83189
25405 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25406 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25409 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25411 PR middle-end/83575
25412 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25413 when in layout mode.
25414 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25415 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25418 2018-01-10 Michael Collison <michael.collison@arm.com>
25420 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25421 * config/aarch64/aarch64-option-extension.def: Add
25422 AARCH64_OPT_EXTENSION of 'fp16fml'.
25423 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25424 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25425 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25426 * config/aarch64/constraints.md (Ui7): New constraint.
25427 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25428 (VFMLA_SEL_W): Ditto.
25431 (VFMLA16_LOW): New int iterator.
25432 (VFMLA16_HIGH): Ditto.
25433 (UNSPEC_FMLAL): New unspec.
25434 (UNSPEC_FMLSL): Ditto.
25435 (UNSPEC_FMLAL2): Ditto.
25436 (UNSPEC_FMLSL2): Ditto.
25437 (f16mac): New code attribute.
25438 * config/aarch64/aarch64-simd-builtins.def
25439 (aarch64_fmlal_lowv2sf): Ditto.
25440 (aarch64_fmlsl_lowv2sf): Ditto.
25441 (aarch64_fmlalq_lowv4sf): Ditto.
25442 (aarch64_fmlslq_lowv4sf): Ditto.
25443 (aarch64_fmlal_highv2sf): Ditto.
25444 (aarch64_fmlsl_highv2sf): Ditto.
25445 (aarch64_fmlalq_highv4sf): Ditto.
25446 (aarch64_fmlslq_highv4sf): Ditto.
25447 (aarch64_fmlal_lane_lowv2sf): Ditto.
25448 (aarch64_fmlsl_lane_lowv2sf): Ditto.
25449 (aarch64_fmlal_laneq_lowv2sf): Ditto.
25450 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25451 (aarch64_fmlalq_lane_lowv4sf): Ditto.
25452 (aarch64_fmlsl_lane_lowv4sf): Ditto.
25453 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25454 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25455 (aarch64_fmlal_lane_highv2sf): Ditto.
25456 (aarch64_fmlsl_lane_highv2sf): Ditto.
25457 (aarch64_fmlal_laneq_highv2sf): Ditto.
25458 (aarch64_fmlsl_laneq_highv2sf): Ditto.
25459 (aarch64_fmlalq_lane_highv4sf): Ditto.
25460 (aarch64_fmlsl_lane_highv4sf): Ditto.
25461 (aarch64_fmlalq_laneq_highv4sf): Ditto.
25462 (aarch64_fmlsl_laneq_highv4sf): Ditto.
25463 * config/aarch64/aarch64-simd.md:
25464 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25465 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25466 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25467 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25468 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25469 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25470 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25471 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25472 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25473 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25474 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25475 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25476 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25477 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25478 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25479 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25480 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25481 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25482 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25483 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25484 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25485 (vfmlsl_low_u32): Ditto.
25486 (vfmlalq_low_u32): Ditto.
25487 (vfmlslq_low_u32): Ditto.
25488 (vfmlal_high_u32): Ditto.
25489 (vfmlsl_high_u32): Ditto.
25490 (vfmlalq_high_u32): Ditto.
25491 (vfmlslq_high_u32): Ditto.
25492 (vfmlal_lane_low_u32): Ditto.
25493 (vfmlsl_lane_low_u32): Ditto.
25494 (vfmlal_laneq_low_u32): Ditto.
25495 (vfmlsl_laneq_low_u32): Ditto.
25496 (vfmlalq_lane_low_u32): Ditto.
25497 (vfmlslq_lane_low_u32): Ditto.
25498 (vfmlalq_laneq_low_u32): Ditto.
25499 (vfmlslq_laneq_low_u32): Ditto.
25500 (vfmlal_lane_high_u32): Ditto.
25501 (vfmlsl_lane_high_u32): Ditto.
25502 (vfmlal_laneq_high_u32): Ditto.
25503 (vfmlsl_laneq_high_u32): Ditto.
25504 (vfmlalq_lane_high_u32): Ditto.
25505 (vfmlslq_lane_high_u32): Ditto.
25506 (vfmlalq_laneq_high_u32): Ditto.
25507 (vfmlslq_laneq_high_u32): Ditto.
25508 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25509 (AARCH64_FL_FOR_ARCH8_4): New.
25510 (AARCH64_ISA_F16FML): New ISA flag.
25511 (TARGET_F16FML): New feature flag for fp16fml.
25512 (doc/invoke.texi): Document new fp16fml option.
25514 2018-01-10 Michael Collison <michael.collison@arm.com>
25516 * config/aarch64/aarch64-builtins.c:
25517 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25518 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25519 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25520 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25521 (AARCH64_ISA_SHA3): New ISA flag.
25522 (TARGET_SHA3): New feature flag for sha3.
25523 * config/aarch64/iterators.md (sha512_op): New int attribute.
25524 (CRYPTO_SHA512): New int iterator.
25525 (UNSPEC_SHA512H): New unspec.
25526 (UNSPEC_SHA512H2): Ditto.
25527 (UNSPEC_SHA512SU0): Ditto.
25528 (UNSPEC_SHA512SU1): Ditto.
25529 * config/aarch64/aarch64-simd-builtins.def
25530 (aarch64_crypto_sha512hqv2di): New builtin.
25531 (aarch64_crypto_sha512h2qv2di): Ditto.
25532 (aarch64_crypto_sha512su0qv2di): Ditto.
25533 (aarch64_crypto_sha512su1qv2di): Ditto.
25534 (aarch64_eor3qv8hi): Ditto.
25535 (aarch64_rax1qv2di): Ditto.
25536 (aarch64_xarqv2di): Ditto.
25537 (aarch64_bcaxqv8hi): Ditto.
25538 * config/aarch64/aarch64-simd.md:
25539 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25540 (aarch64_crypto_sha512su0qv2di): Ditto.
25541 (aarch64_crypto_sha512su1qv2di): Ditto.
25542 (aarch64_eor3qv8hi): Ditto.
25543 (aarch64_rax1qv2di): Ditto.
25544 (aarch64_xarqv2di): Ditto.
25545 (aarch64_bcaxqv8hi): Ditto.
25546 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25547 (vsha512h2q_u64): Ditto.
25548 (vsha512su0q_u64): Ditto.
25549 (vsha512su1q_u64): Ditto.
25550 (veor3q_u16): Ditto.
25551 (vrax1q_u64): Ditto.
25552 (vxarq_u64): Ditto.
25553 (vbcaxq_u16): Ditto.
25554 * config/arm/types.md (crypto_sha512): New type attribute.
25555 (crypto_sha3): Ditto.
25556 (doc/invoke.texi): Document new sha3 option.
25558 2018-01-10 Michael Collison <michael.collison@arm.com>
25560 * config/aarch64/aarch64-builtins.c:
25561 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25562 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25563 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25564 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25565 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25566 (AARCH64_ISA_SM4): New ISA flag.
25567 (TARGET_SM4): New feature flag for sm4.
25568 * config/aarch64/aarch64-simd-builtins.def
25569 (aarch64_sm3ss1qv4si): Ditto.
25570 (aarch64_sm3tt1aq4si): Ditto.
25571 (aarch64_sm3tt1bq4si): Ditto.
25572 (aarch64_sm3tt2aq4si): Ditto.
25573 (aarch64_sm3tt2bq4si): Ditto.
25574 (aarch64_sm3partw1qv4si): Ditto.
25575 (aarch64_sm3partw2qv4si): Ditto.
25576 (aarch64_sm4eqv4si): Ditto.
25577 (aarch64_sm4ekeyqv4si): Ditto.
25578 * config/aarch64/aarch64-simd.md:
25579 (aarch64_sm3ss1qv4si): Ditto.
25580 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25581 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25582 (aarch64_sm4eqv4si): Ditto.
25583 (aarch64_sm4ekeyqv4si): Ditto.
25584 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25585 (sm3part_op): Ditto.
25586 (CRYPTO_SM3TT): Ditto.
25587 (CRYPTO_SM3PART): Ditto.
25588 (UNSPEC_SM3SS1): New unspec.
25589 (UNSPEC_SM3TT1A): Ditto.
25590 (UNSPEC_SM3TT1B): Ditto.
25591 (UNSPEC_SM3TT2A): Ditto.
25592 (UNSPEC_SM3TT2B): Ditto.
25593 (UNSPEC_SM3PARTW1): Ditto.
25594 (UNSPEC_SM3PARTW2): Ditto.
25595 (UNSPEC_SM4E): Ditto.
25596 (UNSPEC_SM4EKEY): Ditto.
25597 * config/aarch64/constraints.md (Ui2): New constraint.
25598 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25599 * config/arm/types.md (crypto_sm3): New type attribute.
25600 (crypto_sm4): Ditto.
25601 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25602 (vsm3tt1aq_u32): Ditto.
25603 (vsm3tt1bq_u32): Ditto.
25604 (vsm3tt2aq_u32): Ditto.
25605 (vsm3tt2bq_u32): Ditto.
25606 (vsm3partw1q_u32): Ditto.
25607 (vsm3partw2q_u32): Ditto.
25608 (vsm4eq_u32): Ditto.
25609 (vsm4ekeyq_u32): Ditto.
25610 (doc/invoke.texi): Document new sm4 option.
25612 2018-01-10 Michael Collison <michael.collison@arm.com>
25614 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25615 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25616 (AARCH64_FL_FOR_ARCH8_4): New.
25617 (AARCH64_FL_V8_4): New flag.
25618 (doc/invoke.texi): Document new armv8.4-a option.
25620 2018-01-10 Michael Collison <michael.collison@arm.com>
25622 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25623 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25624 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25625 * config/aarch64/aarch64-option-extension.def: Add
25626 AARCH64_OPT_EXTENSION of 'sha2'.
25627 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25628 (crypto): Disable sha2 and aes if crypto disabled.
25629 (crypto): Enable aes and sha2 if enabled.
25630 (simd): Disable sha2 and aes if simd disabled.
25631 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25633 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25634 (TARGET_SHA2): New feature flag for sha2.
25635 (TARGET_AES): New feature flag for aes.
25636 * config/aarch64/aarch64-simd.md:
25637 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25638 conditional on TARGET_AES.
25639 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25640 (aarch64_crypto_sha1hsi): Make pattern conditional
25642 (aarch64_crypto_sha1hv4si): Ditto.
25643 (aarch64_be_crypto_sha1hv4si): Ditto.
25644 (aarch64_crypto_sha1su1v4si): Ditto.
25645 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25646 (aarch64_crypto_sha1su0v4si): Ditto.
25647 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25648 (aarch64_crypto_sha256su0v4si): Ditto.
25649 (aarch64_crypto_sha256su1v4si): Ditto.
25650 (doc/invoke.texi): Document new aes and sha2 options.
25652 2018-01-10 Martin Sebor <msebor@redhat.com>
25654 PR tree-optimization/83781
25655 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25658 2018-01-11 Martin Sebor <msebor@gmail.com>
25659 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25661 PR tree-optimization/83501
25662 PR tree-optimization/81703
25664 * tree-ssa-strlen.c (get_string_cst): Rename...
25665 (get_string_len): ...to this. Handle global constants.
25666 (handle_char_store): Adjust.
25668 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
25669 Jim Wilson <jimw@sifive.com>
25671 * config/riscv/riscv-protos.h (riscv_output_return): New.
25672 * config/riscv/riscv.c (struct machine_function): New naked_p field.
25673 (riscv_attribute_table, riscv_output_return),
25674 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25675 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25676 (riscv_compute_frame_info): Only compute frame->mask if not a naked
25678 (riscv_expand_prologue): Add early return for naked function.
25679 (riscv_expand_epilogue): Likewise.
25680 (riscv_function_ok_for_sibcall): Return false for naked function.
25681 (riscv_set_current_function): New.
25682 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25683 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25684 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25685 * doc/extend.texi (RISC-V Function Attributes): New.
25687 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
25689 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25690 check for 128-bit long double before checking TCmode.
25691 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25692 128-bit long doubles before checking TFmode or TCmode.
25693 (FLOAT128_IBM_P): Likewise.
25695 2018-01-10 Martin Sebor <msebor@redhat.com>
25697 PR tree-optimization/83671
25698 * builtins.c (c_strlen): Unconditionally return zero for the empty
25700 Use -Warray-bounds for warnings.
25701 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25702 for non-constant array indices with COMPONENT_REF, arrays of
25703 arrays, and pointers to arrays.
25704 (gimple_fold_builtin_strlen): Determine and set length range for
25705 non-constant character arrays.
25707 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
25709 PR middle-end/81897
25710 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25713 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
25715 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25717 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25720 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25721 VECTOR_MEM_ALTIVEC_OR_VSX_P.
25722 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25723 indexed_or_indirect_operand predicate.
25724 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25725 (*vsx_le_perm_load_v8hi): Likewise.
25726 (*vsx_le_perm_load_v16qi): Likewise.
25727 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25728 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25729 (*vsx_le_perm_store_v8hi): Likewise.
25730 (*vsx_le_perm_store_v16qi): Likewise.
25731 (eight unnamed splitters): Likewise.
25733 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25735 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25736 * config/rs6000/emmintrin.h: Likewise.
25737 * config/rs6000/mmintrin.h: Likewise.
25738 * config/rs6000/xmmintrin.h: Likewise.
25740 2018-01-10 David Malcolm <dmalcolm@redhat.com>
25743 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25745 * tree.c (tree_nop_conversion): Return true for location wrapper
25747 (maybe_wrap_with_location): New function.
25748 (selftest::check_strip_nops): New function.
25749 (selftest::test_location_wrappers): New function.
25750 (selftest::tree_c_tests): Call it.
25751 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25752 (maybe_wrap_with_location): New decl.
25753 (EXPR_LOCATION_WRAPPER_P): New macro.
25754 (location_wrapper_p): New inline function.
25755 (tree_strip_any_location_wrapper): New inline function.
25757 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
25760 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25761 stack_realign_offset for the largest alignment of stack slot
25763 (ix86_find_max_used_stack_alignment): New function.
25764 (ix86_finalize_stack_frame_flags): Use it. Set
25765 max_used_stack_alignment if we don't realign stack.
25766 * config/i386/i386.h (machine_function): Add
25767 max_used_stack_alignment.
25769 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25771 * config/arm/arm.opt (-mbranch-cost): New option.
25772 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25775 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25778 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25779 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25781 2018-01-10 Richard Biener <rguenther@suse.de>
25784 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25785 early out so it also covers the case where we have a non-NULL
25788 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25790 PR tree-optimization/83753
25791 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25792 for non-strided grouped accesses if the number of elements is 1.
25794 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25797 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25798 * i386.h (TARGET_USE_GATHER): Define.
25799 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25801 2018-01-10 Martin Liska <mliska@suse.cz>
25804 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25805 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25807 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25808 CLEANUP_NO_PARTITIONING is not set.
25810 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25812 * doc/rtl.texi: Remove documentation of (const ...) wrappers
25813 for vectors, as a partial revert of r254296.
25814 * rtl.h (const_vec_p): Delete.
25815 (const_vec_duplicate_p): Don't test for vector CONSTs.
25816 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25817 * expmed.c (make_tree): Likewise.
25820 * common.md (E, F): Use CONSTANT_P instead of checking for
25822 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25823 checking for CONST_VECTOR.
25825 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25827 PR middle-end/83575
25828 * predict.c (force_edge_cold): Handle in more sane way edges
25829 with no prediction.
25831 2018-01-09 Carl Love <cel@us.ibm.com>
25833 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25835 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25836 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25837 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25838 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
25839 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25840 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
25841 * config/rs6000/rs6000-protos.h: Add extern defition for
25842 rs6000_generate_float2_double_code.
25843 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25845 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25846 (float2_v2df): Add define_expand.
25848 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
25851 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25852 op_mode in the force_to_mode call.
25854 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25856 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25857 instead of checking each element individually.
25858 (aarch64_evpc_uzp): Likewise.
25859 (aarch64_evpc_zip): Likewise.
25860 (aarch64_evpc_ext): Likewise.
25861 (aarch64_evpc_rev): Likewise.
25862 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25863 instead of checking each element individually. Return true without
25865 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25866 whether all selected elements come from the same input, instead of
25867 checking each element individually. Remove calls to gen_rtx_REG,
25868 start_sequence and end_sequence and instead assert that no rtl is
25871 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25873 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25874 order of HIGH and CONST checks.
25876 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25878 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25879 if the destination isn't an SSA_NAME.
25881 2018-01-09 Richard Biener <rguenther@suse.de>
25883 PR tree-optimization/83668
25884 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25886 (canonicalize_loop_form): ... here, renamed from ...
25887 (canonicalize_loop_closed_ssa_form): ... this and amended to
25888 swap successor edges for loop exit blocks to make us use
25889 the RPO order we need for initial schedule generation.
25891 2018-01-09 Joseph Myers <joseph@codesourcery.com>
25893 PR tree-optimization/64811
25894 * match.pd: When optimizing comparisons with Inf, avoid
25895 introducing or losing exceptions from comparisons with NaN.
25897 2018-01-09 Martin Liska <mliska@suse.cz>
25900 * asan.c (shadow_mem_size): Add gcc_assert.
25902 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
25904 Don't save registers in main().
25907 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25908 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25909 * config/avr/avr.c (avr_set_current_function): Don't error if
25910 naked, OS_task or OS_main are specified at the same time.
25911 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25913 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25915 * common/config/avr/avr-common.c (avr_option_optimization_table):
25916 Switch on -mmain-is-OS_task for optimizing compilations.
25918 2018-01-09 Richard Biener <rguenther@suse.de>
25920 PR tree-optimization/83572
25921 * graphite.c: Include cfganal.h.
25922 (graphite_transform_loops): Connect infinite loops to exit
25923 and remove fake edges at the end.
25925 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25927 * ipa-inline.c (edge_badness): Revert accidental checkin.
25929 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25932 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25933 symbols; not inline clones.
25935 2018-01-09 Jakub Jelinek <jakub@redhat.com>
25938 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25939 hard registers. Formatting fixes.
25941 PR preprocessor/83722
25942 * gcc.c (try_generate_repro): Pass
25943 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25944 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25947 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
25948 Kito Cheng <kito.cheng@gmail.com>
25950 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25951 (riscv_leaf_function_p): Delete.
25952 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25954 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25956 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25958 (do_ifelse): New function.
25959 (do_isel): New function.
25960 (do_sub3): New function.
25961 (do_add3): New function.
25962 (do_load_mask_compare): New function.
25963 (do_overlap_load_compare): New function.
25964 (expand_compare_loop): New function.
25965 (expand_block_compare): Call expand_compare_loop() when appropriate.
25966 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25967 option description.
25968 (-mblock-compare-inline-loop-limit): New option.
25970 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25973 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25974 Reverse order of second and third operands in first alternative.
25975 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25976 of first and second elements in UNSPEC_VPERMR vector.
25977 (altivec_expand_vec_perm_le): Likewise.
25979 2018-01-08 Jeff Law <law@redhat.com>
25981 PR rtl-optimizatin/81308
25982 * tree-switch-conversion.c (cfg_altered): New file scoped static.
25983 (process_switch): If group_case_labels makes a change, then set
25985 (pass_convert_switch::execute): If a switch is converted, then
25986 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
25988 PR rtl-optimization/81308
25989 * recog.c (split_all_insns): Conditionally cleanup the CFG after
25992 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
25994 PR target/83663 - Revert r255946
25995 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25996 generation for cases where splatting a value is not useful.
25997 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25998 across a vec_duplicate and a paradoxical subreg forming a vector
25999 mode to a vec_concat.
26001 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26003 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26004 -march=armv8.3-a variants.
26005 * config/arm/t-multilib: Likewise.
26006 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
26008 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26010 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26012 (cceq_ior_compare_complement): Give it a name so I can use it, and
26013 change boolean_or_operator predicate to boolean_operator so it can
26014 be used to generate a crand.
26015 (eqne): New code iterator.
26016 (bd/bd_neg): New code_attrs.
26017 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26018 a single define_insn.
26019 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26020 decrement (bdnzt/bdnzf/bdzt/bdzf).
26021 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26022 with the new names of the branch decrement patterns, and added the
26023 names of the branch decrement conditional patterns.
26025 2018-01-08 Richard Biener <rguenther@suse.de>
26027 PR tree-optimization/83563
26028 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26031 2018-01-08 Richard Biener <rguenther@suse.de>
26033 PR middle-end/83713
26034 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26036 2018-01-08 Richard Biener <rguenther@suse.de>
26038 PR tree-optimization/83685
26039 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26040 references to abnormals.
26042 2018-01-08 Richard Biener <rguenther@suse.de>
26045 * dwarf2out.c (output_indirect_strings): Handle empty
26046 skeleton_debug_str_hash.
26047 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26049 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26051 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26052 (emit_store_direct): Likewise.
26053 (arc_trampoline_adjust_address): Likewise.
26054 (arc_asm_trampoline_template): New function.
26055 (arc_initialize_trampoline): Use asm_trampoline_template.
26056 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26057 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26058 * config/arc/arc.md (flush_icache): Delete pattern.
26060 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26062 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26063 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26066 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26069 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26070 by not USED_FOR_TARGET.
26071 (make_pass_resolve_sw_modes): Likewise.
26073 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26075 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26078 2018-01-08 Richard Biener <rguenther@suse.de>
26080 PR middle-end/83580
26081 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26083 2018-01-08 Richard Biener <rguenther@suse.de>
26085 PR middle-end/83517
26086 * match.pd ((t * 2) / 2) -> t): Add missing :c.
26088 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
26090 PR middle-end/81897
26091 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26092 basic blocks with a small number of successors.
26093 (convert_control_dep_chain_into_preds): Improve handling of
26095 (dump_predicates): Split apart into...
26096 (dump_pred_chain): ...here...
26097 (dump_pred_info): ...and here.
26098 (can_one_predicate_be_invalidated_p): Add debugging printfs.
26099 (can_chain_union_be_invalidated_p): Improve check for invalidation
26101 (uninit_uses_cannot_happen): Avoid unnecessary if
26102 convert_control_dep_chain_into_preds yielded nothing.
26104 2018-01-06 Martin Sebor <msebor@redhat.com>
26106 PR tree-optimization/83640
26107 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26108 subtracting negative offset from size.
26109 (builtin_access::overlap): Adjust offset bounds of the access to fall
26110 within the size of the object if possible.
26112 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
26114 PR rtl-optimization/83699
26115 * expmed.c (extract_bit_field_1): Restrict the vector usage of
26116 extract_bit_field_as_subreg to cases in which the extracted
26117 value is also a vector.
26119 * lra-constraints.c (process_alt_operands): Test for the equivalence
26120 substitutions when detecting a possible reload cycle.
26122 2018-01-06 Jakub Jelinek <jakub@redhat.com>
26125 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26126 by default if flag_selective_schedling{,2}. Formatting fixes.
26128 PR rtl-optimization/83682
26129 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26130 if it has non-VECTOR_MODE element mode.
26131 (vec_duplicate_p): Likewise.
26133 PR middle-end/83694
26134 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26135 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26137 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26140 * config/i386/i386-builtin.def
26141 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26142 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26143 Require also OPTION_MASK_ISA_AVX512F in addition to
26144 OPTION_MASK_ISA_GFNI.
26145 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26146 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26147 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26148 to OPTION_MASK_ISA_GFNI.
26149 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26150 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26151 OPTION_MASK_ISA_AVX512BW.
26152 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26153 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26154 addition to OPTION_MASK_ISA_GFNI.
26155 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26156 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26157 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26158 to OPTION_MASK_ISA_GFNI.
26159 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26160 a requirement for all ISAs rather than any of them with a few
26162 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26164 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26165 bitmasks to be enabled with 3 exceptions, instead of requiring any
26166 enabled ISA with lots of exceptions.
26167 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26168 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26169 Change avx512bw in isa attribute to avx512f.
26170 * config/i386/sgxintrin.h: Add license boilerplate.
26171 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
26172 to __AVX512F__ and __AVX512VL to __AVX512VL__.
26173 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26174 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26176 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26177 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26178 temporarily sse2 rather than sse if not enabled already.
26181 * config/i386/sse.md (VI248_VLBW): Rename to ...
26182 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
26183 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26184 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26185 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26186 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26187 mode iterator instead of VI248_VLBW.
26189 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
26191 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26192 (record_modified): Skip clobbers; add debug output.
26193 (param_change_prob): Use sreal frequencies.
26195 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26197 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26198 punt for user-aligned variables.
26200 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26202 * tree-chrec.c (chrec_contains_symbols): Return true for
26205 2018-01-05 Sudakshina Das <sudi.das@arm.com>
26208 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26209 of (x|y) == x for BICS pattern.
26211 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26213 PR tree-optimization/83605
26214 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26215 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26218 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
26220 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26221 * config/epiphany/rtems.h: New file.
26223 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26224 Uros Bizjak <ubizjak@gmail.com>
26227 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26228 QIreg_operand instead of register_operand predicate.
26229 * config/i386/i386.c (ix86_rop_should_change_byte_p,
26230 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26231 comments instead of -fmitigate[-_]rop.
26233 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26236 * cgraphunit.c (symbol_table::compile): Switch to text_section
26237 before calling assembly_start debug hook.
26238 * run-rtl-passes.c (run_rtl_passes): Likewise.
26241 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26243 * tree-vrp.c (extract_range_from_binary_expr_1): Check
26244 range_int_cst_p rather than !symbolic_range_p before calling
26245 extract_range_from_multiplicative_op_1.
26247 2018-01-04 Jeff Law <law@redhat.com>
26249 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26250 redundant test in assertion.
26252 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26254 * doc/rtl.texi: Document machine_mode wrapper classes.
26256 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26258 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26259 using tree_to_uhwi.
26261 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26263 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26264 the VEC_PERM_EXPR fold to fail.
26266 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26269 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26270 to switched_sections.
26272 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26275 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26276 test for d.testing.
26278 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
26281 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26282 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26284 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26287 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26288 is BLKmode and bitpos not zero or mode change is needed.
26290 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26293 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26296 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
26299 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26300 instead of MULT rtx. Update all corresponding splitters.
26301 (*saddl_se): Ditto.
26302 (*ssub<modesuffix>): Ditto.
26303 (*ssubl_se): Ditto.
26304 (*cmp_sadd_di): Update split patterns.
26305 (*cmp_sadd_si): Ditto.
26306 (*cmp_sadd_sidi): Ditto.
26307 (*cmp_ssub_di): Ditto.
26308 (*cmp_ssub_si): Ditto.
26309 (*cmp_ssub_sidi): Ditto.
26310 * config/alpha/predicates.md (const23_operand): New predicate.
26311 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26312 Look for ASHIFT, not MULT inner operand.
26313 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26315 2018-01-04 Martin Liska <mliska@suse.cz>
26317 PR gcov-profile/83669
26318 * gcov.c (output_intermediate_file): Add version to intermediate
26320 * doc/gcov.texi: Document new field 'version' in intermediate
26321 file format. Fix location of '-k' option of gcov command.
26323 2018-01-04 Martin Liska <mliska@suse.cz>
26326 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26328 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26330 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26332 2018-01-03 Martin Sebor <msebor@redhat.com>
26334 PR tree-optimization/83655
26335 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26336 checking calls with invalid arguments.
26338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26340 * tree-vect-stmts.c (vect_get_store_rhs): New function.
26341 (vectorizable_mask_load_store): Delete.
26342 (vectorizable_call): Return false for masked loads and stores.
26343 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
26344 instead of gimple_assign_rhs1.
26345 (vectorizable_load): Handle IFN_MASK_LOAD.
26346 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26350 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26352 (vectorizable_mask_load_store): ...here.
26353 (vectorizable_load): ...and here.
26355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26357 * tree-vect-stmts.c (vect_build_all_ones_mask)
26358 (vect_build_zero_merge_argument): New functions, split out from...
26359 (vectorizable_load): ...here.
26361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26363 * tree-vect-stmts.c (vect_check_store_rhs): New function,
26365 (vectorizable_mask_load_store): ...here.
26366 (vectorizable_store): ...and here.
26368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26370 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26372 (vectorizable_mask_load_store): ...here.
26374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26376 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26377 (vect_model_store_cost): Take a vec_load_store_type instead of a
26379 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26380 (vect_model_store_cost): Take a vec_load_store_type instead of a
26382 (vectorizable_mask_load_store): Update accordingly.
26383 (vectorizable_store): Likewise.
26384 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26388 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26389 IFN_MASK_LOAD calls here rather than...
26390 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26393 Alan Hayward <alan.hayward@arm.com>
26394 David Sherwood <david.sherwood@arm.com>
26396 * expmed.c (extract_bit_field_1): For vector extracts,
26397 fall back to extract_bit_field_as_subreg if vec_extract
26400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26401 Alan Hayward <alan.hayward@arm.com>
26402 David Sherwood <david.sherwood@arm.com>
26404 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26405 they are variable or constant sized.
26406 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26407 slots for constant-sized data.
26409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26410 Alan Hayward <alan.hayward@arm.com>
26411 David Sherwood <david.sherwood@arm.com>
26413 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26414 handling COND_EXPRs with boolean comparisons, try to find a better
26415 basis for the mask type than the boolean itself.
26417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26419 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26420 is calculated and how it can be overridden.
26421 * genmodes.c (max_bitsize_mode_any_mode): New variable.
26422 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26424 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26428 Alan Hayward <alan.hayward@arm.com>
26429 David Sherwood <david.sherwood@arm.com>
26431 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26432 Remove the mode argument.
26433 (aarch64_simd_valid_immediate): Remove the mode and inverse
26435 * config/aarch64/iterators.md (bitsize): New iterator.
26436 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26437 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26438 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26439 aarch64_simd_valid_immediate.
26440 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26441 (aarch64_reg_or_bic_imm): Likewise.
26442 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26443 with an insn_type enum and msl with a modifier_type enum.
26444 Replace element_width with a scalar_mode. Change the shift
26445 to unsigned int. Add constructors for scalar_float_mode and
26446 scalar_int_mode elements.
26447 (aarch64_vect_float_const_representable_p): Delete.
26448 (aarch64_can_const_movi_rtx_p)
26449 (aarch64_simd_scalar_immediate_valid_for_move)
26450 (aarch64_simd_make_constant): Update call to
26451 aarch64_simd_valid_immediate.
26452 (aarch64_advsimd_valid_immediate_hs): New function.
26453 (aarch64_advsimd_valid_immediate): Likewise.
26454 (aarch64_simd_valid_immediate): Remove mode and inverse
26455 arguments. Rewrite to use the above. Use const_vec_duplicate_p
26456 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26457 and aarch64_float_const_representable_p on the result.
26458 (aarch64_output_simd_mov_immediate): Remove mode argument.
26459 Update call to aarch64_simd_valid_immediate and use of
26460 simd_immediate_info.
26461 (aarch64_output_scalar_simd_mov_immediate): Update call
26464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26465 Alan Hayward <alan.hayward@arm.com>
26466 David Sherwood <david.sherwood@arm.com>
26468 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26469 (mode_nunits): Likewise CONST_MODE_NUNITS.
26470 * machmode.def (ADJUST_NUNITS): Document.
26471 * genmodes.c (mode_data::need_nunits_adj): New field.
26472 (blank_mode): Update accordingly.
26473 (adj_nunits): New variable.
26474 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26476 (emit_mode_size_inline): Set need_bytesize_adj for all modes
26477 listed in adj_nunits.
26478 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26479 listed in adj_nunits. Don't emit case statements for such modes.
26480 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26481 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
26482 nothing if adj_nunits is nonnull.
26483 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26484 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26485 (emit_mode_fbit): Update use of print_maybe_const_decl.
26486 (emit_move_size): Likewise. Treat the array as non-const
26488 (emit_mode_adjustments): Handle adj_nunits.
26490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26492 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26493 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26494 (VECTOR_MODES): Use it.
26495 (make_vector_modes): Take the prefix as an argument.
26497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26498 Alan Hayward <alan.hayward@arm.com>
26499 David Sherwood <david.sherwood@arm.com>
26501 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26502 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26503 for MODE_VECTOR_BOOL.
26504 * machmode.def (VECTOR_BOOL_MODE): Document.
26505 * genmodes.c (VECTOR_BOOL_MODE): New macro.
26506 (make_vector_bool_mode): New function.
26507 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26509 * lto-streamer-in.c (lto_input_mode_table): Likewise.
26510 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26512 * stor-layout.c (int_mode_for_mode): Likewise.
26513 * tree.c (build_vector_type_for_mode): Likewise.
26514 * varasm.c (output_constant_pool_2): Likewise.
26515 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26516 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
26517 for MODE_VECTOR_BOOL.
26518 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26519 of mode class checks.
26520 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26521 instead of a list of mode class checks.
26522 (expand_vector_scalar_condition): Likewise.
26523 (type_for_widest_vector_mode): Handle BImode as an inner mode.
26525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26526 Alan Hayward <alan.hayward@arm.com>
26527 David Sherwood <david.sherwood@arm.com>
26529 * machmode.h (mode_size): Change from unsigned short to
26531 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26532 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26533 or if measurement_type is not polynomial.
26534 (fixed_size_mode::includes_p): Check for constant-sized modes.
26535 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26536 return a poly_uint16 rather than an unsigned short.
26537 (emit_mode_size): Change the type of mode_size from unsigned short
26538 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
26539 (emit_mode_adjustments): Cope with polynomial vector sizes.
26540 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26542 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26544 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26545 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26546 * caller-save.c (setup_save_areas): Likewise.
26547 (replace_reg_with_saved_mem): Likewise.
26548 * calls.c (emit_library_call_value_1): Likewise.
26549 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26550 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26551 (gen_lowpart_for_combine): Likewise.
26552 * convert.c (convert_to_integer_1): Likewise.
26553 * cse.c (equiv_constant, cse_insn): Likewise.
26554 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26555 (cselib_subst_to_values): Likewise.
26556 * dce.c (word_dce_process_block): Likewise.
26557 * df-problems.c (df_word_lr_mark_ref): Likewise.
26558 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26559 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26560 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26561 (rtl_for_decl_location): Likewise.
26562 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26563 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26564 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26565 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26566 (expand_expr_real_1): Likewise.
26567 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26568 (pad_below): Likewise.
26569 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26570 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26571 * ira.c (get_subreg_tracking_sizes): Likewise.
26572 * ira-build.c (ira_create_allocno_objects): Likewise.
26573 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26574 (ira_sort_regnos_for_alter_reg): Likewise.
26575 * ira-costs.c (record_operand_costs): Likewise.
26576 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26577 (resolve_simple_move): Likewise.
26578 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26579 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26580 (lra_constraints): Likewise.
26581 (CONST_POOL_OK_P): Reject variable-sized modes.
26582 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26583 (add_pseudo_to_slot, lra_spill): Likewise.
26584 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26585 * optabs-query.c (get_best_extraction_insn): Likewise.
26586 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26587 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26588 (expand_mult_highpart, valid_multiword_target_p): Likewise.
26589 * recog.c (offsettable_address_addr_space_p): Likewise.
26590 * regcprop.c (maybe_mode_change): Likewise.
26591 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26592 * regrename.c (build_def_use): Likewise.
26593 * regstat.c (dump_reg_info): Likewise.
26594 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26595 (find_reloads, find_reloads_subreg_address): Likewise.
26596 * reload1.c (eliminate_regs_1): Likewise.
26597 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26598 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26599 (simplify_binary_operation_1, simplify_subreg): Likewise.
26600 * targhooks.c (default_function_arg_padding): Likewise.
26601 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26602 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26603 (verify_gimple_assign_ternary): Likewise.
26604 * tree-inline.c (estimate_move_cost): Likewise.
26605 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26606 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26607 (get_address_cost_ainc): Likewise.
26608 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26609 (vect_supportable_dr_alignment): Likewise.
26610 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26611 (vectorizable_reduction): Likewise.
26612 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26613 (vectorizable_operation, vectorizable_load): Likewise.
26614 * tree.c (build_same_sized_truth_vector_type): Likewise.
26615 * valtrack.c (cleanup_auto_inc_dec): Likewise.
26616 * var-tracking.c (emit_note_insn_var_location): Likewise.
26617 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26618 (ADDR_VEC_ALIGN): Likewise.
26620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26621 Alan Hayward <alan.hayward@arm.com>
26622 David Sherwood <david.sherwood@arm.com>
26624 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26626 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26627 or if measurement_type is polynomial.
26628 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26629 * combine.c (make_extraction): Likewise.
26630 * dse.c (find_shift_sequence): Likewise.
26631 * dwarf2out.c (mem_loc_descriptor): Likewise.
26632 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26633 (extract_bit_field, extract_low_bits): Likewise.
26634 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26635 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26636 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26637 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26638 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26639 * reload.c (find_reloads): Likewise.
26640 * reload1.c (alter_reg): Likewise.
26641 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26642 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26643 * tree-if-conv.c (predicate_mem_writes): Likewise.
26644 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26645 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26646 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26647 * valtrack.c (dead_debug_insert_temp): Likewise.
26648 * varasm.c (mergeable_constant_section): Likewise.
26649 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26652 Alan Hayward <alan.hayward@arm.com>
26653 David Sherwood <david.sherwood@arm.com>
26655 * expr.c (expand_assignment): Cope with polynomial mode sizes
26656 when assigning to a CONCAT.
26658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26659 Alan Hayward <alan.hayward@arm.com>
26660 David Sherwood <david.sherwood@arm.com>
26662 * machmode.h (mode_precision): Change from unsigned short to
26664 (mode_to_precision): Return a poly_uint16 rather than an unsigned
26666 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26667 or if measurement_type is not polynomial.
26668 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
26669 in which the mode is already known to be a scalar_int_mode.
26670 * genmodes.c (emit_mode_precision): Change the type of mode_precision
26671 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
26673 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26674 for GET_MODE_PRECISION.
26675 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26676 for GET_MODE_PRECISION.
26677 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26679 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26680 (expand_field_assignment, make_extraction): Likewise.
26681 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26682 (get_last_value): Likewise.
26683 * convert.c (convert_to_integer_1): Likewise.
26684 * cse.c (cse_insn): Likewise.
26685 * expr.c (expand_expr_real_1): Likewise.
26686 * lra-constraints.c (simplify_operand_subreg): Likewise.
26687 * optabs-query.c (can_atomic_load_p): Likewise.
26688 * optabs.c (expand_atomic_load): Likewise.
26689 (expand_atomic_store): Likewise.
26690 * ree.c (combine_reaching_defs): Likewise.
26691 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26692 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26693 * tree.h (type_has_mode_precision_p): Likewise.
26694 * ubsan.c (instrument_si_overflow): Likewise.
26696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26697 Alan Hayward <alan.hayward@arm.com>
26698 David Sherwood <david.sherwood@arm.com>
26700 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26701 polynomial numbers of units.
26702 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26703 (valid_vector_subparts_p): New function.
26704 (build_vector_type): Remove temporary shim and take the number
26705 of units as a poly_uint64 rather than an int.
26706 (build_opaque_vector_type): Take the number of units as a
26707 poly_uint64 rather than an int.
26708 * tree.c (build_vector_from_ctor): Handle polynomial
26709 TYPE_VECTOR_SUBPARTS.
26710 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26711 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26712 (build_vector_from_val): If the number of units is variable,
26713 use build_vec_duplicate_cst for constant operands and
26714 VEC_DUPLICATE_EXPR otherwise.
26715 (make_vector_type): Remove temporary is_constant ().
26716 (build_vector_type, build_opaque_vector_type): Take the number of
26717 units as a poly_uint64 rather than an int.
26718 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26720 * cfgexpand.c (expand_debug_expr): Likewise.
26721 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26722 (store_constructor, expand_expr_real_1): Likewise.
26723 (const_scalar_mask_from_tree): Likewise.
26724 * fold-const-call.c (fold_const_reduction): Likewise.
26725 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26726 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26727 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26728 (fold_relational_const): Likewise.
26729 (native_interpret_vector): Likewise. Change the size from an
26730 int to an unsigned int.
26731 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26732 TYPE_VECTOR_SUBPARTS.
26733 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26734 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26735 duplicating a non-constant operand into a variable-length vector.
26736 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26737 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26738 * ipa-icf.c (sem_variable::equals): Likewise.
26739 * match.pd: Likewise.
26740 * omp-simd-clone.c (simd_clone_subparts): Likewise.
26741 * print-tree.c (print_node): Likewise.
26742 * stor-layout.c (layout_type): Likewise.
26743 * targhooks.c (default_builtin_vectorization_cost): Likewise.
26744 * tree-cfg.c (verify_gimple_comparison): Likewise.
26745 (verify_gimple_assign_binary): Likewise.
26746 (verify_gimple_assign_ternary): Likewise.
26747 (verify_gimple_assign_single): Likewise.
26748 * tree-pretty-print.c (dump_generic_node): Likewise.
26749 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26750 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26751 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26752 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26753 (vect_shift_permute_load_chain): Likewise.
26754 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26755 (expand_vector_condition, optimize_vector_constructor): Likewise.
26756 (lower_vec_perm, get_compute_type): Likewise.
26757 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26758 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26759 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26760 (vect_recog_mask_conversion_pattern): Likewise.
26761 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26762 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26763 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26764 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26765 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26766 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26767 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26768 (supportable_widening_operation): Likewise.
26769 (supportable_narrowing_operation): Likewise.
26770 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26772 * varasm.c (output_constant): Likewise.
26774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26775 Alan Hayward <alan.hayward@arm.com>
26776 David Sherwood <david.sherwood@arm.com>
26778 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26779 so that both the length == 3 and length != 3 cases set up their
26780 own permute vectors. Add comments explaining why we know the
26781 number of elements is constant.
26782 (vect_permute_load_chain): Likewise.
26784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26785 Alan Hayward <alan.hayward@arm.com>
26786 David Sherwood <david.sherwood@arm.com>
26788 * machmode.h (mode_nunits): Change from unsigned char to
26790 (ONLY_FIXED_SIZE_MODES): New macro.
26791 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26792 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26793 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26795 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26796 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26797 or if measurement_type is not polynomial.
26798 * genmodes.c (ZERO_COEFFS): New macro.
26799 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26801 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26802 Use ZERO_COEFFS when emitting initializers.
26803 * data-streamer.h (bp_pack_poly_value): New function.
26804 (bp_unpack_poly_value): Likewise.
26805 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26806 for GET_MODE_NUNITS.
26807 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26808 for GET_MODE_NUNITS.
26809 * tree.c (make_vector_type): Remove temporary shim and make
26810 the real function take the number of units as a poly_uint64
26811 rather than an int.
26812 (build_vector_type_for_mode): Handle polynomial nunits.
26813 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26814 * emit-rtl.c (const_vec_series_p_1): Likewise.
26815 (gen_rtx_CONST_VECTOR): Likewise.
26816 * fold-const.c (test_vec_duplicate_folding): Likewise.
26817 * genrecog.c (validate_pattern): Likewise.
26818 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26819 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26820 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26821 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26822 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26823 * rtlanal.c (subreg_get_info): Likewise.
26824 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26825 (vect_grouped_load_supported): Likewise.
26826 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26827 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26828 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26829 (simplify_const_unary_operation, simplify_binary_operation_1)
26830 (simplify_const_binary_operation, simplify_ternary_operation)
26831 (test_vector_ops_duplicate, test_vector_ops): Likewise.
26832 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26833 instead of CONST_VECTOR_NUNITS.
26834 * varasm.c (output_constant_pool_2): Likewise.
26835 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26836 explicit-encoded elements in the XVEC for variable-length vectors.
26838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26840 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26843 Alan Hayward <alan.hayward@arm.com>
26844 David Sherwood <david.sherwood@arm.com>
26846 * coretypes.h (fixed_size_mode): Declare.
26847 (fixed_size_mode_pod): New typedef.
26848 * builtins.h (target_builtins::x_apply_args_mode)
26849 (target_builtins::x_apply_result_mode): Change type to
26850 fixed_size_mode_pod.
26851 * builtins.c (apply_args_size, apply_result_size, result_vector)
26852 (expand_builtin_apply_args_1, expand_builtin_apply)
26853 (expand_builtin_return): Update accordingly.
26855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26857 * cse.c (hash_rtx_cb): Hash only the encoded elements.
26858 * cselib.c (cselib_hash_rtx): Likewise.
26859 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26860 CONST_VECTOR encoding.
26862 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26863 Jeff Law <law@redhat.com>
26866 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26867 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26868 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26869 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26872 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26873 explicitly probe *sp in a noreturn function if there were any callee
26874 register saves or frame pointer is needed.
26876 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26879 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26880 BLKmode for ternary, binary or unary expressions.
26883 * var-tracking.c (delete_vta_debug_insn): New inline function.
26884 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26885 insns from get_insns () to NULL instead of each bb separately.
26886 Use delete_vta_debug_insn. No longer static.
26887 (vt_debug_insns_local, variable_tracking_main_1): Adjust
26888 delete_vta_debug_insns callers.
26889 * rtl.h (delete_vta_debug_insns): Declare.
26890 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26891 instead of variable_tracking_main.
26893 2018-01-03 Martin Sebor <msebor@redhat.com>
26895 PR tree-optimization/83603
26896 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26897 arguments past the endof the argument list in functions declared
26898 without a prototype.
26899 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26900 Avoid checking when arguments are null.
26902 2018-01-03 Martin Sebor <msebor@redhat.com>
26905 * doc/extend.texi (attribute const): Fix a typo.
26906 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26907 issuing -Wsuggest-attribute for void functions.
26909 2018-01-03 Martin Sebor <msebor@redhat.com>
26911 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26912 offset_int::from instead of wide_int::to_shwi.
26913 (maybe_diag_overlap): Remove assertion.
26914 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26915 * gimple-ssa-sprintf.c (format_directive): Same.
26916 (parse_directive): Same.
26917 (sprintf_dom_walker::compute_format_length): Same.
26918 (try_substitute_return_value): Same.
26920 2018-01-03 Jeff Law <law@redhat.com>
26922 PR middle-end/83654
26923 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26924 non-constant residual for zero at runtime and avoid probing in
26925 that case. Reorganize code for trailing problem to mirror handling
26928 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26930 PR tree-optimization/83501
26931 * tree-ssa-strlen.c (get_string_cst): New.
26932 (handle_char_store): Call get_string_cst.
26934 2018-01-03 Martin Liska <mliska@suse.cz>
26936 PR tree-optimization/83593
26937 * tree-ssa-strlen.c: Include tree-cfg.h.
26938 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26939 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26940 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26942 (strlen_dom_walker::before_dom_children): Call
26943 gimple_purge_dead_eh_edges. Dump tranformation with details
26945 (strlen_dom_walker::before_dom_children): Update call by adding
26946 new argument cleanup_eh.
26947 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26949 2018-01-03 Martin Liska <mliska@suse.cz>
26952 * cif-code.def (VARIADIC_THUNK): New enum value.
26953 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26956 2018-01-03 Jan Beulich <jbeulich@suse.com>
26958 * sse.md (mov<mode>_internal): Tighten condition for when to use
26959 vmovdqu<ssescalarsize> for TI and OI modes.
26961 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26963 Update copyright years.
26965 2018-01-03 Martin Liska <mliska@suse.cz>
26968 * ipa-visibility.c (function_and_variable_visibility): Skip
26969 functions with noipa attribure.
26971 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26973 * gcc.c (process_command): Update copyright notice dates.
26974 * gcov-dump.c (print_version): Ditto.
26975 * gcov.c (print_version): Ditto.
26976 * gcov-tool.c (print_version): Ditto.
26977 * gengtype.c (create_file): Ditto.
26978 * doc/cpp.texi: Bump @copying's copyright year.
26979 * doc/cppinternals.texi: Ditto.
26980 * doc/gcc.texi: Ditto.
26981 * doc/gccint.texi: Ditto.
26982 * doc/gcov.texi: Ditto.
26983 * doc/install.texi: Ditto.
26984 * doc/invoke.texi: Ditto.
26986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26988 * vector-builder.h (vector_builder::m_full_nelts): Change from
26989 unsigned int to poly_uint64.
26990 (vector_builder::full_nelts): Update prototype accordingly.
26991 (vector_builder::new_vector): Likewise.
26992 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26993 (vector_builder::operator ==): Likewise.
26994 (vector_builder::finalize): Likewise.
26995 * int-vector-builder.h (int_vector_builder::int_vector_builder):
26996 Take the number of elements as a poly_uint64 rather than an
26998 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26999 from unsigned int to poly_uint64.
27000 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27001 (vec_perm_indices::new_vector): Likewise.
27002 (vec_perm_indices::length): Likewise.
27003 (vec_perm_indices::nelts_per_input): Likewise.
27004 (vec_perm_indices::input_nelts): Likewise.
27005 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27006 number of elements per input as a poly_uint64 rather than an
27007 unsigned int. Use the original encoding for variable-length
27008 vectors, rather than clamping each individual element.
27009 For the second and subsequent elements in each pattern,
27010 clamp the step and base before clamping their sum.
27011 (vec_perm_indices::series_p): Handle polynomial element counts.
27012 (vec_perm_indices::all_in_range_p): Likewise.
27013 (vec_perm_indices_to_tree): Likewise.
27014 (vec_perm_indices_to_rtx): Likewise.
27015 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27016 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27017 (tree_vector_builder::new_binary_operation): Handle polynomial
27018 element counts. Return false if we need to know the number
27019 of elements at compile time.
27020 * fold-const.c (fold_vec_perm): Punt if the number of elements
27021 isn't known at compile time.
27023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27025 * vec-perm-indices.h (vec_perm_builder): Change element type
27026 from HOST_WIDE_INT to poly_int64.
27027 (vec_perm_indices::element_type): Update accordingly.
27028 (vec_perm_indices::clamp): Handle polynomial element_types.
27029 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27030 (vec_perm_indices::all_in_range_p): Likewise.
27031 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27033 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27034 polynomial vec_perm_indices element types.
27035 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27036 * fold-const.c (fold_vec_perm): Likewise.
27037 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27038 * tree-vect-generic.c (lower_vec_perm): Likewise.
27039 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27040 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27041 element type to HOST_WIDE_INT.
27043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27044 Alan Hayward <alan.hayward@arm.com>
27045 David Sherwood <david.sherwood@arm.com>
27047 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27048 rather than an int. Use plus_constant.
27049 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27050 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27053 Alan Hayward <alan.hayward@arm.com>
27054 David Sherwood <david.sherwood@arm.com>
27056 * calls.c (emit_call_1, expand_call): Change struct_value_size from
27057 a HOST_WIDE_INT to a poly_int64.
27059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27060 Alan Hayward <alan.hayward@arm.com>
27061 David Sherwood <david.sherwood@arm.com>
27063 * calls.c (load_register_parameters): Cope with polynomial
27064 mode sizes. Require a constant size for BLKmode parameters
27065 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
27066 forces a parameter to be padded at the lsb end in order to
27067 fill a complete number of words, require the parameter size
27068 to be ordered wrt UNITS_PER_WORD.
27070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27071 Alan Hayward <alan.hayward@arm.com>
27072 David Sherwood <david.sherwood@arm.com>
27074 * reload1.c (spill_stack_slot_width): Change element type
27075 from unsigned int to poly_uint64_pod.
27076 (alter_reg): Treat mode sizes as polynomial.
27078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27079 Alan Hayward <alan.hayward@arm.com>
27080 David Sherwood <david.sherwood@arm.com>
27082 * reload.c (complex_word_subreg_p): New function.
27083 (reload_inner_reg_of_subreg, push_reload): Use it.
27085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27086 Alan Hayward <alan.hayward@arm.com>
27087 David Sherwood <david.sherwood@arm.com>
27089 * lra-constraints.c (process_alt_operands): Reject matched
27090 operands whose sizes aren't ordered.
27091 (match_reload): Refer to this check here.
27093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27094 Alan Hayward <alan.hayward@arm.com>
27095 David Sherwood <david.sherwood@arm.com>
27097 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27098 that the mode size is in the set {1, 2, 4, 8, 16}.
27100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27101 Alan Hayward <alan.hayward@arm.com>
27102 David Sherwood <david.sherwood@arm.com>
27104 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27105 Use plus_constant instead of gen_rtx_PLUS.
27107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27108 Alan Hayward <alan.hayward@arm.com>
27109 David Sherwood <david.sherwood@arm.com>
27111 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27112 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27113 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27114 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27115 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27116 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27117 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27118 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27119 * config/i386/i386.c (ix86_push_rounding): ...this new function.
27120 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27122 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27123 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27124 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27125 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27126 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27127 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27128 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27129 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27130 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27131 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27133 * expr.c (emit_move_resolve_push): Treat the input and result
27134 of PUSH_ROUNDING as a poly_int64.
27135 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27136 (emit_push_insn): Likewise.
27137 * lra-eliminations.c (mark_not_eliminable): Likewise.
27138 * recog.c (push_operand): Likewise.
27139 * reload1.c (elimination_effects): Likewise.
27140 * rtlanal.c (nonzero_bits1): Likewise.
27141 * calls.c (store_one_arg): Likewise. Require the padding to be
27142 known at compile time.
27144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27145 Alan Hayward <alan.hayward@arm.com>
27146 David Sherwood <david.sherwood@arm.com>
27148 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27149 Use plus_constant instead of gen_rtx_PLUS.
27151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27152 Alan Hayward <alan.hayward@arm.com>
27153 David Sherwood <david.sherwood@arm.com>
27155 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27156 rather than an int.
27158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27159 Alan Hayward <alan.hayward@arm.com>
27160 David Sherwood <david.sherwood@arm.com>
27162 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27163 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27164 via stack temporaries. Treat the mode size as polynomial too.
27166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27167 Alan Hayward <alan.hayward@arm.com>
27168 David Sherwood <david.sherwood@arm.com>
27170 * expr.c (expand_expr_real_2): When handling conversions involving
27171 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27172 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
27173 as a poly_uint64 too.
27175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27176 Alan Hayward <alan.hayward@arm.com>
27177 David Sherwood <david.sherwood@arm.com>
27179 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27182 Alan Hayward <alan.hayward@arm.com>
27183 David Sherwood <david.sherwood@arm.com>
27185 * combine.c (can_change_dest_mode): Handle polynomial
27186 REGMODE_NATURAL_SIZE.
27187 * expmed.c (store_bit_field_1): Likewise.
27188 * expr.c (store_constructor): Likewise.
27189 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27190 and polynomial REGMODE_NATURAL_SIZE.
27191 (gen_lowpart_common): Likewise.
27192 * reginfo.c (record_subregs_of_mode): Likewise.
27193 * rtlanal.c (read_modify_subreg_p): Likewise.
27195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27196 Alan Hayward <alan.hayward@arm.com>
27197 David Sherwood <david.sherwood@arm.com>
27199 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27200 numbers of elements.
27202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27203 Alan Hayward <alan.hayward@arm.com>
27204 David Sherwood <david.sherwood@arm.com>
27206 * match.pd: Cope with polynomial numbers of vector elements.
27208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27209 Alan Hayward <alan.hayward@arm.com>
27210 David Sherwood <david.sherwood@arm.com>
27212 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27213 in a POINTER_PLUS_EXPR.
27215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27216 Alan Hayward <alan.hayward@arm.com>
27217 David Sherwood <david.sherwood@arm.com>
27219 * omp-simd-clone.c (simd_clone_subparts): New function.
27220 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27221 (ipa_simd_modify_function_body): Likewise.
27223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27224 Alan Hayward <alan.hayward@arm.com>
27225 David Sherwood <david.sherwood@arm.com>
27227 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27228 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27229 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27230 (expand_vector_condition, vector_element): Likewise.
27231 (subparts_gt): New function.
27232 (get_compute_type): Use subparts_gt.
27233 (count_type_subparts): Delete.
27234 (expand_vector_operations_1): Use subparts_gt instead of
27235 count_type_subparts.
27237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27238 Alan Hayward <alan.hayward@arm.com>
27239 David Sherwood <david.sherwood@arm.com>
27241 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27242 (vect_compile_time_alias): ...this new function. Do the calculation
27243 on poly_ints rather than trees.
27244 (vect_prune_runtime_alias_test_list): Update call accordingly.
27246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27247 Alan Hayward <alan.hayward@arm.com>
27248 David Sherwood <david.sherwood@arm.com>
27250 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27252 (vect_schedule_slp_instance): Likewise.
27254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27255 Alan Hayward <alan.hayward@arm.com>
27256 David Sherwood <david.sherwood@arm.com>
27258 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27259 constant and extern definitions for variable-length vectors.
27260 (vect_get_constant_vectors): Note that the number of units
27261 is known to be constant.
27263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27264 Alan Hayward <alan.hayward@arm.com>
27265 David Sherwood <david.sherwood@arm.com>
27267 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27268 of units as polynomial. Choose between WIDE and NARROW based
27271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27272 Alan Hayward <alan.hayward@arm.com>
27273 David Sherwood <david.sherwood@arm.com>
27275 * tree-vect-stmts.c (simd_clone_subparts): New function.
27276 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27279 Alan Hayward <alan.hayward@arm.com>
27280 David Sherwood <david.sherwood@arm.com>
27282 * tree-vect-stmts.c (vectorizable_call): Treat the number of
27283 vectors as polynomial. Use build_index_vector for
27284 IFN_GOMP_SIMD_LANE.
27286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27287 Alan Hayward <alan.hayward@arm.com>
27288 David Sherwood <david.sherwood@arm.com>
27290 * tree-vect-stmts.c (get_load_store_type): Treat the number of
27291 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27292 for variable-length vectors.
27293 (vectorizable_mask_load_store): Treat the number of units as
27294 polynomial, asserting that it is constant if the condition has
27295 already been enforced.
27296 (vectorizable_store, vectorizable_load): Likewise.
27298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27299 Alan Hayward <alan.hayward@arm.com>
27300 David Sherwood <david.sherwood@arm.com>
27302 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27303 of units as polynomial. Punt if we can't tell at compile time
27304 which vector contains the final result.
27306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27307 Alan Hayward <alan.hayward@arm.com>
27308 David Sherwood <david.sherwood@arm.com>
27310 * tree-vect-loop.c (vectorizable_induction): Treat the number
27311 of units as polynomial. Punt on SLP inductions. Use an integer
27312 VEC_SERIES_EXPR for variable-length integer reductions. Use a
27313 cast of such a series for variable-length floating-point
27316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27317 Alan Hayward <alan.hayward@arm.com>
27318 David Sherwood <david.sherwood@arm.com>
27320 * tree.h (build_index_vector): Declare.
27321 * tree.c (build_index_vector): New function.
27322 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27323 of units as polynomial, forcibly converting it to a constant if
27324 vectorizable_reduction has already enforced the condition.
27325 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
27326 to create a {1,2,3,...} vector.
27327 (vectorizable_reduction): Treat the number of units as polynomial.
27328 Choose vectype_in based on the largest scalar element size rather
27329 than the smallest number of units. Enforce the restrictions
27332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27333 Alan Hayward <alan.hayward@arm.com>
27334 David Sherwood <david.sherwood@arm.com>
27336 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27337 number of units as polynomial.
27339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27340 Alan Hayward <alan.hayward@arm.com>
27341 David Sherwood <david.sherwood@arm.com>
27343 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27344 * target.def (autovectorize_vector_sizes): Return the vector sizes
27345 by pointer, using vector_sizes rather than a bitmask.
27346 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27347 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27348 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27350 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27351 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27352 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27353 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27354 * omp-general.c (omp_max_vf): Likewise.
27355 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27356 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27357 * tree-vect-loop.c (vect_analyze_loop): Likewise.
27358 * tree-vect-slp.c (vect_slp_bb): Likewise.
27359 * doc/tm.texi: Regenerate.
27360 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27362 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27363 the vector size as a poly_uint64 rather than an unsigned int.
27364 (current_vector_size): Change from an unsigned int to a poly_uint64.
27365 (get_vectype_for_scalar_type): Update accordingly.
27366 * tree.h (build_truth_vector_type): Take the size and number of
27367 units as a poly_uint64 rather than an unsigned int.
27368 (build_vector_type): Add a temporary overload that takes
27369 the number of units as a poly_uint64 rather than an unsigned int.
27370 * tree.c (make_vector_type): Likewise.
27371 (build_truth_vector_type): Take the number of units as a poly_uint64
27372 rather than an unsigned int.
27374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27375 Alan Hayward <alan.hayward@arm.com>
27376 David Sherwood <david.sherwood@arm.com>
27378 * target.def (get_mask_mode): Take the number of units and length
27379 as poly_uint64s rather than unsigned ints.
27380 * targhooks.h (default_get_mask_mode): Update accordingly.
27381 * targhooks.c (default_get_mask_mode): Likewise.
27382 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27383 * doc/tm.texi: Regenerate.
27385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27386 Alan Hayward <alan.hayward@arm.com>
27387 David Sherwood <david.sherwood@arm.com>
27389 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27390 * omp-general.c (omp_max_vf): Likewise.
27391 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27392 (expand_omp_simd): Handle polynomial safelen.
27393 * omp-low.c (omplow_simd_context): Add a default constructor.
27394 (omplow_simd_context::max_vf): Change from int to poly_uint64.
27395 (lower_rec_simd_input_clauses): Update accordingly.
27396 (lower_rec_input_clauses): Likewise.
27398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27399 Alan Hayward <alan.hayward@arm.com>
27400 David Sherwood <david.sherwood@arm.com>
27402 * tree-vectorizer.h (vect_nunits_for_cost): New function.
27403 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27404 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27405 (vect_analyze_slp_cost): Likewise.
27406 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27407 (vect_model_load_cost): Likewise.
27409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27410 Alan Hayward <alan.hayward@arm.com>
27411 David Sherwood <david.sherwood@arm.com>
27413 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27414 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27415 from an unsigned int * to a poly_uint64_pod *.
27416 (calculate_unrolling_factor): New function.
27417 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
27419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27420 Alan Hayward <alan.hayward@arm.com>
27421 David Sherwood <david.sherwood@arm.com>
27423 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27424 from an unsigned int to a poly_uint64.
27425 (_loop_vec_info::slp_unrolling_factor): Likewise.
27426 (_loop_vec_info::vectorization_factor): Change from an int
27428 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27429 (vect_get_num_vectors): New function.
27430 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27431 (vect_get_num_copies): Use vect_get_num_vectors.
27432 (vect_analyze_data_ref_dependences): Change max_vf from an int *
27433 to an unsigned int *.
27434 (vect_analyze_data_refs): Change min_vf from an int * to a
27436 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27437 than an unsigned HOST_WIDE_INT.
27438 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27439 (vect_analyze_data_ref_dependence): Change max_vf from an int *
27440 to an unsigned int *.
27441 (vect_analyze_data_ref_dependences): Likewise.
27442 (vect_compute_data_ref_alignment): Handle polynomial vf.
27443 (vect_enhance_data_refs_alignment): Likewise.
27444 (vect_prune_runtime_alias_test_list): Likewise.
27445 (vect_shift_permute_load_chain): Likewise.
27446 (vect_supportable_dr_alignment): Likewise.
27447 (dependence_distance_ge_vf): Take the vectorization factor as a
27448 poly_uint64 rather than an unsigned HOST_WIDE_INT.
27449 (vect_analyze_data_refs): Change min_vf from an int * to a
27451 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27452 vfm1 as a poly_uint64 rather than an int. Make the same change
27453 for the returned bound_scalar.
27454 (vect_gen_vector_loop_niters): Handle polynomial vf.
27455 (vect_do_peeling): Likewise. Update call to
27456 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27457 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27459 * tree-vect-loop.c (vect_determine_vectorization_factor)
27460 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27461 (vect_get_known_peeling_cost): Likewise.
27462 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27463 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27464 (vect_transform_loop): Likewise. Use the lowest possible VF when
27465 updating the upper bounds of the loop.
27466 (vect_min_worthwhile_factor): Make static. Return an unsigned int
27467 rather than an int.
27468 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27469 polynomial unroll factors.
27470 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27471 (vect_make_slp_decision): Likewise.
27472 (vect_supported_load_permutation_p): Likewise, and polynomial
27474 (vect_analyze_slp_cost): Handle polynomial vf.
27475 (vect_slp_analyze_node_operations): Likewise.
27476 (vect_slp_analyze_bb_1): Likewise.
27477 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27478 than an unsigned HOST_WIDE_INT.
27479 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27480 (vectorizable_load): Handle polynomial vf.
27481 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27483 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27486 Alan Hayward <alan.hayward@arm.com>
27487 David Sherwood <david.sherwood@arm.com>
27489 * match.pd: Handle bit operations involving three constants
27490 and try to fold one pair.
27492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27494 * tree-vect-loop-manip.c: Include gimple-fold.h.
27495 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27496 niters_maybe_zero parameters. Handle other cases besides a step of 1.
27497 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27498 Add a path that uses a step of VF instead of 1, but disable it
27500 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27501 and niters_no_overflow parameters. Update calls to
27502 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27503 Create a new SSA name if the latter choses to use a ste other
27504 than zero, and return it via niters_vector_mult_vf_var.
27505 * tree-vect-loop.c (vect_transform_loop): Update calls to
27506 vect_do_peeling, vect_gen_vector_loop_niters and
27507 slpeel_make_loop_iterate_ntimes.
27508 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27509 (vect_gen_vector_loop_niters): Update declarations after above changes.
27511 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
27513 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27514 128-bit round to integer instructions.
27515 (ceil<mode>2): Likewise.
27516 (btrunc<mode>2): Likewise.
27517 (round<mode>2): Likewise.
27519 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27521 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27522 unaligned VSX load/store on P8/P9.
27523 (expand_block_clear): Allow the use of unaligned VSX
27524 load/store on P8/P9.
27526 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27528 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27530 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27531 swap associated with both a load and a store.
27533 2018-01-02 Andrew Waterman <andrew@sifive.com>
27535 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27536 * config/riscv/riscv.md (clear_cache): Use it.
27538 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
27540 * web.c: Remove out-of-date comment.
27542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27544 * expr.c (fixup_args_size_notes): Check that any existing
27545 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27546 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27547 (emit_single_push_insn): ...here.
27549 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27551 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27552 (const_vector_encoded_nelts): New function.
27553 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27554 (const_vector_int_elt, const_vector_elt): Declare.
27555 * emit-rtl.c (const_vector_int_elt_1): New function.
27556 (const_vector_elt): Likewise.
27557 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27558 of CONST_VECTOR_ELT.
27560 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27562 * expr.c: Include rtx-vector-builder.h.
27563 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27564 directly on the tree encoding.
27565 (const_vector_from_tree): Likewise.
27566 * optabs.c: Include rtx-vector-builder.h.
27567 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27568 sequence of "u" values.
27569 * vec-perm-indices.c: Include rtx-vector-builder.h.
27570 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27571 directly on the vec_perm_indices encoding.
27573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27575 * doc/rtl.texi (const_vector): Describe new encoding scheme.
27576 * Makefile.in (OBJS): Add rtx-vector-builder.o.
27577 * rtx-vector-builder.h: New file.
27578 * rtx-vector-builder.c: Likewise.
27579 * rtl.h (rtx_def::u2): Add a const_vector field.
27580 (CONST_VECTOR_NPATTERNS): New macro.
27581 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27582 (CONST_VECTOR_DUPLICATE_P): Likewise.
27583 (CONST_VECTOR_STEPPED_P): Likewise.
27584 (CONST_VECTOR_ENCODED_ELT): Likewise.
27585 (const_vec_duplicate_p): Check for a duplicated vector encoding.
27586 (unwrap_const_vec_duplicate): Likewise.
27587 (const_vec_series_p): Check for a non-duplicated vector encoding.
27588 Say that the function only returns true for integer vectors.
27589 * emit-rtl.c: Include rtx-vector-builder.h.
27590 (gen_const_vec_duplicate_1): Delete.
27591 (gen_const_vector): Call gen_const_vec_duplicate instead of
27592 gen_const_vec_duplicate_1.
27593 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27594 (gen_const_vec_duplicate): Use rtx_vector_builder.
27595 (gen_const_vec_series): Likewise.
27596 (gen_rtx_CONST_VECTOR): Likewise.
27597 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27598 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27599 Build a new vector rather than modifying a CONST_VECTOR in-place.
27600 (handle_special_swappables): Update call accordingly.
27601 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27602 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27603 Build a new vector rather than modifying a CONST_VECTOR in-place.
27604 (handle_special_swappables): Update call accordingly.
27606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27608 * simplify-rtx.c (simplify_const_binary_operation): Use
27609 CONST_VECTOR_ELT instead of XVECEXP.
27611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27613 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27614 the selector elements to be different from the data elements
27615 if the selector is a VECTOR_CST.
27616 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27617 ssizetype for the selector.
27619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27621 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27622 before testing each element individually.
27623 * tree-vect-generic.c (lower_vec_perm): Likewise.
27625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27627 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27628 * selftest-run-tests.c (selftest::run_tests): Call it.
27629 * vector-builder.h (vector_builder::operator ==): New function.
27630 (vector_builder::operator !=): Likewise.
27631 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27632 (vec_perm_indices::all_from_input_p): New function.
27633 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27634 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27635 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27636 instead of reading the VECTOR_CST directly. Detect whether both
27637 vector inputs are the same before constructing the vec_perm_indices,
27638 and update the number of inputs argument accordingly. Use the
27639 utility functions added above. Only construct sel2 if we need to.
27641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27643 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27644 the broadcast of the low byte.
27645 (expand_mult_highpart): Use an explicit encoding for the permutes.
27646 * optabs-query.c (can_mult_highpart_p): Likewise.
27647 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27648 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27649 (vectorizable_bswap): Likewise.
27650 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27651 explicit encoding for the power-of-2 permutes.
27652 (vect_permute_store_chain): Likewise.
27653 (vect_grouped_load_supported): Likewise.
27654 (vect_permute_load_chain): Likewise.
27656 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27658 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27659 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27660 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27661 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27662 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27663 (vect_gen_perm_mask_any): Likewise.
27665 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27667 * int-vector-builder.h: New file.
27668 * vec-perm-indices.h: Include int-vector-builder.h.
27669 (vec_perm_indices): Redefine as an int_vector_builder.
27670 (auto_vec_perm_indices): Delete.
27671 (vec_perm_builder): Redefine as a stand-alone class.
27672 (vec_perm_indices::vec_perm_indices): New function.
27673 (vec_perm_indices::clamp): Likewise.
27674 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27675 (vec_perm_indices::new_vector): New function.
27676 (vec_perm_indices::new_expanded_vector): Update for new
27677 vec_perm_indices class.
27678 (vec_perm_indices::rotate_inputs): New function.
27679 (vec_perm_indices::all_in_range_p): Operate directly on the
27680 encoded form, without computing elided elements.
27681 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27682 encoding. Update for new vec_perm_indices class.
27683 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27684 the given vec_perm_builder.
27685 (expand_vec_perm_var): Update vec_perm_builder constructor.
27686 (expand_mult_highpart): Use vec_perm_builder instead of
27687 auto_vec_perm_indices.
27688 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27689 vec_perm_indices instead of auto_vec_perm_indices. Use a single
27690 or double series encoding as appropriate.
27691 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27692 vec_perm_indices instead of auto_vec_perm_indices.
27693 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27694 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27695 (vect_permute_store_chain): Likewise.
27696 (vect_grouped_load_supported): Likewise.
27697 (vect_permute_load_chain): Likewise.
27698 (vect_shift_permute_load_chain): Likewise.
27699 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27700 (vect_transform_slp_perm_load): Likewise.
27701 (vect_schedule_slp_instance): Likewise.
27702 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27703 (vectorizable_mask_load_store): Likewise.
27704 (vectorizable_bswap): Likewise.
27705 (vectorizable_store): Likewise.
27706 (vectorizable_load): Likewise.
27707 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27708 vec_perm_indices instead of auto_vec_perm_indices. Use
27709 tree_to_vec_perm_builder to read the vector from a tree.
27710 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27711 vec_perm_builder instead of a vec_perm_indices.
27712 (have_whole_vector_shift): Use vec_perm_builder and
27713 vec_perm_indices instead of auto_vec_perm_indices. Leave the
27714 truncation to calc_vec_perm_mask_for_shift.
27715 (vect_create_epilog_for_reduction): Likewise.
27716 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27717 from auto_vec_perm_indices to vec_perm_indices.
27718 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27719 instead of changing individual elements.
27720 (aarch64_vectorize_vec_perm_const): Use new_vector to install
27721 the vector in d.perm.
27722 * config/arm/arm.c (expand_vec_perm_d::perm): Change
27723 from auto_vec_perm_indices to vec_perm_indices.
27724 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27725 instead of changing individual elements.
27726 (arm_vectorize_vec_perm_const): Use new_vector to install
27727 the vector in d.perm.
27728 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27729 Update vec_perm_builder constructor.
27730 (rs6000_expand_interleave): Likewise.
27731 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27732 (rs6000_expand_interleave): Likewise.
27734 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27736 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27737 to qimode could truncate the indices.
27738 * optabs.c (expand_vec_perm_var): Likewise.
27740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27742 * Makefile.in (OBJS): Add vec-perm-indices.o.
27743 * vec-perm-indices.h: New file.
27744 * vec-perm-indices.c: Likewise.
27745 * target.h (vec_perm_indices): Replace with a forward class
27747 (auto_vec_perm_indices): Move to vec-perm-indices.h.
27748 * optabs.h: Include vec-perm-indices.h.
27749 (expand_vec_perm): Delete.
27750 (selector_fits_mode_p, expand_vec_perm_var): Declare.
27751 (expand_vec_perm_const): Declare.
27752 * target.def (vec_perm_const_ok): Replace with...
27753 (vec_perm_const): ...this new hook.
27754 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27755 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27756 * doc/tm.texi: Regenerate.
27757 * optabs.def (vec_perm_const): Delete.
27758 * doc/md.texi (vec_perm_const): Likewise.
27759 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27760 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27761 expand_vec_perm for constant permutation vectors. Assert that
27762 the mode of variable permutation vectors is the integer equivalent
27763 of the mode that is being permuted.
27764 * optabs-query.h (selector_fits_mode_p): Declare.
27765 * optabs-query.c: Include vec-perm-indices.h.
27766 (selector_fits_mode_p): New function.
27767 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27768 is defined, instead of checking whether the vec_perm_const_optab
27769 exists. Use targetm.vectorize.vec_perm_const instead of
27770 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27771 fit in the vector mode before using a variable permute.
27772 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27773 vec_perm_indices instead of an rtx.
27774 (expand_vec_perm): Replace with...
27775 (expand_vec_perm_const): ...this new function. Take the selector
27776 as a vec_perm_indices rather than an rtx. Also take the mode of
27777 the selector. Update call to shift_amt_for_vec_perm_mask.
27778 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27779 Use vec_perm_indices::new_expanded_vector to expand the original
27780 selector into bytes. Check whether the indices fit in the vector
27781 mode before using a variable permute.
27782 (expand_vec_perm_var): Make global.
27783 (expand_mult_highpart): Use expand_vec_perm_const.
27784 * fold-const.c: Includes vec-perm-indices.h.
27785 * tree-ssa-forwprop.c: Likewise.
27786 * tree-vect-data-refs.c: Likewise.
27787 * tree-vect-generic.c: Likewise.
27788 * tree-vect-loop.c: Likewise.
27789 * tree-vect-slp.c: Likewise.
27790 * tree-vect-stmts.c: Likewise.
27791 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27793 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27794 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27795 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27796 (aarch64_vectorize_vec_perm_const): ...this new function.
27797 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27798 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27799 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27800 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27801 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27802 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27803 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27805 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
27806 check for NEON modes.
27807 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27808 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27809 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27810 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27812 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
27813 the old VEC_PERM_CONST conditions.
27814 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27815 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27816 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27817 (ia64_vectorize_vec_perm_const_ok): Merge into...
27818 (ia64_vectorize_vec_perm_const): ...this new function.
27819 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27820 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27821 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27822 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27823 * config/mips/mips.c (mips_expand_vec_perm_const)
27824 (mips_vectorize_vec_perm_const_ok): Merge into...
27825 (mips_vectorize_vec_perm_const): ...this new function.
27826 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27827 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27828 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27829 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27830 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27831 (rs6000_expand_vec_perm_const): Delete.
27832 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27834 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27835 (altivec_expand_vec_perm_const_le): Take each operand individually.
27836 Operate on constant selectors rather than rtxes.
27837 (altivec_expand_vec_perm_const): Likewise. Update call to
27838 altivec_expand_vec_perm_const_le.
27839 (rs6000_expand_vec_perm_const): Delete.
27840 (rs6000_vectorize_vec_perm_const_ok): Delete.
27841 (rs6000_vectorize_vec_perm_const): New function.
27842 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27843 an element count and rtx array.
27844 (rs6000_expand_extract_even): Update call accordingly.
27845 (rs6000_expand_interleave): Likewise.
27846 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27847 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27848 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27849 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27850 (rs6000_expand_vec_perm_const): Delete.
27851 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27852 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27853 (altivec_expand_vec_perm_const_le): Take each operand individually.
27854 Operate on constant selectors rather than rtxes.
27855 (altivec_expand_vec_perm_const): Likewise. Update call to
27856 altivec_expand_vec_perm_const_le.
27857 (rs6000_expand_vec_perm_const): Delete.
27858 (rs6000_vectorize_vec_perm_const_ok): Delete.
27859 (rs6000_vectorize_vec_perm_const): New function. Remove stray
27860 reference to the SPE evmerge intructions.
27861 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27862 an element count and rtx array.
27863 (rs6000_expand_extract_even): Update call accordingly.
27864 (rs6000_expand_interleave): Likewise.
27865 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27866 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27868 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27872 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27873 vector mode and that that mode matches the mode of the data
27875 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27876 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
27877 directly using expand_vec_perm_1 when forcing selectors into
27879 (expand_vec_perm_var): New function, split out from expand_vec_perm.
27881 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27883 * optabs-query.h (can_vec_perm_p): Delete.
27884 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27885 * optabs-query.c (can_vec_perm_p): Split into...
27886 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27887 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27888 particular selector is valid.
27889 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27890 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27891 (vect_grouped_load_supported): Likewise.
27892 (vect_shift_permute_load_chain): Likewise.
27893 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27894 (vect_transform_slp_perm_load): Likewise.
27895 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27896 (vectorizable_bswap): Likewise.
27897 (vect_gen_perm_mask_checked): Likewise.
27898 * fold-const.c (fold_ternary_loc): Likewise. Don't take
27899 implementations of variable permutation vectors into account
27900 when deciding which selector to use.
27901 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27902 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27903 with a false third argument.
27904 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27905 to test whether the constant selector is valid and can_vec_perm_var_p
27906 to test whether a variable selector is valid.
27908 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27910 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27911 * optabs-query.c (can_vec_perm_p): Likewise.
27912 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27913 instead of vec_perm_indices.
27914 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27915 (vect_gen_perm_mask_checked): Likewise,
27916 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27917 (vect_gen_perm_mask_checked): Likewise,
27919 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27921 * optabs-query.h (qimode_for_vec_perm): Declare.
27922 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27923 (qimode_for_vec_perm): ...this new function.
27924 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27926 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27928 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27929 does not have a conditional at the top.
27931 2018-01-02 Richard Biener <rguenther@suse.de>
27933 * ipa-inline.c (big_speedup_p): Fix expression.
27935 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27938 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27941 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27945 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27946 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27947 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27948 cond_taken_branch_cost 3->4.
27950 2018-01-01 Jakub Jelinek <jakub@redhat.com>
27952 PR tree-optimization/83581
27953 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27954 TODO_cleanup_cfg if any changes have been made.
27956 PR middle-end/83608
27957 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27958 convert_modes if target mode has the right side, but different mode
27961 PR middle-end/83609
27962 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27963 last argument when extracting from CONCAT. If either from_real or
27964 from_imag is NULL, use expansion through memory. If result is not
27965 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27966 the parts directly to inner mode, if even that fails, use expansion
27969 PR middle-end/83623
27970 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27971 check for bswap in mode rather than HImode and use that in expand_unop
27974 Copyright (C) 2018 Free Software Foundation, Inc.
27976 Copying and distribution of this file, with or without modification,
27977 are permitted in any medium without royalty provided the copyright
27978 notice and this notice are preserved.